Scrabble Solver

For nearly two years, I have been trying to get some acknowledgement for what I believe to be record play and game in the variant SuperScrabble, as offered on the website www.thepixiepit.co.uk (and also available for "real life" play, face to face).

An image of the game can be seen here:

http://www.geocities.com/rexbickers/FourSuperBingosMay27.jpg

As you will note...

... the highest scoring play was 6259 for ANTHROPOMORPHIZATIONS (43 X 144 = 6192 plus 50 for the bingo = 6242), OM (12), TO (2) and IT (3) The rack contained AROITIS and the board contained the words NTH, OP, MORPH, ZA and ON.

My "opponent" (Christine) had the second-highest scoring play: 6110 for COMPARTMENTALIZATIONS (42 x 144 = 6048 plus 50 for the bingo = 6098), UP (7), ET (3) and ET(2).

Higher scoring individual plays could be made for either of these words, but I assert that it is not possible to achieve a higher scoring combined game score.

Best wishes,

Rex Bickers
Floyds Knobs, Indiana (USA)

wow !!

My best result in Slovak version with regular game 15 x 15 is 4459 points. I hope, this is the best world result with 99 Scrabble stones:

No. Supply Pos.Words Points Cumul.
1. VYHŔŇAP 8E VYHŔŇA 56 56
2. PONEVŠÍ 8C POVYHŔŇA 32 88
3. NEVŠÍMI 8A NEPOVYHŔŇAVŠÍMI 468+50 518 606
4. OPRAVIL J1 OPRAVILA 16+50 66 672
5. NSRÁD?O 1I NO 2 674
6. SRÁD?OV 1I NOS 3 677
7. RÁD?OVK F1 RÁD(I)OVKY 25+50 75 752
8. EAMÄSŽR 1F RE 2 754
9. AMÄSŽRA 1E ARE 3 757
10. MÄSŽRAÉ 5C MÄSOŽRAVÉ 116+50 166 923
11. ZAJOIAĎ C1 ZAJOM 22 945
12. IMAĎARÓ N8 MI 3 948
13. MAĎARÓN N1 MAĎARÓNMI 94+50 144 1092
14. BEFBŤAI 1A BEZFAREBNOSŤAMI 1431+50 1481 2573
15. OSTEĽTE C8 POSTEĽTE 32+50 82 2655
16. LODA?OM G8 HLODA(N)OM 15+50 65 2720
17. KOČTÝSL 15E KOM 5 2725
18. ČTÝSLED F14 ČO,ČO 32 2757
19. TÝSLEDO 10F TO 4 2761
20. ÝSLEDOK K8 VÝSLEDOK 30+50 80 2841
21. NIPTGEJ 15I NIK 4 2845
22. PTGEJÔN 14J PO,PI 14 2859
23. TGEJÔNE 14J POT 4 2863
24. GEJÔNEX M13 GEJ, POTE 29 2892
25. ÔNEXUUÚ 10J ÔS 25 2917
26. NEXUUÚC 15A NEEXKOMUNIKUJÚC,TU 1492+50 1542 4459

You can see it on the official site of Slovak Scrabble Company http://hramescrabble.sk/ Then click to icon "Zo života"

Have a Nice day,

Ivan Nosál
Bratislava, Slovakia

As far as I can tell, sesquioxidizing is not a word. Never.

Why?

Well, I assume the idea is contracted from oxidizing/oxidize. To oxidize something, you need it to combine (e.g.: Mg + O = MgO) with oxygen (1 definition) or get the element to lose valence electrons (another definition).

Now, a sesquioxide is a compound where there are 2 radicals (elements other than O) and 3 O's. e.g. Al2O3 is a sesquioxide.

Now as far as I know, you can only get a compound to react with something else, not combine. And then it won't be a sesquioxide - as it is strictly a compound with 2 radicals and 3 oxygen atoms.

And oxidize as in definition 2 can only relate to that - it cannot have any suffix or prefix, as in sesquioxidize (worded badly). You can't sesquioxidize something in the same way. You either oxidize it or leave it.

And if you can't sesquioxidize something, then you can't be sesquioxidizing. Simple (okay, maybe not, but if you know what I am talking about, it is).

If you happen to look up sesquioxidize/ing on Wikipedia, it defines it as creating a sesquioxide, that is wrong. That is called synthesizing a sesquioxide, as the term can only go under a strict molecule structure, and can not be altered from X2O3. whereas an oxide just needs and oxygen atom from a reaction.

So there you go. That is why lexicographers ignore sesquioxidizing.

Someone please correct me if I am wrong with anything.

--Just expanding, in paragra2 I mean that I'm pretty sure you cant combine a compound with a compound. e.g. Comp. + Comp. = Comp. + Comp. (and more, if applicable)

Actually, I may be wrong. One of the products may be a sesquioxide? But yet again, it's formed through reaction, not combination (where I think an oxide can only be formed under combination (e.g.: Mg + O = MgO)and not reaction. Where I'm implying it's only possible to make a sesquioxide by reaction.

Who knows?

-sorry, I meant "(where I think oxidization can only work through combination)"

"Requalified" is not a word in TWL.

I think "Requalified" actually is in TWL. What is your source?

Hey PlasticGuitar,

I'm impressed with your knowledge of Chemistry, but the mere fact that a thing isn't possible in reality doesn't mean the word representing it doesn't exist and can't be used to express a valid thought or give actual information.

What about your own sentence, "If you can't sesquioxidize something, then you can't be sesquioxidizing." ?

By your reasoning words like UNICORN shouldn't be in the lexicon either.

IMHO

Cheers!
jumpinmf

Unlike for many a word in Scrabble dictionaries, the Oxford English dictionary does have an entry for sesquioxide, which includes the sentence: "Hence sesquioxidation, conversion into a sesquioxide; sesquioxidic a.; sesquioxidized pa. pple., converted into a sesquioxide." Followed by some quotes from the 19th and 20th century. If "sesquioxidation" and "sesquioxidized" are considered English, "sesquioxidizing" should be too ;-)

With respect to Josepha Heifetz Byrne's 2069 point word play: this probably included many more words from her 1976 "Dictionary of Unusual, Obscure, and Preposterous Words" (in which she also introduced sesquioxidizing) that you can't find in SOWPODS.

Arian Smit

@asmit- Nice find. What I was trying to say in that half-novel above (I have a tendency to over-complicate things)is that I thought sesquioxidation was the same as oxidation, oxidizing, etc. but with 2 other elements involved. But now that I word it like that, and read your post, it becomes all clear, and sesquioxidizing does make sense.

Oxidation is actually loss of electrons in a reduction-oxidation reaction, but pretty much applies to creation of oxides.

Yes, it's all so clear now! But why isn't it in the lexicon yet? Neither in 'branes', which is short for membranes (relating to the String Theory, which is a bit too complex to go into here.)

Also, I found a 2037 point play using SESQUIOXIDIZING here (note: it's a .pdf file):

http://digitalcommons.butler.edu/cgi/viewcontent.cgi?article=1771&context=wordways

...which just happens to be done by Josepha Heifetz. (I think it's the same one, at least). However, the words aren't valid according to a book I have called "Collins Scrabble Words - The Official Word List" (which - if anyone's curious - has over 250,000 words).

Also, asmit, knowing how much of a whiz you are at the puzzles, how many points do you think you could get, if you were given a rack of DGQSZXI in the puzzles? =P

--PlasticGuitar

P.S. REQUALIFIED is a word in the TWL06. Check it using the solver - set the dictionary to TWL, type QUALIFIED and put in a rack of RE and there you go.

Wait, what? I come back to check the puzzles, and I see "DGQSZXI"? Oh man, that made my day.

--PlasticGuitar

@PlasticGuitar – Hey, I hadn’t seen this pdf yet. Nice digging and fun to read.

That must have been so much work, back in 1972, without the use of computers nor a unified list of words. Imagine the number of dictionaries leaved through and the number of scores hand-calculated. Though "W.E.R. Bone", as a group of Cambridge mathematicians, probably had access to some mainframe. Pity then that their benzoxycamphors was accepted and not Heifetz’s sesquioxidizing (despite Webster’s & OED’s knoledge of it). Apparently benzoxycamphor was in Webster’s Second (1934) (Ross Eckler mentions this himself in his "Word ways: the journal of recreational linguistics”, see http://books.google.com/books?id=m4ksAAAAMAAJ&q=benzoxycamphor&dq=benzoxycamphor&client=safari ). It also shows up in a Webster’s wordlist [http://0x80.org/wordlist/webster-dictionary.txt ]. This list (204125 2-15 letter words without plurals or conjugations) also has Bone's “squandermaniac”, which you can’t find anywhere ele. Jinnyrickshaw isn’t even in here though; it must be the only way you can't spell jinricksha.

Taking up Eckler’s 37-year old “impossible” challenge to improve Heifetz's score by 10 points using whatever you can find lying around seems even to me like a waste of time. Unwittingly, I stranded at a 7 points improvement using SOWPODS words only, but if any combination of word lists can be used, the sky is the limit.

--asmit

P.S. Now I understand why that familiar set of letters showed up in the puzzles. Without you-know-what-word, I'm stuck so far at a mere 729.

YOU WROTE:

"Using only words from TWL06 and only possible though unlikely moves for the entire game, I found a 1780 points play with the rack ABEOPXZ to make oxyphenbutazone."

If TWL06 is the dictionary also used on Facebook Scrabble (I think it is), your 1780 point word won't work. I know, because I tried it. REQUALIFIED is not a word in that dictionary.

Sorry. 1,778 is the limit.

Ok, I'm a bit confused on these puzzles... Is there a requirement when solving the puzzle that their be a valid path to the game state prior to playing the high scoring word? In many cases, previously existing words are longer than 7 letters and hence must have been formed by extending smaller words. It doesn't always seem obvious what the smaller words were that led to that board arrangement. Shouldn't the solution include how the board evolved to that state-- or at least the intermediate words that make up the longer words?

If there is no valid path to arrive at the state, that makes the 'solution' a lot less interesting in that it could never 'naturally' occur, even in the most contrived game play.

Okay, I have completely revamped my idea on 'sesquioxidizing' (which, as far as I know, is now scientifically correct!). However, because it's too long to post here, here's my blog post:

http://angryrantingkid.blogspot.com/2010/01/blog-post.html

@ jvriezen: I'm pretty sure it's a requirement for the board to be replicable under normal conditions. Once I tried using the solitaire original Lexulous to replicate one of the SOWPODS oxyphenbutazone plays above. It worked!

re:TenzingHillary

I didn't have TWL06 in book form, but used a list supposedly representing TWL06 (its name is "TWL06.txt" and you can download it from multiple sites). Both Dave Wood and I checked all words in Scrabulizer and requalified is accepted in the TWL mode. Thus, unless my TWL06.txt doesn't represent the real thing and Scrabulizer doesn't use TWL06 the 1780 solution seems to be valid. Perhaps Facebook Scrabble uses TWL98. Requalified's absence in that dictionary would explain why the previously known high score was 1778.

Arian

yes well that all sounds pretty impressive, but i dont think we will be seeing very many words of that calibre springing up in too many games up and down the country.
to be fair to myself and the average scrabble player to get a 50-100 pointer is very good and can certainly win a game,
but its nice to see what can be achieved.
jools south yorkshire.

my brain hurts

asmit,

I'm not trying to take away anything from your 1780, which is very impressive. I'm just saying it wouldn't be accepted on Facebook. (If it were, I guarantee you you'd already have had a dozen copycats who'd have posted it on Facebook.)

If you want to try to beat the records on Facebook, you need to check obscure words in the Facebook dictionary itself. I suggest you check out the recent posts on a Facebook discussion page (link below), and then download the spreadsheet I posted a link to there. I have a feeling that, if anyone can beat my Facebook High Game score (3962, single player), you may be the person.

Good luck.

jvriezen,

Valid words do require that the player find some actually possible way to lay out the words, and that he not use more letters than he actually has available. In other words, as you insist, the word must actually be possible (however unlikely) in real life. Asmit's word here satisfies all those criteria. Whether it counts also depends on the dictionary one uses. In the TWL06 dictionary, it would work, but not in whatever version of TWL is used on Facebook, since that dictionary doesn't include "requalified." With other dictionaries, as noted here, even higher word scores are possible.

Though asmit's word would not be acceptable in Facebook, I must say that the way he got to it here is extremely impressive. I too got to a 1780-point word, also using "prequalified" but the way I got there was far less imaginative than what I see here from asmit. (Unfortunately, I learned the hard way that Facebook won't accept "requalified." The lesson there is that you need to check words on the Facebook dictionary to be sure, not just on a TWL word list.)

What's the highest total combined Scrabble game score achievable from playing all 100 tiles? I'm becoming obsessed with the problem, but haven't even reached 3500 points yet.

The best strategy will probably be similar that for attaining the highest 7-letter score, but with two fifteen letter words played on 3X3X3 margins. A mid-board 15-letter word covering the other two triple word scores works well as a platform for building bingos to the margins.

My best so far? Marginal words SESQUICENTENARY and OXYPHENBUTAZONE produced a total score of 3426. Marginal words DEMYTHOLOGIZERS and PREFIGURATIVELY produced a 3291 score. I'm just a hacker. Is this a puzzle that others are working to solve already?

The Scrabble FAQ I linked to back in 2008 (http://home.teleport.com/~stevena/scrabble/faqtext.html#TWL04) now has an updated text reflecting records with dictionaries called "OWL2" (up to 8 letter words) and "OLWL" (longer words). It now reports theoretical high scores of 1782 for a single word and 3986 for a complete game. From the list of created words, it appears that the single play is similar to my 1780 and 1785 point solutions with TWL06 and SOWPODS, but that OWL2+OLWL differs from TWL06 at least in containing the word "zoogamete" and from SOWPODS in missing at least "ameliorative" and "(b)lithesomeness" (and/or "yeven"). All these dictionaries have "requalified", setting the Facebook dictionary apart again. My own TWL06 list so far has matched Scrabulizer's list perfectly, but apparently it doesn't exactly match the National Scrabble Association's "official list". Of course, if the differences are limited to 12 letter words or so, no real scrabble player will mind, but only people hunting for theoretical scores. Still, Dave may be interested to know that there are differences.

re:sourdough
The highest combined score reported on the Scrabble FAQ page (with OWL2+OLWL) is 3986 and is shown in full, with the familiar words "oxyphenbutazone", "microtechniques" and "overidentifying". In this case, all words used are in TWL06. There is a typo in the score table in that the word DE instead of WO is formed in the oxyphenbutazone play (the score is right), but the game otherwise seems to be correct and playable. Given this, 4000+ pts with SOWPODS should be doable.

re:TenzingHillary
Thanks for the compliments, but I'll try to restrain from trying to beat anyone's score on this. It would take a lot of time. Perhaps you can copy the 3986 pt game on Facebook to maintain your top position, though you should be fine where you are, having been the first on top of the world (solo!).

Arian

I found some new high scores for both SOWPODS and the TWL!

TWL:

1458 OXYPHENBUTAZONE
63 OPACIFYING
10 XIS
30 PREQUALIFIED
63 BRAINWASHING
15 AVOIDABLE
30 ZOOGAMETES
63 EJACULATING
50 *BONUS
______
1782

- - - - - -
SOWPODS:

1458 OXYPHENBUTAZONE
66 OECUMENICAL
13 YAFF
25 PREADJUSTEDty
63 BRAINWASHING
19 AQUIVER
31 ZINKIEST
66 EVAPORABILITIES
50 *BONUS
________
1791

It should be fairly straightforward to figure out how these words get laid out on the board - and I guarantee it is possible to make a playable setup with the remaining tiles. Maybe I'll post my board positions somewhere soon..

Ben Woo

Here's some news on the TWL record! There's a close variation on Ben's solution that scores 1783. I almost feel bad taking the record with such a small change.

However, more interesting is this: I have written a program to prove that 1783 is optimal for the TWL! There may be other solutions that score 1783-I didn't check them all-but unless I've made a mistake, there's nothing higher. If anyone wants my program just contact me (it could also be used to look at SOWPODS, for example).

Here is the board, and a construction that gets you there:

OPACIFYING
XIS
YE
PREQUALIFIED
H O
E BE
NO D DORK
BRAINWASHED
US V
T ULna
AMELIORATIVE
ZOOGAMETES
OWNER T
N R
EJACULATING

WASHED
DORK
BE
DOer
Ed
QUALIFIed
RAINwashed
NO
HEn
REqualified
YE
PIer
AS
paCIFYING
DiVULGE
uLNA
aRE
rAT
MElIOratIVE
tETRI
OWNeR
OOgAMeTeS
TiNG
JACULAting
oN
US
uT
OXyPhenButAZonE

Kyle Cronan

Alright, I guess it's stripping spaces, let's try that again:

OPACIFYING_____
XIS____________
YE_____________
PREQUALIFIED___
H__________O___
E_________BE___
NO_D_____DORK__
BRAINWASHED____
US_V___________
T__ULna________
AMELIORATIVE___
ZOOGAMETES_____
OWNER___T______
N_______R______
EJACULATING____

Opposing player is left with a T. :)

I notice the High Word record on Facebook Scrabble has been raised to 1780 by one Tyler Wagner. Presumably that reflects the addition of "REQUALIFIED" to the Facebook Scrabble dictionary, which had previously been absent (thus limiting the High Word record score to 1778, held by 19 people). I predict that you'll soon see numerous others scoring 1780, since the solution has long been available on the Internet (at this very page, for example), awaiting only the addition "REQUALIFIED" to the Facebook Scrabble dictionary.

The Facebook Single Player record score remains at 3962, which has been there for two years now. (NOTE: the high COMBINED SCORES game is 3986, but the winning score in that game was several hundred points lower than the Facebook Single Player record score).

In working on my 3962 point High Game score (single player, not combined scores), I previously hit what I thought was a "ceiling" at 3953, until Jon Prentice scored a 3958 (since increased to 3959). I found it hard to believe anything above 3953 was possible, but there was Jon's score. I then stumbled onto what I'm guessing was Jon's approach (though I really don't know how he got there), and tweaked it a bit to get 3962, which has been the record for two years now. I've tried to get it higher, with no luck. (About a year ago, I thought I had it, but learned that yet another word in the TWL dictionary is not in the Facebook dictionary -- it's always necessary to check first).

There certainly are some players here who might be able to top 3962, assuming it's possible. I tend to think it can't be topped, but I thought the same thing about 3953, so who knows?

Good luck, to anyone who cares to try.

-- Tenzing Hillary

Now that "REQUALIFIED" has worked its way into the Facebook Scrabble dictionary, I doubt that Tyler Wagner's 1780 "High Word" record will last very long.

I just tried both Ben Woo's 1782 game, which plays out, and asmit's 1783 game, which also plays out (though there's no need to play the second "B"). It's only a matter of (a short) time before Facebook players copy these "High Word" solutions and end up with a 1783 score.

Those of you who like such challenges should try for the "High Game" score on Facebook. I've never published the solution, though I've been tempted to, and I don't know where you can find much help on the Internet -- which is what distinguishes the High Game record from the High Word record. You could start with the combined-score high game record (3986), which can be found all over the Internet, but simply copying that game won't even get you into the top ten on Facebook.

To anyone who makes the effort, good luck. I can't imagine myself scoring more than 3962 (unless they add a few new words to the Facebook Scrabble dictionary, in which case I could imagine scoring a few more points). I'll be impressed if someone else can beat it.

I just realized I mistakenly gave asmit credit for coming up with the 1783 High Word score. I should have credited Kyle Cronan. My apology, Kyle, and congratulations.

I came up with a layout that permits Ben Woo's 1782-point game, and played it on Facebook Scrabble. I'm not sure what layout you had in mind, Ben, since you didn't show any here, but I will not that you have to come up with something that leaves you more than 7 tiles in the bag when you're done. (My layout left 9). For that reason, I don't think Kyle's 1783-point game is actually playable. I found a way to increase the number of tiles left in the bag, but only to 5 tiles, and that's not enough. Unless you have more than 7, you can't keep swapping tiles to get what you need for both players (unless you're extremely lucky), and so the game just ends with a whimper. If someone can figure a way around this problem, Kyle's 1783-point game looks like a winner to me: it plays out on paper, just not in real life so far for me.

Sourdough asked:

"What's the highest total combined Scrabble game score achievable from playing all 100 tiles? I'm becoming obsessed with the problem, but haven't even reached 3500 points yet."

It depends on what dictionary one uses. Using the Facebook dictionary, and playing an actually playable game, I believe the answer would be 3986 (combined), for which you can find the solution at many places on the Internet. (3986 is not shown as a record on Facebook because Facebook's "High Game" score counts only one player's score, not the combined scores. The 3986 falls hundreds of points short by that measure.)

Bear in mind that there will always be at least one tile left at the end in an actual game, so 99 is the largest number of tiles that could actually be played in a game. If, theoretically, you played that 100th letter, the 3986 game (as published on the Internet) could be boosted by 10 points, to 3996.

Even more important, bear in mind that it all depends on the dictionary being used. Facebook recently expanded its dictionary, notably adding "REQUALIFIED." This has made possible a higher "High Word" score than formerly appeared on Facebook. The record (mine, for now) is up to 1782 from 1778. I don't take much credit for that High Word, though, since I played the words suggested here by Ben Woo, who thus deserves most of the credit. The 1782 point word did, however, require one to find an actual layout (Ben didn't offer one, though he correctly pointed out such a layout was possible and indicated he might post one later), with all of the "connectors" between the words, and with at least 7 letters left over at the end. (If you end up with fewer than 7 letters in the bag, you can no longer "swap," which brings any high-scoring effort to a screeching halt.) The actual playing-out of Ben's solution (based on my lay-out, at least -- the only one I know of so far) also required some very careful sequencing, a problem understood only by those who've played this "high score" games.

Nonetheless, I don't doubt at all that Ben Woo could have found (and probably did find, as he claimed) a layout for a 1782 point word. That word became possible only because Facebook added "REQUALIFIED" to its dictionary. (Before then, 1778 had been the record, held by 19 people, not including me.)

The tougher task is to beat the "High Game Score" record on Facebook (also mine - 3962). There is no "model" for that on the Internet. The high "combined score" game (3986) is a good starting point, but it requires a very great deal of "reverse engineering" to get to a very high "single player" score (which was "only" 3690 in that game). If you are a serious "high score" player, you should be focusing on that one, since you would be justified in taking credit for topping 3962. I am pleased with that effort far more than my 1782 point word, for which I added just a little bit of thought to what was mostly Ben Woo's accomplishment.

By the way, the 1783 point word offered by Kyle Cronan above, while very imaginative, can't be played out on Facebook. You need at least 7 tiles left at the end, and I can't see any way to have that many left with Kyle's game. So, as far as I know, 1782 is the High Word using Facebook's dictionary. With other dictionaries, of course, much higher words are possible.

-- Tenzing Hillary

Alright, I want to finally get in the conversation. I have the fourth highest place "High Game Score" on Facebook, which is really a tie for third with Andrew Strack, though he got it first. I assume we have similar boards and methods.

When I played, I noticed Scrabble didn't give me double points for the remaining tiles when the winning player bingos and uses the last 7 tiles in the game. That would make 3985, not 3986, really the upper limit under Scrabbles rules. Of course, wouldn't the "total score" really be just 3984? Since this is the total of the scores, wouldn't the winning player's bonus for the losing player's unused tiles be negated by the losing player's penalty for those same tiles?

I would love to somehow work together to come up with a higher scoring high game or single play, but I'm not sure how to do that since everyone seems to want to set the record as an individual.

I just played Kyle Cronan's 1783 game out on Facebook. It wasn't that difficult...

Ryan,

I'm impressed to hear you were able to play out Kyle's 1783 game. A question: Were you able to come up with a layout which left 7 (or more) tiles in the bag when you played OXYPHENBUTAZONE? If so, I'm extremely impressed, since I've looked at it over and over and can't come up with more than 5 left in the bag. (Kyle's solution left only 1, but he was trying merely to present a "theoretical" game, not one that could actually be played on Facebook).

Alternatively, if you were able to play Kyle's game even though you had FEWER than 7 tiles in the bag when you played OXYPHENBUTAZONE, then obviously I'm incorrect to have concluded that you need to have at least 7.

I'm curious to know which one it was. Either alternative makes your 1783-point game very impressive.

As for why you didn't get double points at the end, Facebook uses North American rules, which add your opponent's tile scores and subtract those tile scores from the opponent's total. If Official Scrabble rules (same as European tournament rules, I understand) were instead used, you'd have received "double" your opponent's remaining tile scores, as you'd expected. As you point out, the 3986 game would be 3984 if it were played out on Facebook.

Incidentally, some time ago I concluded that the 3986 game could NOT be played out on Facebook, for the simple reason that one would end up with fewer than 7 tiles in the bag near the end, preventing swaps. Based on your success in playing out Kyle's 1783 point word, if you were able to accomplish that with fewer than 7 tiles left, it would seem to follow that I was incorrect to conclude that the 3986 (or 3984) game is impossible to play.

Be that as it may, the highest combined score I can achieve with my 3962 game is 3976 (or 3978, if it were scored under Official Scrabble/European tournament rules). I've not tried to maximize the combined score, since I've focused only on the "single player" score for Facebook purposes, but I'm not optimistic I could boost the combined score above the 3976 (3978) I've achieved so far.

As far as "working together" to come up with a higher scoring game, I'm actually all for that (though my youngest son tells me I shouldn't be). That 3962 point game has been sitting at the top for two years now, and whatever satisfaction I've gotten from it is being overcome by my desire to see whether it really can be broken. I've not yet decided whether just to "publish" it, but am leaning strongly that way despite my son's remonstrations. In the meantime, I'll give you a hint that may be enough for you.

First some background: I gave Andrew Strack a hint that enabled him to get from 3951 to 3953, which I gave him only because he'd essentially figured out how to do it anyway, and had missed only a few minor "tweaks" along the way that enable one to add 2 points; Andrew struck me as a smart guy who'd figure it out those tweaks sooner or later anyway, and probably "sooner" rather than "later." Unfortunately for Andrew (and me, since my high score was also 3953 at the time), the very day that Andrew's 3953 showed up, Jon Prentice's 3958 also showed up. Andrew and I instantly recognized that Jon must have made some fundamental additional change in the board, since we'd concluded (and still believe) that 3953 is tops under the approach Andrew and I had used.

I don't know what Jon's approach was, but obviously it sent me back to the drawing board. I came up with a modified approach that yielded 3960, and then I found a way to tweak that to 3962. I've tried to come up with ways to boost it higher, but with no luck.

So now the hint: I'm guessing your 3953 game included RAVIGOTE in the same place where it appears in the 3986 combined-score record game. My 3962 point game does not include RAVIGOTE, or any other word in that spot. Instead, it includes an 8-letter word coming down from the first "C" in MICROTECHNIQUES (played across the top row -- I play my games with the long words running horizontally rather than vertically, an arbitrary preference as I'm sure you know).

See whether that hint is enough for you to match or top 3962. I should warn you that I don't see any way to TOP 3962 using my approach, and so you might have to try something different if you don't want merely to top out at 3962.

For now, let me repeat my congratulations on topping my High Word score with your 1783. Either you came up with a much more "efficient" layout than Kyle offered, in which case I'm extremely impressed, or I'm just dead wrong to have concluded you need 7 tiles in the bag at the end. (A third possibility is that you just got extremely lucky with your swaps at the end, but the odds of that are very low).

Good luck, Ryan.

Ryan,

I just figured out how you played out Kyle's 1783 point game (or at least I think I have: the fact that I'd thought it was impossible naturally leaves me doubting my intelligence at the moment).

I'm not sure why it wasn't obvious to me earlier, but you deserve credit for figuring it out while I labored under the mistaken impression it was impossible.

Ryan,

Looks like I was dead wrong about needing to have 7 letters left at the end. Obviously the "dummy" will end up getting whatever he needs after it's too late to swap. The "winner," of course, has to have OXPBAZE by the time there are 7 letters left in the bag, but that's all that's necessary.

Now the interesting question for the "High Word" game will be whether Kyle is correct that 1783 is the top possible score. My hunch so far is that he's correct.

Ryan,

A clarification on the hint I gave you earlier: the 8-letter word coming down from the first "C" in MICROTECHNIQUES is 8 letters WITHOUT the C, 9 letters WITH the C.

Okay, I'll work on it. You're right on the method, by the way. Just play until there are tiles left in the bag, then swap until both players have their next 7 ready to play. OXYPHENBUTAZONE was my last play of the game.

*play until there are just enough tiles left in the bag...

Ryan,

There is one risk to keep in mind at the end, though, practically speaking, it's not a risk when you're playing out a game such as the one Kyle laid out for us. When swapping is no longer permitted, you have to be certain that the bag doesn't contain a tile that must be played BEFORE other tiles in the "opponent's" rack may be played. For example, suppose there are 5 tiles left in the bag, and that there are, say, 9 more letters that need to be played from the opponent's rack. If those 9 more letters are to be played in 3 different places on the board, there will be a very high likelihood that tiles in the opponent's rack will include one or more tiles that can be played in one or more spots, so that the opponent can then draw tiles from the bag until there are no more tiles left and, therefore, he will be able to play whatever letters remain to be played.

But suppose that the 9 letters to be played are all in one location of the board and that there is a "bottleneck" between an already-played area of the board and that to-be-played area, and that a letter needed to get through that bottleneck is still in the bag?

For an example, suppose that, before playing any other tiles, the opponent must play "BIKE," after which he can play his remaining letters by playing short words from the B and the I. Suppose the E in BIKE is already there on the board, but he needs the BIK. If the K is still in the bag with only 5 tiles left in the bag, the "opponent" will be unable to play (other than, of course, to play some other word which will put an end to the scripted game).

This need isn't obvious when there are several spots to play. For example, if he could play either BIKE in one location or IN or TAP in another location, chances are pretty high that he can play in one of those other two locations, and then draw tiles until the K shows up. The board at the end of Kyle's game was like that. But if a "bottleneck" situation arose, it would be necessary to be sure, before swapping became impossible, that the opponent's rack contains enough certainly-playable tiles that, once those tiles have been played, he will have drawn all of the remaining tiles in the bag.

I suspect this was already clear to you, but it may help others, and it at least clarifies my own thinking.

Good luck on the High Game effort. I think it's more interesting to figure it out. Though we deserve some credit for playing out a 1783 point word (especially you, since I'd mistakenly believed at first it wasn't possible), the real credit belongs to Kyle Cronan for figuring out the words and the layout.

If you top my High Game score, you'll deserve all the credit. I have a hunch that, if it's possible, you'll do it. I'm just not sure it's possible.

TenzingHillary,

I've come up with a solution that scores 3954, which I just played (it takes a few days to show). That should move me into third place from my current fourth.

It looks like I'm going to have to claw for each of the remaining 8 points though - I'm more impressed with your 3962 score than ever.

Ryan,

3954 is quite interesting, because that puts you over the 3953 ceiling that I and Andrew Strack thought was tops until Jon Prentice's 3958 showed up (since boosted to 3959). What puzzles me, though, is how both you and Jon got above 3953 in a way that may be different from how I got there. If you did, then I may be wrong to conclude 3962 is the maximum. On the other hand, if you did get there the same way I did, you'd almost certainly get above 3954. That suggests you have a different method.

The bottom line seems to be that there is more than one way to get there, leaving me less sure than ever that 3962 is the maximum. Keep me posted if you get closer, and good luck.

TenzingHillary,

I'm up to 3962. I actually see 4 different (though very similar) boards that would play out to that amount, but nothing more yet.

How long does it take you play a 3950+ game on facebook? The highest single score was like 2 hours, but highest total score takes me like 6 at least.

I'll keep you posted (though I assume you'll also just see on Facebook) if I can break 3962.

Congratulations, Ryan.

There indeed are four different boards that play out to 3962 (assuming you got there by figuring out the word I hinted at above), though the only difference lies in where you play one letter. If you've found something even more different than that, I must confess to having no idea what it is, and you deserve a great deal of credit for finding it.

If I can think a way of getting over 3962, I'll let you know and at least give you some guidance as to how I got there. I've been looking again lately, though, and once again am convinced it just can't be done. I'll be amazed, and extremely impressed, if you can do it.

By the way, now you're at 3962, don't you share my puzzlement at how Jon Prentice could have "broken through" the 3953 "ceiling" and yet not have reached all the way to 3962? That still puzzles me. Once one makes the necessary breakthrough, there's still considerably more tinkering required to squeeze out those last few points, but you can get at least within a couple of points without all that much difficulty.

It again makes me think that Jon took a different route from you and me. (And I can't even be sure, of course, that you followed the same route as I. Maybe there is more than one.)

In any case, congratulations, and good luck trying to top 3962.

Ryan,

I forgot to answer your question about how long it takes to play out one of these long games. I'm not sure I recall exactly, since it's been two years since I've actually played one out. As I recall, though, it took about 3 hours to play 3962. Playing the "high word" score was quicker - about 90 minutes, as I recall - though I'm not sure why that should be, since you end up (as you well know) playing out practically the whole game just to play that one big word.

This undertaking illustrates the wisdom of the old adage: "Measure twice, cut once." I can't even imagine playing out a game that one hasn't laid out ahead of time from start to finish (though I've had a little confusion about "end games" that has more than once brought me to a frustrating dead end, as some of my comments above should indicate).

TenzingHillary,

The highest single word game goes so much faster because both players can play letters every turn (until the super-scoring play). This makes the game go twice as fast, as it's essentially like having a 14 tile rack instead of a 7 tile one for most of the game - there's much less letter swapping.

I've played my 3962 game now. It might be another day before it shows up. I'm still working on improving it, but that seems pretty tough. I've come up with some interesting variations, but they typically score 5-10 points less.

Are there any hints you would like to know about my 3962 board? I'm thinking of working on the highest combined score (that doesn't use SESQUIOXIDIZING) next. What's your value for that?

Oh, also the highest single word game goes faster because you can play more letters at a time. In the highest total game, there are more turns where you play a single letter, then another single letter, etc., in order to score as many times as possible, while on the highest single play game, I play as many letters as I can at once. Also, on the highest single play game, you can save a little time by not even playing all 99 of the letters (this only saves a few turns, but still would make the game shorter).

Ryan,

I'm curious to know whether your 3962 game played the 9-letter word down from the first C in MICROTECHNIQUES, as I'd mentioned. If so, then I'm not confident one can top 3962, as I've given that a great deal of thought already.

But if your layout does not include that 9-letter word, I'd be curious just to know that. That simple fact would be enough of a hint, since it would once again make me doubt my conclusion that 3962 is the highest possible Facebook game. As I mentioned earlier, Jon Prentice essentially gave me such a "hint" when his 3958 game (since boosted to 3959) made clear that I'd been mistaken to believe that 3953 was the maximum.

Ryan,

On the "Highest Combined Score" question, I think the 3986 record is the best I've seen. As I mentioned, that hasn't been my focus (having concentrated instead on the "Highest Game - Single Player" record on Facebook). Nonetheless, when I have looked at it, I've come up with nothing better than the 3986 that's already been played.

In my 3962 point game, the opponent's tiles can be played in one of two ways, yielding either 11 or 14 points. The latter way yields 3978 combined points, under the "European" scoring system used to score the 3986 combined-score record game (or 3976 points, under the alternate North American/Facebook scoring rules), still 8 points short.

Ryan,

You wrote:

"I'm still working on improving it, but that seems pretty tough. I've come up with some interesting variations, but they typically score 5-10 points less."

As you get higher and higher, that indeed is the problem: interesting variations, but finding one that increases the total score gets tougher and tougher. But that's what makes the challenge interesting.

Ryan,

By the way, now that you're on top too, you can share some of the abuse from a few odd ducks on Facebook who insist that achieving these high scores amounts to "cheating." Some of them get quite angry and even threaten bodily harm, though most people appear simply not to understand that these "high score" games are different from competitive games: no cheating at all is involved, just careful scripting after a very great deal of analysis.

Electronic Arts and Facebook really should establish a "dual track" for high scores -- one for competitive games, the other for non-competitive games. As will be obvious to anyone who's played a "scripted" game, it would be child's play to distinguish the categories simply by setting a maximum number of swaps and passes for competitive games. This would allow competitive-game high scores to get the recognition they deserve, without preventing the satisfaction that many of us get from the intellectual challenge of maximizing the possible score in theoretical (but playable) games.

And last, but not least, "dual track" records might settle down some of the strange birds on Facebook who get so upset that they threaten bodily harm. It's amazing and disappointing to me that anonymity on the Internet exposes such a violent nature in many people.

Because word challenges aren't allowed, moves aren't time limited, there's an in-game dictionary, and there's no way to monitor use of a word generator, I just don't think the Facebook version of Scrabble is a good place to keep records for "competitive" play. Theoretically, I could play a friend, and we could be on the phone discussing what letters we each have, setting each other up for plays, etc., and there wouldn't even be any evidence of it through Facebook. I look to Facebook for cooperative records, and the Scrabble Association (whatever it's called) for sanctioned, monitored games non-"tool assisted" records.

In regards to my 3962 solutions, I did replace RAVIGOTE with a C word, which allowed me to go from 3953 to 3962. My solutions all sound like the one you're describing - I expect ours are essentially the same at this point, but I haven't given up yet.

Quick question, under North American rules, in a game where one player scored 3985 and used all his tiles and ended his final turn with a bingo, and the second player had the "A" tile remaining, would the second player's score be - zero, negative one, or negative two points? I just want to clear up what the "total" should be. Thanks.

I meant to say "scored 3984...", in reference to the current "3986" game.

Under North American rules, assuming the second player has never played a word, I don't know whether his score would be zero or -1. It certainly wouldn't be -2. Under European rules, his score would be zero, and 1 point would be added to the first player's score.

Kyle,

I'd love to look at the program. I've been doing my stuff by Excel spreadsheets. How can I contact you privately?

Ryan,

I've still not found a solution to top 3962, but I have found a much different layout that ties it. It does without quite a few words that most people use, such as ALAMEDAS, RAVIGOTE, JIBB, FLAMED, WAB, QATS, and all of the smaller words that make up those longer words. Since it's so much different, it's very bad luck that it just works out to a tie, and so I won't bother to play it out. I do wish, though, that I'd thought of this solution first, since it's so much different. I'll keep looking for a way to squeeze another point out of it, but I sure don't see one yet.

If you want to reach Kyle, I suggest a Google search. I don't know whether he still checks this website.

TenzingHillary,

I've been working on the highest single scoring play, and I've proved the theoretical maximum (under TWL06) down to 1830 right now. It's very possible/likely that Kyle was right, and 1783 is the maximum under that dictionary. I'm going to keep working on that until I either prove him wrong or reach his conclusion as well.

In regards to your board - That's awesome! I feel completely confident the 3962/3986 score has not been maximized, especially under TWL06. I'm more of a fine-tuner/maximizer than I am an innovater though, so I'm always jealous of finds like yours. I'd love to see the board once you've reached your limit (and set the new record if applicable).

Ryan,

I'm disappointed that my new layout only ties the 3962 record, and so I'm all but certain I'll publish it or at least let you know, just because it's interesting to see how different it is. I'm going to keep trying for a little while, though, to see whether I can raise it (or find some other way).

I don't understand how you and Kyle prove maximum scores, but looking over my word lists led me to conclude he was probably right about 1783 being tops for a single play. I assume your 1830 is possible because Facebook doesn't use TWL06, correct?

TenzingHillary,

Would you care to take this conversation private? If not, that's okay.

In regards to the highest scoring turn, I first checked the points awarded for all the 15 letter words in TWL06. OXYPHENBUTAZONE is the highest, and the next highest is 135 points less.

I then created a list of all the valid words that became other valid words when you added a letter to the front of them. Even if you got 7 copies of the highest of those, and somehow got full points for all of them, you wouldn't exceed 1783 by 135 points, meaning that you have to use OXYPHENBUTAZONE.

I then found the 7 highest scoring words you could make from each of the 15 letters in OXYPHENBUTAZONE. When I added the 1458 for Oxyphenbutazone, the 50 point bonus for using all the tiles, and the highest scoring 7 secondary words, I think the score was either 1831. This means that without the 50 point bonus, any solution would be lower than our 1783, so 7 secondary words do have to be made to exceed 1783, if that's possible.

Then, I realized that some of the words had either an X or a Z in them, which wouldn't score their full value, so I had to replace those with the next highest scoring words, which lowered the score to 1829 I believe.

I think two of my words use a Q, so I know that both can't receive the full 10 points. There are obviously some letters that are used more times than allowed as well, so I'm going to try to find a way to come up with the highest scoring combination of words that don't require more blanks that I have available. Then, I'm going to have to make sure they're legal to play together, etc.

For the highest scoring board, I was looking at connecting the G in overidentifying with either the S in microtechniques or the E in oxyphenbutazone. When you say you didn't use all of the smaller words to make up the longer words you didn't use, I'm not sure I understand. Do you mean you didn't use QAT (because you didn't use QATS), and that you didn't use AM, LAM, FLAM, and FLAME (because you didn't use FLAMED)? My 3962 solution already didn't use QATS and RAVIGOTE - does yours use the same C word that you used in the 3962 solution you played on Facebook?

Disregard the FLAME question, as I realize you could swap another letter (such as B or T) for the M there. I'd be more curious to know if you used ALAMEDA (you said you didn't use ALAMEDAS, and that you didn't use the smaller words, but I would just want to clarify. I think ALAMEDA and NAMETAG (starting next to the first E in Microtechniques) were the only good 7 letter plays I remember that played well next to microtechniques (and created "arousing").

Anyway, I'm rambling now, so let me end this.

Ryan,

Take a look at AMANITA instead of ALAMEDA in the upper right corner. I did use the same C word, though I played one less letter off of it at the top. I made changes in the left, center and right portions of the bottom of the board, including playing 4 letters above the X and Y that yielded a whopping 63 points in two plays. One of those letters came from the center-bottom, which in turn required a change to the right-bottom. So, bottom line, I changed the upper right and lower left corners a lot, the center-bottom, right-bottom and left-top a little, and voila: 3962 points. Those extra points in the lower-left corner offset "losses" in other areas resulting from changes necessary to free up those letters. Alas, the offset was perfect, so that I didn't actually generate even one additional point, despite the substantially different look of the layout. Unlike my (and your, I gather) earlier 3962 game, I didn't have any "left-over" letters at the end, and so the opportunity to boost the score a bit higher just wasn't there. I haven't given up yet, but I'm almost at that point with this layout. On the other hand, the very fact that I found another layout that gets the same 3962 result has persuaded me that there may be yet another one out there, possibly one that yields more than 3962 points.

I like to leave it in your capable hands to fill in the gaps, which I'm confident you can do. I should mention that I also found it necessary to change a bit the letters played at the end by the opponent, which ended up reducing his score to 9 points (down to 8 after subtracting his remaining one-point letter).

Ryan,

One additional hint: the opponent can end up with either an L or a T at the end. You can form a word in the lower-left corner with either letter.

I think you'll find it interesting to try to figure out this new layout, and you might even have better luck than I've had in squeezing out an extra point.

Ryan,

Incidentally, I just noticed I've been misstating the rule differences applicable to the end-game. It's the North American tournament rules that add double the remaining tile scores to the score of the player who goes out first. That differs from Facebook and (I gather) European tournament rules, which instead add the remaining tile scores (just once) to the score of the player who goes out, and subtract the same number from the other player's score.

"Scrabble rules differ in how they adjust scores for value of tiles remaining at end. Under official rules of Scrabble (used on Facebook) and European tournament rules, value of remaining tiles is added to score of player who ends game and subtracted from score of player who holds those tiles, so COMBINED score is not changed. Under North American tournament rules, DOUBLE the value of the remaining tiles is added to score of player who ends game, but nothing is subtracted from other player's score, so COMBINED score is increased by amount of adjustment. Margin of victory is the same under either set of rules."

Ryan,

A correction: in the upper left corner, I still played 3 letters off of the C word, but they were in columns 2-4 rather than 1-3.

So I have a solution that meets most of your criteria, (the 63 points on the XY in two plays, the lack of RAVIGOTE, the C connection from Microtechniques to Overidentifying, letters in columns 2-4 off of it, the center bottom change to go above XY, the right bottom change to fix the center bottom, amanita substitution, the L or T at the end because either can be used in the lower left, the 9 points scored by the low scoring player (could be 12 if so desired in mine), etc.

My board is only scoring 3912 before the letter bonus though.... I'm thinking maybe I did something different around my alternatives to JIBB or FLAMED.

Disregard that. I had forgotten a 50 point bonus, but I had also over-counted 7 points as well. Still working on it...

Ryan,

Not sure what you're doing differently, but I have a hunch based on your comment about the lower scoring player being able to score either 9 or 12 points. That's not an option in my layout. If you figure out why that's so, you may eliminate the one difference between our layouts.

I may be mistaken in assuming what you've come up with, of course, but I suspect it has something to do with this.

Ryan,

Another observation: You wonder whether you did something different in your alternative to JIBB. There's only one alternative: JAMB. So I suspect we're doing the same thing there.

Ryan,

Well, I just played out my new layout, hoping that maybe I'd mistakenly calculated the score. But it came out 3962 to 8, as expected. Much like the first layout, this one doesn't appear (to me, at least) to be susceptible to any increase. With all the changes, there is no particular reason it had to work out that way -- it could just as easily have ended up higher or lower -- but that's the way it appears to work out. Unless you can get more from it (assuming you soon figure it out, as I'm confident you will), it looks like some other layout will be needed -- unless 3962 is just the max.

Ryan,

By the way, if you like, I'll be glad to post the new layout here, though you'll probably get more pleasure out of figuring it out yourself. Nonetheless, if you prefer, just let me know.

Ryan,

I take back my earlier comment on the losing player having an option between 9 and 12: he DOES have that option, depending on whether he plays ROTE or ROT, ROTE. I just had him play ROTE, since there was no point in trying for a high combined-score game with this layout.

If you are doing the same thing as I did there, I guess I'm at a loss to figure out why you're not reaching 3962. I suggest you take a careful look at the places where you've made the suggested changes, and be sure you're getting the maximum possible points in each spot.

Good luck.

Luck? Who needs luck. I just squeezed 3966 out of a variation of the AMANITA board. I'm still working on optimizing it, though I'm not sure how much more it will yield. I will keep you posted.

I can post that layout here if you want, or we can exchange it privately. This is a public forum - I'm not sure if you want it displayed for everyone indefinitely.

Ryan,

Wow, 3966! Very impressive. Obviously you've figured it out and then some.

No need to post your variation. Now that you've let me know I was wrong to believe that layout maxed out at 3962, I'll try to figure out what I've been overlooking.\\

Well have fun! I think 3966 may be my maximum on this board. I realize that since you had already been optimizing for a while, there wasn't likely to be much left. It looks like you were pretty thorough already. Perhaps with another larger word change somehow.

I've actually been working on different variations of this and the previous 3962 board, but I haven't found anything worthy of mention yet. I hope you find a way to go beyond 3966.

I emailed Nathan Hedt a while back, by the way (he is credited with the 3986 play in the FAQ). He said he had once found a 3995 solution (I'm assuming this was 3993 plus 2 for an unused tile), but that he had never reached 4000. Those are my highest total goals - reach 3995, break 3995, reach 4000, then break 4000. Of course, along the way, I'll be simultaneously working on the highest game score for one player.

Ryan,

To reach 4000 would be quite an accomplishment. We're so close, but those last few points are elusive.

At the moment, I'll be happy just to figure out how you squeezed four more points out of my AMANITA layout. I hope to take a closer look over the weekend at what I may have overlooked, but my brief look today leaves me utterly baffled for the time being. Great work on that.

TenzingHillary,

Anything yet? I've been working on the 3962/3966 boards, but I don't see much more to add to them at the moment. Do you have any other variations you've found that score in the same range? I'd love something else to analyze.

Let me know if you want any hints or the solution on the 3966 game.

Ryan,

I finally figured out your 4-point enhancement of my AMANITA board (and played it out earlier today). I found your enhancement to be extremely impressive.

Probably needless to say, I wouldn't have mentioned my new layout if I hadn't been fairly confident I'd maxed it out. I do like to share what I’ve learned -- albeit through hints, as you’ve learned here, so that a great deal of effort and skill is still required -- but I doubt that even Mother Theresa would have mentioned that new layout without first figuring out whether she could use it herself to boost the record score. I thought I’d looked it over long enough, but obviously I was wrong about that.

When you finally figured it out, and added 4 points to boot, I naturally thought I'd been too hasty. I expected I'd revisit the layout and figure out what I'd overlooked in an hour or so. I'd like to say that's what happened, but I’d be far less than honest to say that. Much to my surprise and frustration, I spent very many hours trying to figure out what you'd done.

You'd said you were an "optimizer" rather than an "innovator," and so I started out by looking where you might have moved the one obviously-movable tile in my layout (I'm sure you know which tile I mean). Since removing that tile from where I had initially played it immediately subtracts 14 points from one's score, I naturally looked for an alternate place where the tile would yield at least 18 points (since you'd added 4 points to my initial layout total).

As I'm sure you found, there were no obvious alternative spots. The place you chose (and I finally stumbled upon) was not anywhere near the top of the list, since it appeared at first glance to offer only a few points. Frankly, I don't recall even considering that spot early in my search, and certainly wouldn't have devoted more than 3 seconds to it if I had. Nor did any of the other alternate locations for that “movable” letter yield anything close to 18 points.

At that point, I concluded you were being modest in claiming only to be an “optimizer,” and so I started evaluating more substantial changes, all of which led to dead ends (some promising enough that I may mention them to you later, but dead ends nonetheless). After many such disappointing efforts, I restored my original AMANITA layout, stared at it for the hundredth time, and asked myself for the hundredth time: "How in the world did Ryan squeeze 4 more points out of this?" And after all those hours, I was no closer to answering that question than when I'd started.

In one sense, though, that was not quite true: All of my "dead ends" had at least left me persuaded that I hadn't left many stones unturned. That meant I knew quite a few things you had NOT done, and that struck me as helpful. Helpful or not, I was nonetheless clueless about what you HAD done.

So I considered yet again what I’d started with: the possibility that you had only tweaked the AMANITA layout after all, and I tried yet again to figure out exactly how you’d tweaked it. That’s when I finally stumbled upon your solution, which, as I said above, I found very impressive. It impressed me because it requires several plays (three) to yield all of the extra points, and depends on the proper sequencing of those three plays. Once one figures that out, what appears at first glance to be a “trade” of 14 points for 5 points reveals what it really is: a “trade” of 26 points for 30 points.

Had you not figured out that enhancement, I’d never have thought it possible and would not have even looked for it. Nothing kept me looking except the inescapable fact that you’d done it. What I learned from all this is that those who claim to be only “optimizers” in fact add a lot more value than “innovators” in many cases – at least in this pursuit. Finding new layouts is not that hard. Finding new layouts that yield as much as earlier proven layouts is harder, of course. But taking a new layout and exploiting its full potential is often much harder, and I think that was the case here. I think 99 out of 100 people would simply have looked at the spot you found and passed it by without a second glance. Regrettably but honestly, I must put myself in that “99” category. Fortunately for me, you were the 1 in 100. By showing that more was possible, you enabled me to pull myself out of that “99” category, and I appreciate that.

Having been proven wrong twice now – once when I thought my 3953 record score was unbeatable, and a second time when I thought my 3962 record score was unbeatable (though the latter record did last two years) – I’m nevertheless inclined to risk making the same mistake yet another time: I don’t think anyone can top 3966.

If anyone can prove that wrong, you’re probably the guy. So keep up the good work, and good luck.

-- Tenzing Hillary

Ryan,

I can’t say these will lead you anywhere, but given your proven analytical abilities, I’ll mention a few efforts that were dead ends for me but may well be fruitful for you. Bear in mind, when I describe board areas, that I arrange my long words horizontally rather than vertically.

In the lower right corner, try FLAVOR horizontally and AVO vertically. It yields more in that corner than what you and I have there now, though at the obvious cost of eliminating the word descending from the first “C” in MICROTECHNIQUES.

Also in the lower right corner, consider REINJURED and DITZ, rather than REINJURES and SPAZ.

I’ve come up with nothing promising in lieu of WOLFLIKE.

Trying to replace REINJURES with anything but REINJURED is difficult. Bear in mind that 2 letters in the word must be played before the bingo is played, since it’s a nine-letter word. I’d thought READJUSTS might work, since that would free up useful letters (R, I and N), but there’s no way to play a second letter at the usual “connector” locations (either the second letter or the second-last letter), and so there’s no practical way to play READJUSTS.

I think my AMANITA substitution for ALAMEDAS is a significant improvement since it frees up one letter. On the other hand, it makes it necessary to connect to ROTE from the upper left corner, which precludes playing ADO in the upper left corner (which, as you know, later yields 12 additional points when you play MICROTECHNIQUES (because MA is formed). If you can play a letter to the left of the A in CAROUSING, you can connect to ROTE from the right end, thus freeing up the “connector” letter underneath the R in ROTE and allowing ADO in the upper left corner (or simply DO, if you can find a better use for the A).

In lieu of CAROUSING, I’ve experimented with CRAVENING and CRANKLING, but to no avail.

Here’s a significant change that I toyed with only briefly, but may (or may not) be worth looking further into. Eliminate the long word descending from the first C in MICROTECHNIQUES. Replace it with a long word from the final E in MICROTECHNIQUES down to the second N in OVERIDENTIFYING. I’ve tried ESTRAGON and ESTROGEN, since those words use many of the letters from the to-be-abandoned long word descending from the first C in MICROTECHNIQUES.

Finally, I tried, again without success, to connect from the left side of REINJURES down to the center bottom. Making the connection is not difficult (and allows you to eliminate the letter to the left of the M in JAMB if you choose to do so), but I wasn’t able to find a way to do so that generated enough points to make the effort worthwhile.

As I said above, none of these efforts yielded much for me, but I’m confident they’ll yield more for you. Whether they’ll yield enough, who knows? But one or two of them might be worth pursuing.

Good luck.

Tenzing Hillary,

It sounds like you've found the same four point improvement that I did. I had secretly hoped you'd found another, but after reading your description, it's clear it's the same as mine. Good job on that.

I had played around with the other "-FLIKE" words (CALF, GULK, HOOF, LEAF, ROOF, RUFF, SERF, SURF, TURF, and WAIF) before, but hadn't found much to do with them, as freeing the W from WOLFLIKE costs some points that I was having trouble restoring. I think LEAFLIKE was useable under one of my recent boards, but I wasn't able to make up the lost points.

I'm still inclined to believe that a 3993 board (before bonuses) exists. I would expect, therefore, that there's a way to manipulate that 2nd player moves of that board to score somewhere from 3970 to 3980. Of course, I haven't found a board that scores 3993 before bonuses, yet, but I believe it exists until I'm proven otherwise.

I wouldn't say there's no practical way to play READJUSTS. REINJURES scored 146 in my 3962 game, and it scored 148 in my 3966 game (because RE was made before REINJURES was played).In regards to READJUSTS, I count 98 points for READJUST, 34 points for READJUSTS, and a 5 point gain on SPA now being worth 10 instead of 5, for a total of 137. It's definitely less, but probably worth reviewing if freeing up the INRE is a beneficial trade for the ADST. You don't get the "double double play" on the J that way though, which is like my 3962 and not like my 3966. Of course, I could be making some dumb mistake in my analysis here - always feel free to let me know if that's the case.

I had tried CRANKLING before as well, but not CRAVENING (I'm thinking I was looking for WOLFLIKE alternatives at the time.) I may get back to that and CRAVENING as well, but not right away.

I will take a crack at the FLAVOR variant.

Just to be clear, to play READJUST>READJUSTS/SPA without a word in the RAVIGOTE position, you'd have to replace WOLFLIKE with a word that ends with T.

Or you could remove WOLFLIKE and play RAVIGOTE (or another word that ends with E and fits there) to play the letters RADJUST to make READJUST, then follow with READJUSTS/SPA.

Just my thoughts. Let me know what you think.

Ryan,

Sounds like you've already tried a lot of what I tried.

Interesting observation on READJUSTS, but, as you note, it still leaves you short, especially on the J. Freeing up the I, N and R by playing READJUSTS could be worth it even so, I suppose, if you could find very productive uses for those letters. As you've undoubtedly noticed, all three letters are in very short supply, and thus valuable, but I'm not sure they can be put to good enough use to offset the significant loss of points.

Unless one can change one or more of CAROUSING, WOLFLIKE and REINJURES, I'm not sure where else improvements can be made. About the only major change that strikes me as worth pursuing is to replace the long word coming down from the first C in MICROTECHNIQUES, but I've tried that several times and ended up with fewer points from the alternative. It might also be that replacing AMANITA with an 8-letter word, thus making possible a better play in the upper left corner, might yield some improvement. But I've played with that possibility at some considerable length, and AMANITA has been the best I've come up with so far. I suspect you've spent some time working on possibilities there too.

As our score gets higher and higher, opportunity costs also get higher and higher. It may be that some dramatic changes could make a difference, but I'm inclined to doubt that. Disturbing any of the three long words, for example, seems unpromising. I've looked several times for a replacement for OVERIDENTIFYING (OVERSIMPLIFYING, for example, and a couple of other lesser candidates), but I always end up concluding OVERIDENTIFYING is best.

Tenzing

P.S. Incidentally, I think you got 168 for REINJURES in your 3966 game, not counting the two points for RE (146 + 20 +2), no?

P.S. What about REINJURED and DITZ? Have you tried that?

Ryan,

I should add that the main shortcoming of both CRANKLING and CRAVENING is that each of them contains a second N. N's are in very short supply, since 5 of the 6 are used up by the three long words (one shared by whatever word is played on the very center square on the first turn), and the sixth by REINJURES. Since AMANITA uses both blanks, that leaves you stuck.

Well, the 3962 board didn't fully maximize the J, and it still was a very nice board, but without the gain there to offset the loss on not having the J doubled in READJUSTS, it does seem like the deficit there would be hard to overcome to surpass the 3966 board. Of course, if you could show high scoring opportunities for the INRE letters, it would be worth at least entertaining.

The W and K in WOLFLIKE are getting multiplied by 4. Moving the K to CRANKLING would get it scored 5 times (ANKLING, RANKLING, AND CRANKLINGx3), so if the W could get scored highly (perhaps in the WE/WAB position), it would seem like a gain to move those letters out of wolflike and into higher scoring spots, and move lower scoring letters out of CAROUSING or other spots into the 4X location where wolflike is now.

I hadn't done much with REINJURED/DITZ. Where would you like to have the "P" from SPA? What would you play instead of the words we've been using in the FLAMED position? Is there one that scores as much (or nearly as much) as FLAMED (in several plays) from the newly available set of letters?

Ryan,

You get 6x for the K if you play CRANKLING (ANKLING is 2x). Even so, the use of 2 N's makes it hard to justify CRANKLING or CRAVENING.

I agree it would be hard to profit from READJUSTS, but do note that IRN could be useful letters to free up: N for CRANKLING or CRAVENING, for example, R for RAMADAS (which would free up one blank, and permit a right-end connection to ROTE), I for JIBB (though your clever enhancement in that area makes one wonder whether it's wise to revert to JIBB even if that becomes possible), since your enhancement wouldn't work with JIBB).

I don't know that REINJURED/DITZ leads anywhere. I only note that it's really the only other practical choice there. In the end, I don't see any hope for improvement. Possibly, though probably not likely, with the FLAVOR/AVO alternative, but that requires a lot of reworking elsewhere. Possibly too with the substitution, for the "C" word down from MICROTECHNIQUES, of a word down from the second E in MICROTECHNIQUES, but that too will require a lot of reworking.

In other words, I'm pretty much out of ideas, for the time being, except for these few. I hope you can make something out of one of them.

Well, I'll keep working through some of the ideas we've discussed. If I come up with anything I'll discuss it here.

If you get a board that comes close to the 3950s-3960s and you give up on improving it further, I'd be glad to take a shot at it.

So when I've tried ESTROGEN and ESTRAGON, it doesn't seem productive. I'm losing one of the multipliers on those letters (that were originally scored 3 times, twice as an 8 letter word and once as a 9 letter word) because STRAGON/STROGEN don't hook to anything, so it's only scored as a double. And then I lose some of the multipliers on the "V", which was originally tripled (first a double, then a single) when connecting the first C in MICROTECHNIQUES to the first E in OVERIDENTIFYING. I just don't see any major gains anywhere to make up all those lost points.

I'll keep working on FLAVOR and DITZ next.

Ryan,

It is frustrating as we move higher and higher. I can't imagine changing OXYPHENBUTAZONE or MICROTECHNIQUES, but I have toyed with changing OVERIDENTIFYING. As I mentioned earlier, I've never had any luck in that effort, but I can't say for sure that it's pointless to keep trying. OVERSIMPLIFYING seems promising, but it puts P right on the very center square, and it's hard to find anything coming down to that P from the second C in MICROTECHNIQUES.

Maybe we're just maxed out. Again, though, I've said that before and been wrong twice already.

Maybe Nathan Hedt has some ideas. You mentioned he said he'd increased his "combined score" game to 3993 (before bonus), and maybe some of what he changed would be useful for "single player" boards too.

He looked for it and then said he couldn't find his 3995 board, but he didn't seem interested in sharing anything anyway at the time. Maybe I'll try him again someday. Until then, I'll keep working at it. I think there should be some combination of moves from the 3986, 3962, and 3966 boards that should allow a few more points to be squeezed out.

You're more optimistic than I, which is commendable.

Merry Christmas, Ryan.

Tenzing Hillary,

Merry (late) Christmas to you!

I was thinking more about the 9 point improvement to the board used to score 3953. There's a fair chance that since it scores 9 total points more, that we'd max out at 3953+9, which just happens to be 3962. The odds that it allows you to play above 3966 is unlikely.

If the 9 point improvement doesn't use QATS, then you wouldn't lose the 12 points that are lost from repositioning that S, and I would estimate 3953+9+12, or 3974, as a realistic maximum.

It's a longshot that the 9 point improvement doesn't use QATS, and therefore unlikely that the 9 point improvement will lead to a solution that exceeds 3966 for Player 1, but since I'm already interested in the 9 point improvement on its own, I will keep pursuing it. Our 3962 and 3966 solutions score less than the board that was used to derive the 3953 board (partly due to the QATS issue), so I think finding the 9 point improvement is in a different direction than we've been going.

Also, I checked out FLAVOR as an alternative to FLAMED/FLAWED, but it seems FLAMED/FLAWED and the points from the extra A placed elsewhere exceed the points scored by FLAVOR (and PA/AVO). Did FLAVOR seem promising, or did you just want it reviewed as a potential alternative?

"Did FLAVOR seem promising, or did you just want it reviewed as a potential alternative?"

Realistically the latter, with a hope that the former would turn out to be true. Since it "fits" there, it seems appropriate to explore its possibilities. It sounds like you've done so as much as I did, or more, and neither of us came up with anything. I thought you might, given your success in finding refinements I'd overlooked.

On the rest of your comment, I can't say I grasp what you're saying, so please clarify. For what it's worth, none of the boards I've used - to achieve 3953, 3962 or 3966, used QATS, for the simple reason that ATS is not a word and the Q is played in the very last turn of the game (with MICROTECHNIQUES) in all of my games (to make either QAT or QI). QATS can only be played if the S is played after MICROTECHNIQUES, which doesn't happen in any of my high-scoring games.

Obviously, if MICROTECHNIQUES is not the last word played, QATS would be possible (not to mention attractive). A long time ago, I played around with some layouts in which OXYPHENBUTAZONE was played last, which allows QATS (and I certainly played it in those layouts), but those layouts had shortcomings of their own: only when I switched to MICROTECHNIQUES-last layouts was I able to get even 3950. Nonetheless, it might be worth experimenting with OXYPHENBUTAZONE endings again, just to permit the playing of QATS.

Another alternative, of course, is saving OVERIDENTIFYING for last, as Nathan Hedt did when he played his 3986 game. That isn't feasible for any layout that plays a word off of either the OVER or the ING, as you and I did in our 3962 and 3966 layouts, but Nathan Hedt's game didn't, since he connected only RAVIGOTE and WOLFLIKE to OVERIDENTIFYING.

On balance, finding a way to play QATS may be worth looking into, since it's the only apparent source of a lot more points. Assuming you used the same layout as I did for 3953, playing QATS would increase the value of the S from 14 points to 26 points, a big jump (though not enough unless added to a score already higher than 3953).

Sorry, I have a habit of not being as precise as I can.

Nathan has a board that scores 3995. I am presuming that's a 9 point improvement to his 3986 board (I used his 3986 board to play my 3953 game, except I had to not play QATS).

If the 3995 board still uses QATS, then I doubt we'll be able to use the 3995 board to exceed 3966.

If, however, the 3995 board doesn't use QATS, then I think we could probably score more than 3966 from it.

Tenzing Hillary,

Happy New Year!

I've been looking at FLAVOR and PA/AVO again as an alternative to FLAMED/FLAWED. It actually frees up some replacements for RAVIGOTE, like MiTIGATE, GEMInATE, MEdITATE, AdMITTEE, and EsTIMATE. I'm still only scoring a total in the 3960's though, and like the low 3940's for Player 1. I'm still using ALAMEDAS, though, as I don't have the letters to connect the first C in MICROTECHNIQUES with the first E in OVERIDENTIFYING.

Would you care to try and give it a try?

Let me know.

Actually I might have double counted 16 points on that solution. Anyway, still no luck with FLAVOR. I'll try REINJURED/DITZ in more detail now.

Ryan,

Good to hear from you. I don't know that I'll have time to concentrate on this for a while, though I'll try to make time when I can.

I do remember some alternatives to RAVIGOTE, but I recall only GEMINATE from your list. LEVIGATE was another one, which frees up a valuable R, but as I recall, that wouldn't work for some other reason (perhaps because it uses an extra E, or an L).

Of the two routes you're heading down, I mentioned that neither one really yielded anything, but I do recall being more optimistic about the FLAVOR/AVO approach. Even so, I'm not sure I had any reason to like that one better. In any case, to get above 3966, I still tend to think (as I gather you do) that we need something quite different from what we have so far.

Good luck, and keep me posted on your progress.

I tried REINJURED and DITZ, but I kept losing like 25 points in that area of the board by playing FATTED instead of FLAWED/FLAMED. The W in flaw is scored 6 times there (AW, LAW, FLAW, WO, and FLAWEDx2), so replacing it with a 1 point letter, and then losing the scoring of a 4 letter word really costs some points that I can't find a way to recover elsewhere.

Today I came up with a 3955 solution that uses CRANKLING instead of CAROUSING, OUTFABLE instead of WOLFLIKE, and AmREETAS instead of ALAmEDAs. It lost some points because I have a blank as an "n" in REINJURES, and because it plays RIFT instead of RAFT (it doesn't get the points for the first two letters as their own word.) It looks better than the DITZ variation, but it doesn't look like I can squeeze out more than 11 more points without another major change.

*RIFT instead of REFT, rather. Still 3955 + 1 bonus for letter, so 3956.

I think I'm up to 3963 on the AmREETAS board, unless I've scored something wrong (which is always a possibility).

It's 3959, since I had forgotten to reduce REInJURES by 4 for the use of the blank. I'm working on something else interesting though.

Ryan,

I'm quite impressed that you've made some headway with CRANKLING. I've tried it a few times, but never got very far. I do have a vague recollection that OUTFABLE was also one of my candidates to replace WOLFLIKE. I also tried AmREETAS a long time ago (though with CAROUSING), but I gave it up for one reason or another.

As you've probably already noticed, RAMADAS is worth a look, but the several efforts I've made with that have all come up short. I remember thinking RAMADAN would work very well, but it's not a word in the Facebook Scrabble dictionary.

I'm pretty sure the three long words, and REINJURES, are "untouchables." At the bottom, either JIBB or JABB seems awfully tough to beat (especially since you really maximized the J with your discovery of that 4-point boost in that area). That leaves three or four areas, of course, as we've been discussing, but making those words "untouchable" restricts us greatly, especially for certain letters (such as N).

Sounds like you're getting closer than I'd thought possible, but I should have known: you've amazed me in the past. Keep it up and keep me posted.

http://www.megaupload.com/?d=5DOSBV0C

Ryan,

At the link above, you can find a blank Scrabble board I created using Excel. You just type in your words, one letter per square, and the table below the board automatically updates the number of that letter left (same with blanks: type "Bk" on the board, not case-sensitive). In the scoring sheet at the right, you just type in the "Word Score" each time and it keeps a running total. To distinguish your own score from the opponent's score, type an "X" in the column indicated, and the spreadsheet will automatically count that row's Word Score in the opponent's column.

More detailed instructions appear on the spreadsheet, and so you should be able to figure out how to use it.

The second tabbed worksheet is the same thing, only it includes 3 boards instead of 1 but does not include any scoresheet.

Neither of these helps you figure out new layouts, of course, but they are quite useful for keeping track of remaining tiles and keeping track of scores. Let me know if (1) you're able to download the spreadsheet; and (2) whether or not you find it useful.

Good luck.

I should add, on my blank Scrabble board:

In the "Letters Played" column in the scoring sheet, type the actual letters played on the board in that turn, not the number of letters played. The number of letters played will be calculated and displayed in column Z of that row, and column AA will display the number of letters left after that play.

http://www.megaupload.com/?d=QDB8HJ63

Ryan,

I replaced the spreadsheet with one that corrected a minor formatting problem. You'll need to use the new link above, rather than the one I gave you in my earlier message.

Neat spreadsheet. I had actually been using scrabulizer to track the remaining tiles for me, and to score the multi-word plays, and just typing everything into a spreadsheet.

I created a simple word checker function that checks the validity of a word. I think it would be useful added to your spreadsheet.

I created a tab named TWL 06, and put the words in column A, starting in cell A1 and going to A178691.

The formula (which should be in a single cell without a line break) is

=IFERROR(IF(VLOOKUP(AD2, 'TWL 06'!$A$1:$A$178691, 1, 0) = AD2, "Valid word", 0), "Not valid")

AD2 can be changed to whatever cell you want to type the word to be checked into, and you can change the messages in quotes to whatever you want it to say if it's a valid word or not. The name of the tab can be changed, and if you want the word list somewhere on the same tab as the boards, then the 'TWL 06'! part can be removed entirely.

...and the range in the formula would need to be changed to match the new cell range of where you put the words.

Have you encountered a solution that scores well, but involves getting really lucky with the tiles? I was playing one the other day that involved 9 tiles remaining, drawing 7 of them, and hoping that a specific two of them were in that draw, or I wouldn't be able to get the maximum score. I calculated that there was a 23/72 chance (about 1/3) of it not working out.

Nevermind. I found a better way to play that one that only has a 1/8 chance of not working.

Victory is mine! 3966 has fallen after only 23 days; the new FB record (for now) is 3969.

Wow! I'm amazed at that. I have to give you credit. Optimism + hard work, Good old-fashioned virtues yield great results. Congratulations.

TenzingHillary,

I feel like I've hit a wall for now, so I'm going to look at SIMPLIFY>OVERSIMPLIFYING next.

Getting the M and the P scored 9 times in the 15 letter play, plus some additional times seems very attractive. It's worth 30 more points than OVERIDENTIFYING. Since they're being replaced with lower scoring letters, it's likely that a few (but less than 30) points will be lost elsewhere when I'm now playing D, E, N, and T instead of S, M, P, L. Still, it's potential for a good 5-20 point gain.

The real issue, as you mentioned earlier, is the 100+ points normally earned from building up to CAROUSING or CRANKLING, which seems hard to replace. There aren't any legal 7 or 8 letter words I could find that have the P in the proper position, and can then have the C attached and still form legal words.

There's only a handful of 8 and 9 letter words even worthy of consideration, assuming we'll still want to use that many letters there to connect the two 15 letter words.

CALTHROP(S)
CANTRAIP(S)
CHANGEUP(S)
CHUMSHIP(S)
CLAPTRAP(S)
COOKSHOP(S)
CROWSTEP(S)
CITYSCAPE
COLLOTYPE
COLLOTYPY
CONSCRIPT
CREAMCUPS
CRYOSCOPE
CRYOSCOPY
CYANOTYPE
CYMOGRAPH

Just my thoughts.

Ryan,

"There aren't any legal 7 or 8 letter words I could find that have the P in the proper position, and can then have the C attached and still form legal words."

That was precisely why I gave up on OVERSIMPLIFYING. It does score well if you can play it, but it's awfully hard to figure out how to do that. I don't think your list is helpful, because a word in that position needs to be a word also before you put the C in front of it. Few if any words on your list fit that bill.

Ryan,

You might have in mind that you WON'T connect the second C in MICROTECHNIQUES to OVERSIMPLIFYING. That would solve the problem I just described. But bear in mind that SIMPLIFY has 8 letters, and so you need first to play a vertical word containing a P before you can play SIMPLIFY. You can do so, of course, without ever connecting that first word to anything, but it strikes me that you'd be wasting a lot of letters doing so.

Right, because if you play the C words above, then the H and the I can't be on the board, and then there's the issue of needing to play 8 letters to complete MICROTECHNIQUES, which ruins the whole thing.

I can score in the low 3900's with OVERSIMPLIFYING, but I really don't see it going much further.

I'm not sure where to go next. I may keep trying to improve my 3969 solution, but I'm basically out of ideas there too.

I'll be hard-pressed just to catch up to you, though I'm fairly confident now that I've been proven wrong for the umpteenth time that the record can go no higher than wherever it happens to be when I make that incorrect prediction. I will have limited time to figure out what wizardry you performed this time, so I'm counting on luck this time.

TenzingHillary,

I think I'm on the complete opposite side of the fence. I think I'll keep on thinking there's a higher limit, regardless of how high we get. I'll probably spend months trying for a higher score even after the highest one is reached (someday).

At least mathematics is on your side. There is a limit, and each new record we set is more likely to be it, and is closer to the limit if it isn't. If you keep thinking we've hit the top, you'll eventually be correct.

I'll proudly hold on to the first place position you held for so long until you catch up (or you or someone else surpasses me, which wouldn't surprise me). Hopefully our discussions here will spark some interest.

Do you think we should start a new topic? This was really was about the 1783 single turn solution, rather than maximizing the total for a player.

Ryan,

Not a bad idea to start a new topic, though it might be better to start it on Facebook. I don't use Scrabulizer for anything but our discussions, and I would guess that more people frequent the Facebook discussion groups, but I really don't know. I'll leave that up to you. If you do start a new topic, though, you should be sure to include a link to this discussion up front, so anyone who's interested can read what we've written so far.

I think my last message was unclear. I was essentially saying that, because I've been wrong so many times that the record couldn't go higher, I no longer think that's necessarily the case. Your optimism and hard work has now proved me wrong twice, as Jon Prentice had once before.

I have no idea whether I'll ever pass you up (chances are more likely, I think, that you'll move the record higher), but I'm all but certain I'll tie you sooner or later. I'm taking a very systematic approach, and I'll share whatever I learn in the process, but I think you'll find it more interesting once I've actually accomplished my goal with it.

I will mention one thing, though: CRAVENING doesn't look promising to me at all. I don't know how I'd not noticed this before, but, unlike CRANKLING and CAROUSING, one can't make a 7-letter word at the very start of the game: AVENING is not a word.

Good to hear from you, Ryan. Let me know if and where you create a new topic.

Ryan,

One other preliminary comment, on your effort to substitute something for OVERIDENTIFYING. As you noted, finding a candidate that has a P on the center square is very tough. There are, however, several alternatives to CAROUSING or CRANKLING that have an R in that spot - CAROUSERS being the one that looks most promising. I haven't gone back to check whether there are any OVERIDENTIFYING replacements that have an R on the center square, and I suspect I am forgetting that I previously checked out CAROUSERS with that thought in mind, but that may be worth a look.

Here's another very preliminary thought: Consider MANEUVERABILITY in lieu of OVERIDENTIFYING. D doubt it will yield much, but it's probably worth at least a look. Unlike OVERIDENTIFYING, it won't allow you to place a tile on both of the outside Triple Word squares in the same play, but you nonetheless can get two separate triple words. You'd play the R in another word on the first play (AROUSER, for example). Then play MANEUVER for one Triple Word. Then add ABILITY for another, longer triple word. As I said, I doubt you'd come close to the 293 for OVERIDENTIFYING, but it may be worth a quick look. There aren't many candidates to replace OVERIDENTIFYING that have an R in the center square and also have other key attributes: high score, no XHCQZ, and the ability to form two bingos in the course of creating the word. That, of course, is why one always comes back to OVERIDENTIFYING (so far, at least).

The letters in OVERIDENTIFYING are 26 points, while MANEUVERABILITY only uses 25, so there is one more point in the tile bag to somehow be used later.

I count 92 for MANUEVER, and 128 for adding ABILITY to it, for a total of 220. IDENTIFY scores 69, and OVERIDENTIFYING scores 293, for a total of 362. Starting out with MANEUVERABILITY looks like a 142 point deficit to overcome elsewhere. Essentially (not counting individual letter multipliers) 8 of the letters are scoring 6 times, and 7 of the letters are only scoring 3 times (in OVERIDENTIFYING, 8 of them scored 10 times, and the other 7 scored 9 times).

I think it would be really hard to make up those points, as we're not actually getting any extra letters to use anywhere else to score extra points, just different ones. However, it would be awesome to see it work somehow, as it would involve a somewhat different board. If you get a board using MANEUVERABILITY that scores above 3900, let me know, and I'll dive into it further.

Ryan,

I think you're right about MANEUVERABILITY. It looks like the "Big 3" are untouchable. I tend to think that goes for REINJURES too, with the possible exception of REINJURED. And it looks like either CAROUSING or CRANKLING. I may find out I'm wrong about REINJURES as I try to figure out how you got to 3969 (no luck yet, but 5 points away on another approach), but that's the way it looks now.

TenzingHillary and Ryan,

I've been playing Scrabble on Facebook with a few friends across the country for a few years now I guess and recently became curious about the OXYPHENBUTAZONE I kept seeing in the global stats. To make a long story short I ended up in Scrabulizer and read this blog discussion through. I thought enough hints had been dropped to try my hand at high word and high game. I'm hooked now. I did manage to play the 1783 high word so I then turned to the high game. It's quite a bit more challenging. A few days ago I made my first attempt and scored a sorry 3889, but it got me on the list at 6th I think.

I've read this blog again a few times and still need help. I have a 3943 solution worked out but I can't find the tricks to get higher. I'm playing horizontally and use the same three fifteen letter words as you talk about. I use TenzingHillary's C word in the third column, and I've determined the double double trick with the J. I use a modified ROTE ending that leaves only an O unplayed but I can't seem to find anything else to improve my score.

I'm more than curious about the four-letter yielding a whopping 63 above the XY, that just doesn't seem possible. Anyway, you both have me hooked now so I'm sure to keep working until I find something I've overlooked. Congratulations to you both, you've been having a very healthy competition. Hope to give you a run or at least catch up.

I have several various boards that I'm working on between 3959 and 3969 points, but I don't have one that scores exactly 3964. I'm thinking you've either got something novel I don't, or one of us hasn't maxed out our boards.

It seems like CRAVENING would score about 50 points lower than CAROUSING (the 50 point bonus, and some letter scoring would create the deficit, but the 8 and 9 letter words would score higher than their counterparts in CAROUSING). A 50ish point deficit would be easier to recover than the 142 point one mentioned previously. I say easier, of course, only as a relative term. Some other words like that are:

CENSURING
CENTERING
CHAUNTING
COFFERING
CRANCHING

CRAVENING
CREMATING
CRIMPLING
CRIPPLING
CRUMBLING
CRUMPLING
CUMBERING

On a different note, skipping a word in the WOLFLIKE position, while including one in the RAVIGOTE position, would cost us even fewer points than using RAVENING. I think we had dismissed that a while back, when looking at READJUSTS as an alternative to REINJURES. Of course, I think that's less likely of an option now than it used to be, as you know. Still though, if you're willing to take a temporary point loss, it's an option, and a relatively "cheaper" one at that.

I may look again at REINJURED - I still feel like there's potential for at least some interesting boards.

I thought about somehow connecting the first R in REINJURES with the first O of OXYPHENBUTAZONE somehow, so that I could get triple scoring off that when playing the 7 letter play. It would involve playing perhaps the X or the Y in advance, and not playing one of the other letters (the U seems most affordable, as you'd only lose 3 points from UTA).

I can't seem to make it work though, as it takes a few letters and doesn't seem to average enough points per letter to justify not playing them elsewhere. I haven't been exhaustive with it though, so maybe something would work in one of my variations.

Canad_ian,

Welcome. Glad to see you've become interested. A 3889 game is quite impressive for your first effort.

You might find useful the blank board I've uploaded at http://www.megaupload.com/?d=QDB8HJ63

"I did manage to play the 1783 high word so I then turned to the high game. It's quite a bit more challenging."

It certainly is. It took skill to figure out the 1783-point word, but Kyle Cronan (see above) supplied that skill. The actual playing-out of that word was easy (though I'd mistakenly thought that Kyle's board couldn't actually be played until Ryan corrected me on that). Achieving high GAME scores, as you note, is a whole different kettle of fish.

I see you've used my C word, and Ryan's "double double trick with the J." Both are useful, though I wouldn't conclude they're as "sacred" as the three 15-point words, which I am pretty well convinced must stay where they are (I think Ryan agrees, though he's experimenting with possible replacements for OVERIDENTIFYING). To me so far, REINJURES also appears to be inevitable, but I'm still open to being persuaded otherwise. I do, however, have a layout that does not use Ryan's "double double" enhancement (which nonetheless was quite impressive, as I've written above) and yields 3964. I'm also tempted at times to abandon my "C word," though I have not yet done so in any layout since I first figured it out.

"I'm more than curious about the four-letter yielding a whopping 63 above the XY, that just doesn't seem possible."

You'll probably figure it out eventually. When you do, you'll probably be tempted not to disturb it, given the high "local" score. But bear in mind that disturb it is just what Ryan did, ending up with a four-point boost in his total score that it took me many hours to figure out. As we discussed above, Ryan used one of the four letters used to yield the 63-point score in the lower-left corner to create his "double double trick with the J," which considerably reduced the 63-point score in the lower-left corner but more than made up for it near the J. I'll leave it to you to figure out the 63-point 4-letter combination, but caution you, for the reasons I just gave, that you should not necessarily use it. Nor, for the reasons I gave earlier, should you necessarily use Ryan's "double double J trick." I still am trying to use it myself in new board layouts, and am still considering using the 63-point score in the lower-left corner, but I'm nevertheless willing to forego one or both of those if I can find a better use for those letters.

It sounds like you've already picked up several hints, but I'll mention one again, since I've already mentioned it above, though not in "concentrated" form: The only useful 4-letter words connecting the J and the bottom B are JAMB and JIBB. That doesn't necessarily mean one must play a 4-letter word starting with J there, of course, but that's what it does mean if there is a J at that spot. This enables you to move just a few letters around to try each of those two words at that spot, with the moved letters usually ending up in the lower right corner (FLAWED instead of FLAMED, for example) or on the triple letter square just up and left of the B in OXYPHENBUTAZONE.

Finally, if you haven't done so already, you should look at the 3986-point "combined score" record game by Nathan Hedt and Stephen Root, which you can find on the Internet. You'll need to make many changes, of course, to yield a high "single player" score (the winner in that game had "only" 3690, if I recall correctly), but it's nevertheless important to study that board.

I hope this helps. As you can surmise, I think a few hints are OK but they shouldn't go so far as to take away the pleasure you'll get from figuring out the solutions on your own. (I'm still struggling to figure out better solutions myself.) Good luck, and welcome!

-- Tenzing Hillary

Ryan,

"I have several various boards that I'm working on between 3959 and 3969 points, but I don't have one that scores exactly 3964. I'm thinking you've either got something novel I don't, or one of us hasn't maxed out our boards."

That's interesting. I came up with that 3964 board pretty quickly, which encouraged me, but I've not been able to add a single point to it since then. (I keep hoping I inadvertently underscored it by a few points, but I've checked the scoring several times and it seems that 3964 is correct.) I will mention that it does NOT use your "4-point enhancement" near the J. I have another board, very similar, that does use that, but so far yields only 3962. All in all, it does seem we must be using different approaches.

"It seems like CRAVENING would score about 50 points lower than CAROUSING …"

I'm confused by this. As I mentioned yesterday, I think I was mistaken even to consider using CRAVENING. Because AVENING is not a word and RAVENING has 8 letters, you can't use it unless the N has already been played. That, in turn, would require playing IDENTIFY first, but IDENTIFY also has 8 letters, which means, once again, that the N must already be there. So it appears to me that CRAVENING is useless.

I've recently wondered whether CREMATING might work there, but haven't experimented with it. Obviously one is reluctant to use the M, though it would at least be on a double-letter square. The other candidates you mentioned (CENSURING 
CENTERING 
CHAUNTING 
COFFERING 
CRANCHING CRAVENING 
CREMATING 
CRIMPLING 
CRIPPLING 
CRUMBLING 
CRUMPLING 
CUMBERING) don't look promising to me, though I haven't entirely ruled out CENTERING, CENSURING, CRUMBLING and CUMBERING.

"On a different note, skipping a word in the WOLFLIKE position, while including one in the RAVIGOTE position, would cost us even fewer points than using RAVENING. I think we had dismissed that a while back, when looking at READJUSTS as an alternative to REINJURES. Of course, I think that's less likely of an option now than it used to be, as you know. Still though, if you're willing to take a temporary point loss, it's an option, and a relatively "cheaper" one at that."

I'm confused by this too. If you play a word in the RAVIGOTE position, but not in the WOLFLIKE position, how can you form REINJURES (or READJUSTS, if you haven't ruled that out yet), which has 9 letters? Certainly you could play the first R beforehand, but that would preclude the 4x multiplier. I see no alternative to playing a word in the WOLFLIKE position (assuming, of course, that you play a word in one of those two places at all).

"I may look again at REINJURED - I still feel like there's potential for at least some interesting boards."

I certainly haven't ruled it out either, though I, like you, haven't had much luck with it yet.

"I thought about somehow connecting the first R in REINJURES with the first O of OXYPHENBUTAZONE somehow, so that I could get triple scoring off that when playing the 7 letter play. It would involve playing perhaps the X or the Y in advance, and not playing one of the other letters (the U seems most affordable, as you'd only lose 3 points from UTA)."

I've also toyed with connecting REINJURES with OXYPHENBUTAZONE – for example, using RASP, then OS, so that OY and RASP would be formed when OXYPHENBUTAZONE is played. Another variation yielded RASP, SO, OH/AH, SOD/SAD/SOW/SAW (depending on what letter ends up on the triple letter square to the right), making it unnecessary to play a letter just above the first N in OXYPHENBUTAZONE. I concluded it was NOT worth hooking up the X, since you can do better after OXYPHENBUTAZONE is played by playing on the double-word square. For reasons I can't now recall (presumably: no point gain), I abandoned all of those efforts.

"...using RASP, then OS..." should read "using RAS, then OS..."

I was saying to play "IN" first, as part of RAVENING, which would then leave the "N" in place to score IDENTIFY as normal. It would cost the 50 point bonus that's normally earned when playing ANKLING/ROUSING, however.

I had mentioned a while back, and was just re-iterating, that you could play RAVIGOTE or another word that has the I and the E in those positions, and not play a word in the WOLFLIKE position. The first E in REINJURES would be available, so you could play REINJURE then play SPA. You wouldn't get the full 4X benefit for REINJURES, but the point loss would only be like 20-30, which is less than the point losses we were discussing for some of the other alternatives.

Ryan,

Thanks for clarifying that. I've revisited RAVIGOTE a little lately, but the opportunity cost always seems too high.

I tried going the other way with RASP as well, making WOOS/OX/OY (through several turns), but still wasn't able to come up with anything that didn't lose 10-12 points.

TenzingHillary,

Do you want some hints on my 3969 board, or are you wanting to enjoy the challenge?

Because I'm still curious on what you're doing on your 3964 board, since you think it's optimized.

Ryan,

Thanks for the offer of hints, but you're right that I prefer to slog along on my own. I haven't had any luck yet, though my time's been limited. I did come up with a slight variation of my 3964 board that also yields 3964, but both boards appear to be maxed out. Based on your success with my earlier 3962 board, though, it may be that it's still got potential for more that I'm just unable to spot. My hunch, though, is that your board is a little different from mine -- which seems especially likely since, as you say, none of your boards yields exactly 3964.

I'm still confident I'll get there, but the length of time it's taking me is just further testament to your accomplishment. You deserve great credit.

Thanks for the praise - your work was instrumental to my success, and I still think you take the lead again eventually if you keep working on it. I'll keep working on my boards then, but I am very interested in your 3964 boards.

I'm going to go through all of the REINJURES alternatives, though I'm expecting, as you've said, that REINJURED/REINJURES are the only viable options. It will still be good practice, I guess, and it might yield some new insights.

Oh, I was working on the 1783 board some the other day. By replacing DOER/DORK/BOD/DE with "FORTH", and by replacing ULna with "pUL", I can have 6 tiles leftover at the end of the game. I don't see a way to score anymore points though, and I'm inclined to believe that it has been maximized.

Ryan,

I'd used "FROTH" when I played 1783 -- same thing as your "FORTH," of course -- but hadn't used pUL instead of ULNa, so I ended up with 5. I suspect Kyle Cronan was just trying to make the necessary connections to yield his 1783. You may recall that I mistakenly thought you needed 7 tiles left at the end, which is why I'd tried to optimize it. But by the time I'd got it up to 5 remaining tiles, you'd pointed out above that you don't really need to do that.

I too think 1783 is tops, but that counts less coming from me than from you, since my sense of the "top" has been mistaken several times so far.

Well, I haven't done any exhaustive work on the 1783 game. I'm just presuming that Kyle's claim about his program was accurate, as the program would probably be simple enough to write and small enough to run with such a limited list of words to test.

I think I might have figured out Jon Prentice's 3958/3959 solution as well (it's my 3959 solution that I mentioned earlier. I just remembered that my last improvement to it was squeezing out one extra point). It may not be right, of course, but I think it's optimized at 3959 for the choice of 2nd tier (7-9 letter) words.

Do you think a program can determine the maximum high GAME score? I certainly can see that for high WORD, since the choices are fewer, but high GAME seems hard to program.

I'm pretty sure the programs aren't powerful enough to solve the game for the highest score from the beginning, as the game has so many potential moves. If enough assumptions are made, I expect a program could come up with solutions though. Scrabulizer checks all possible one-turn solutions, for example. The highest scoring two-turn solution might take 30 times as much computing power, so say a minutes instead of 2 seconds. After five or six turns at that pace, it would take days to come up with the highest solution.

However, it might be feasible to place the first 80-90 tiles, and have a program solve the highest total that can be scored from the remaining moves. Most programs for Scrabble are written based upon the expectation that there's a 2-4 player game going on, and that the program user is more concerned with either maximizing each single turn score or with winning the game, even if that means scoring less in total in order to play better defense. There's a much smaller audience for what we're doing here compared to the millions of players who play Scrabble and similar games, so it's likely there aren't many programs (if any) that are trying to do exactly what we're doing.

Ryan,

Sorry I forgot to answer your questions about my 3964 layouts. Both use TREATISE,RANKLING,OUTFABLE. The two layouts vary only in the placement of a P, a D and a blank. I have another layout that yields 3962, using your "double double J" enhancement, and yet another that is essentially the same with a slightly different sequence in the lower right corner, which yields 3961. Obviously, the efforts necessary to achieve your "double double J" aren't worthwhile for those two layouts. The two 3964 layouts (or at least the first of them) occurred to me very quickly once I'd switched to RANKLING and OUTFABLE, but, as I've said, I've been unable to add even 1 point since then.

Ryan,

A couple of clarifications: (1) "RANKLING" should be "CRANKLING" in both places, as you'd probably figured out; and (2) my two layouts that use your "double double J" enhancement are not just "essentially the same." The board is actually identical in both; only the sequence of play (in the lower right corner) varies between them.

3964? I had maxed out at 3963 on that layout. In fact, I had previously thought I'd maxed out at 3965, then realized I'd double counted a two point word. It appears there is more work to be done!

Just to clarify, my board for that layout was 3962 before adjustment for a remaining tile, which would score 3963 on Facebook, and 3964 under the set of rules that doesn't reduce player 2's score.

Ryan,

I just checked the scoring again, and it comes up to 3964 (3963 before the one-point bonus, under Facebook scoring).

I'm not sure why you're coming up one point short. Are you playing HEW/AW and JAMB?

You might also double-check the scoring on MICROTECHNIQUES. As you know, that varies, depending on what small words you have hanging down from it.

Found it. I had 3964, but had incorrectly noted 3 instead of 4 for a play. I hadn't actually played this board out on Facebook, since I developed it after the 3966 board. I did figure out how to play it in about 2 hours though (using a fast computer with a very wide screen and a lot of ram).

In regards to your comments, I hadn't explored the P move, though I now see two different D placements that work with it. Still no gains yet though. I really feel like there's more potential here, since a blank can be moved around, unlike my other boards.

My 3959 board is now 3960, by the way. Apparently I had copied and pasted that error when copying a series of identical moves.

Ryan,

Glad that's cleared up. As you noted, this layout is unusual in that a blank is left over at the end -- rare, indeed. Nonetheless, whether I move it around or just play it, I end up with 3964. My "movement" of the blank, incidentally, had it taking the place of the P, which in turn took the place of a D, which in turn formed a 9-point word in the lower left corner. When the dust cleared, the net change was a whopping zero.

I came up with another 3962 solution today (using a different combination of words in the CAROUSING/ALAMEDA position). I still have three more of those to check, but these were the combinations that didn't look very promising.

Are you finding anything exciting?

I got an email from the guy who has the Scrabble FAQ (I had emailed him a while back). He's corrected the error in the 3986 board that's in the FAQ, and I mentioned to him to update the 1782 score to 1783 score, and to include the record we're working on as well.

Ryan,

I'm still having no luck with 3969. It's tougher this time. Last time, you'd started with my own layout and just tweaked it - tweaked it extremely impressively, but at least I knew what you'd started from. This time, I don't know exactly what layout you started with. Still, no hints, please. I'm confident I'll figure it out sooner or later, frustrating though it's been. Of course, you might be at 3970 by then.

I've wondered, by the way, whether it might be feasible to construct, at the top of the board, an equivalent to the "J" multiplier at the bottom, placing, for example, a W in the corresponding position, forming CROW or CREW or CRAW coming down from the top row. Horizontally would be a word in the position corresponding to REINJURES. Obviously this would require a radical departure from the boards we've been working on, and my limited experimentation has yielded no encouragement so far, but if this ever did pan out, the possibilities could be considerable. Consider how many points get generated, after all, by the J.

Our boards do great on the Z, Q, and X, and I think great on the J as well. The goal would be to better maximize the next tier letters (one of the V's, one of the W's, one of the M's, one of the B's, and one of the D's). Improving the number of times those are multiplied will yield our biggest improvements.

Finding a way to score one of the W's like the J is definitely interesting, but only yields have the pointts (36 points would be gained by scoring the W 13 times instead of 4, for example). Definitely worth looking into, but are you also trying to keep the same number of 50 point bonuses?

I'm sure you'll figure it out my solution if you're going systematically, but I'm more hoping you find something different that scores more. My hope is that you'll break the 3969 barrier, and we'll keep increasing the score from there. Of course, I'm also hoping others find this and join in the effort.

It's funny you mention CROW, as I remember looking at "CROWSTEP" and "CROWKEEPER" (and related words) when I was looking at OVERSIMPLIFYING instead of OVERIDENTIFYING, but I didn't find a suitable word that formed a new one when a "C" was added to the front.

I'll give it a shot at some point, after I finish maximixing a few other boards. I think I have about eight boards now that score 3960 or more, so that obsessive part of me still wants to squeeze a miracle out of one of them.

"only yields half the points" rather (sorry - I was interrupted a few times while typing).

Well, I've managed to learn a few of your tricks and have improved my posting by 65 points to 3954, which places me fourth for now. I'm not sure I'll ever better that but figuring out the 3959, 3962, 3966 and 3969 boards posted mentioned is a puzzle hard to let go of. I'm giving it a rest for awhile though, waiting an hour for one letter was a killer.

Congratulations canad_ian. Glad to hear you're still hooked. A rest will do you good, but it's hard not to get back into it, especially when you know higher scores are being achieved (especially Ryan's 3969, which I've yet to figure out).

Yes, congratulations! Welcome to the discussion. I'm delighted to see someone else involved. 3954 is a great score, because it passes a benchmark that existed for quite some time, yet still has room for improvement to be really motivating.

I'm sure rest will be good, but once you discover that next big score, especially if it moves you in the ranking, those hours don't seem so bad.

In case this helps you in future games, my tile drawing strategy is to swap all 7 of my tiles for the one or more tiles that are the most rare in the tile bag. For example, if I'm drawing for OXYPBZE, and I get OOOEEEE, I would trade all 7 tiles again on the next turn, in order to have a better chance at getting the Z and/or the X. Once I have the Z, I swap for the X (or the next most rare letter(s) I don't have, if I get it on the same turn as the Z) until I'm swapping one tile for the most common letter. That way I get several chances per turn for the rare letters by swapping several tiles, and several chances per turn for the more common letters, by drawing for one of a few copies of those letters. Near the end of most games, though, you're going to be drawing for a set of mostly unique tiles, which does take a while, especially when you have 5 or 6 of the 7 you need, and there are 7 or more tiles in the bag.

Also, once you swap some letters, the other player can immediately go to "my turn games" (whatever it's called) and select the game, then pass or take whatever action is necessary. When I played my 3953 game, I actually waited the 15-60 seconds every turn to be told that Player 2 could go, and it took like 8-10 hours to play out the game. Still exciting, but a very long day.

I had a long reply penned but as this site didn't like my "two words" I'm hard pressed to repeat them other than to say I stole all my moves from this blog and am especially thankful to Ryan's comments before his landmark post.

My game took 4 hours, 57 minutes with over an hour waiting for the K.

I'm able to play much more quickly than either of you are describing.

I have both players' Scrabble games open, on separate browsers, as I assume you do too. I use Chrome and Firefox, and assume here I'm using Chrome for the winning player.

Assume I'm swapping on the Chrome browser and passing on the Firefox browser. In Chrome, I swap my letters, immediatelyI click on the Firefox window, and then hit the "refresh" keys (Apple-R, for me), which speeds up the "transfer" from one player to the other. I then click "Skip" and confirm by clicking "Yes." Then I click on the Chrome window again, hit the "refresh" keys there and repeat the process. After a short while, I figure out exactly where my pointer needs to be while I'm waiting for a button to become active (for example, the "Swap" button on the Chrome browser). Before long, I'm down to about 15 seconds average (30 seconds "round trip").

My hunch is that the big difference may be my hitting "refresh" the moment after I click on the other browser. If you simply wait for the Scrabble game software to switch from one player to another, the wait is much longer. Even if you do hit "refresh" there's still a bit of a wait, but it's much shorter.

I hope this helps.

I think the last game I played out (3966) took a bit under 3 hours, though I can't remember for sure. Incidentally, as I'm sure you noticed, the wait time varies considerably. My last game, I played my first word (all 7 letters) after only about 4 swaps, which I thought was an omen of good luck. That was not to be, though. Other words took as much as 30-40 minutes.

I actually got my play down to just over 2 hours on my 3969 game. I was just talking about when I first started playing and was trying to set my 3953 game (which took multiple tries, due to an error I made in the first game).

To anyone reading:

I had a breakthrough last night! After 19 more days of work since my last one, I came up with a board that scores 3983 for player 1 under FB rules (3984 under Official/international rules). I just finished playing it. This board also sets a new record for total game score versus the Scrabble FAQ. I'm going to spend a while trying to optimize it further, but as always, it initially looks pretty tight.

Congratulations, Ryan.

Just when I'd hoped to catch up. Great credit where due. That's an amazing accomplishment. 4,000 next?

Ryan,

In a way, it's a relief to hear of such a high score. Now, at least, I know there's no point wondering whether maybe I can squeeze an extra few points out of the lower right corner by re-sequencing the play there (as I've done many times, and suspect you have too in the past). Clearly you came up with something radically different, so I'll have to consider a whole lot of options.

Now that your 3969 game is "history," I would be curious to hear a hint as to how you got there -- but ONLY if you can give that hint without also giving anything away about your 3983 game. You deserve immense credit for that, and if anyone ever matches or tops it, they should do so on their own.

Yes, the quest to pass 4000 is ongoing. I was fortunate that this discovering increased both my total and my one player records. I was worried that I might somehow find a way to score 3975 or so but still not break the 3986 total score from the FAQ.

I'll probably keep playing as long as it looks interesting, but getting above 4000 on Facebook seems like a nice place to take a break. I have been looking at this practically every day for almost two months now. Eventually I expect I'll try to get my solutions published in the Scrabble FAQ.

I'm not sure how much of a hint to give, so I'll speak in generalities. The 3969 board involved larger scale changes than the 3962>3966 change. The 3983 game did as well, but was more of a natural progression from the 3969 board, including using one variation I had previously declared not useful on prior boards. I hope that helps.

With all my typos here, I realize how much I've become conditioned by Word and its grammar checking. I meant "discovery" rather than "discovering" above.

Ryan,

I hope you'll reconsider whether to get your solution published in Scrabble FAQ. The "high word" contest is uninteresting because no one has really done anything but play out Kyle Cronan's published solution. People can learn some things from reading our exchange here, which is fine, but coming up with a "high game" solution -- especially yours -- remains something that requires skill and hard work. It's your call, of course, but I vote for you keeping your solution secret. You can simply tell Scrabble FAQ what your "combined player" score was, and they'll undoubtedly concede that it is certainly above the old record total, since you're only two points short of that with your single-player score. If they want to "asterisk" your record, so be it, but it would be difficult for them to deny that you've broken the record.

Remember: you can always change your mind and publish your solution, but you can't change your mind and "unpublish" it.

TenzingHillary,

I won't be publishing anything anytime soon, and I will consider the decision thoroughly before doing so.

There's a program called Quackle which works very similar to the Scrabulizer applet, but seems to have some multiple move solving capabilities. I don't think it will be able to construct a board from scratch, but it might be able to confirm how to maximize the points from a few "loose tiles" (the ones that aren't structurally necessary for the board, but are just distributed to score points.

I think I'll just try to find a way to score 4 more points, because 3987 on Facebook guarantees that 3986 total has been broken.

Ryan,

I'm not familiar with Quackie (or the Scrabulizer applet, for that matter). I can't imagine they do anything you can't do with word lists and blank boards, can they?

As you know from prior comments, I'm skeptical that computer programs can really outdo humans in "high game" playing, simply because there are too many variables (not the case for "high word" play, where I can imagine programs would help). As one nears the end of the game, I suppose, there are fewer variables, and so I can imagine a program helps, assuming it allows you to plug in the state of the board.

I should mention that I was once skeptical about computer chess players too, but I gather they've outplayed many grand masters.

Scrabulizer is the site we're on now. I had mentioned it before when you provided your spreadsheet - I thought you knew of it. If you just go to scrabulizer.com, you can type in whatever's on the board, then type what letters are in your rack, and it will tell you the highest scoring moves available. I was just referring to it as an applet, as it's not a separate application, but rather built into a page. It's what was used to generate the graphic that is at the beginning of this post.

Since computer programs are powerful enough to see every possible move for a given turn, and can play a pretty good endgame, it makes sense that they would rival the top human players when both players are competing for high score. They typically aren't designed to assume one of the players will pass continuously. Most programs will always generate the moves that yield the highest score each turn, or the move that best balances score and blocks the position for the other player.

I was thinking about it my FB total score on my 3983 game.
Player 2 had to have at least 5 points on the rack at the beginning of the game.

Those points were not counted in the 3983 if they were played on the board, and they were only counted once in the 3983 score if they were unplayed. Since official rules count them twice in the total, at least five points aren't being counted in my 3983 score, regardless of whether they were played or not. I therefore argue that Facebook has verified a minimum of 3988 for my game in which Player 1 scored 3983, which does set the new record for total score.

In actuality, of course, my total was higher than 3988, but no one can argue I didn't score at least 3988. Does that math seem right?

3983! Wow! Just when I think the target is in sight it gets moved, ha. I thought about your 3983 and it inspired me to try again. I shortened my game play time to four hours and posted a 3963. I'm hoping something will come to me in my dreams to better that but I'm happy to be third anyway. Again let me say that the constant improvements you post are inspiring Ryan.

It sounds like you're coming for us! I love it! I guess I can't take it easy yet. Any words or strategies you'd care to share that seemed to work really well for you? I'm always looking to hear new ideas.

You and I know that Player 2 always plays all of his tiles except 1. Nonetheless, unless I'm missing something, you'll still have to rely on the Scrabble FAQ people taking your word for it (I certainly would). At least in theory, it's possible that Player 2 never played a turn. In that case, your 3983 (under Facebook rules) would include whatever points Player 2 had in his rack at the end, and Player 2 would end up with a negative score equal to the aggregate points in his rack Under North American tournament rules, your score would include double the aggregate points in Player 2's rack and Player 2 would end up with zero.

I certainly wouldn't need convincing, if I were running the Scrabble FAQ site, but that theoretical possibility remains.

canad_ian,

Congratulations on your 3963. Though Ryan's 3983 is way out ahead of us, bear in mind that your 3963 would have given you the top spot only a few weeks ago.

I have the same reaction as you to Ryan's 3983 -- or reactions, I should say. First and foremost: an extremely impressive accomplishment, for which Ryan deserves great credit. Second: deflating. Third (after the passage of some time from Reaction #2): inspiring. At least now we know such a score is possible, and so we can and will go after it.

Who knows? Maybe even 3984. I'm forever cured of my bad habit of saying it can't go any higher.

RyanM561,

About the only "words or strategies" I can share are that I'd still be at 3889 if I hadn't carefully reviewed all the hints found in this blog. It's funny but most of my increases have come to mind just lying in bed thinking about it not actually looking at a board. Every score I've had at first seemed impossible to top and then in a flash a little wrinkle adds a few more points. I'm topped out until I have another "aha" moment.

TenzingHillary

What's the Scrabble FAQ site? I take it from the conversations that this site gives honour only to the highest combined score. What is that at present?

3983 seems awfully hard to beat but from the previous conversations I guess 3953, 3958, 3962, 3966 and 3969 seemed equally difficult but have all fallen to Ryan and the majority to yourself as well. I still have three to match before even thinking of higher.

canad_ian,

The Scrabble FAQ site lists only the "combined score" record, which is 3986 under the scoring rules they use (which give the player who goes out double the tile scores of the other player, but subtract nothing from that other player's score). That score would be 3985 under Facebook Scrabble rules (which add the other player's tile scores to the score of the player who goes out, and subtract the same number from the other player's score). Presuming as I do that the "other player" had 1 point left when Ryan went out in his 3983 game, his winning score would have been 3984 under the Scrabble FAQ scoring rule. We don't know what the "combined score" would have been, but I have no doubt it's well above the old record, as Ryan says.

Glad to hear you're hooked. Also glad you've been able to benefit from this discussion, though it sounds like you're glad that the hints have not been so extensive that the fun is taken out of it. (Lately, of course, you and I are left totally in the dark as to how Ryan scored 3983, but that's as it should be.)

The FAQ is here:

http://home.teleport.com/~stevena/scrabble/faqtext.html

Arghh (and that's a Scrabble word)! I spent over three hours thinking I had a second place game but I'd screwed up the endgame and it was for naught or might as well have been. I did manage a replay in under that time which scored 3966 so I'm not totally disappointed. Okay Ryan, where the @$#! did you find 17 more?

Ryan and Canad_ian,

As I mentioned earlier, Ryan's 3983 game is liberating in a way, because it makes clear that something more radical is required to get there. No more tweaking a small section of the board to squeeze out an extra point or two. I'm more and more confident I'll get there eventually, though I suspect it will take a while. Nonetheless, I can't stop being impressed, even amazed, that Ryan managed to add 14 points to his already impressive top score.

I should mention, by the way, that all three of us (and several other Facebook players) are about 1,100 points above what the record used to be about three years ago, when I first started working on "high game" play. I think Jon Prentice had the record back then, but it was somewhere in the 2800's, if I recall correctly. Jon rose to the occasion once I'd pushed the record higher, and he now sits at 3959 himself. But Jon and I (and later others) added those 1,100+ points with a lot less effort than I suspect it took Ryan to add the last 14. That is why I find Ryan's accomplishment more impressive.

In my effort to top Ryan's 3983 game, I've reverted to a disciplined, back-to-basics approach. As a first step, I've decided to strip away whatever I don't know for sure should be there. As a second step, I've focused on the high-point letters that remain after the "locked-in" portions of the board have been played – figuring out which of those remaining high-point letters aren't receiving high multipliers when they're placed where we place them in our mid-3960s layouts. As a third step, I've identified multiple possible layouts that might enable me to maximize the multipliers for as many of those high-point letters as possible, assuming that I am able later to find actual words for those layouts that place the high-point tiles in good positions. As a fourth step, I've actually laid out half-completed boards with these possible layouts, but with only an "X" placeholder letter at each spot in the layout (other than the "locked-in" spots, where I play the actual letter). My final step, which I'm at now and expect to be for quite some time, is to experiment with those numerous layouts, always trying to avoid the temptation to charge too soon down some apparently promising road.

As you can imagine, the alternatives start to multiply once one reaches the third step, even more so at the fourth step, and especially once the final "experimenting" step begins. Nonetheless, this approach is helpful for me because (1) the first step narrows my options somewhat, but not more than I can "prove" is warranted; (2) the second step focuses my attention where it should be: on maximizing the "multipliers" on letters for which high multiples will make the biggest difference; and (3) the remaining steps keep me from charging prematurely down some dead-end road. This, of course, leaves me standing at the "crossroads" a great deal longer than I might like, but it beats taking the wrong road.

On the first step, I continue to believe the "Big 3" 15-point words are the best, with MICROTECHNIQUES across the top, OVERIDENTIFYING across the middle, and OXYPHENBUTAZONE across the bottom. (As I've mentioned before, I play my 15-point words horizontally, an arbitrary choice; I recognize they can just as easily be played vertically.)

I'll give my reasons for each choice.

MICROTECHNIQUES: I long ago looked very carefully at the possibility of switching the positions of MICROTECHNIQUES and OXYPHENBUTAZONE, and concluded it can't be done. Among other reasons, there are no 2-letter or 3-letter words that remain a word when Q is added to the end, which precludes putting MICROTECHNIQUES along the bottom unless one is willing to sacrifice many points from the vertical Q word (and can figure out a way to avoid it when connecting to the U or E in MICROTECHNIQUES, one of which letters must be in place before MICROTECHNIQUES can be played). By contrast, if MICROTECHNIQUES is played along the top, several words "hanging down" from it are possible, such as QI, QAT, QAID, QADI. And that is why I leave MICROTECHNIQUES at the top.

OXYPHENBUTAZONE: Certainly there are good arguments for placing OXYPHENBUTAZONE along the top instead of along the bottom – as it is, of course, when played as a "high word." But I can't find any way to maximize the J when I do that – which, needless to say, gets multiplied very many times when it is played where we now play it, 3 squares above the B in OXYPHENBUTAZONE at the bottom of the board. There are two three-letter words – JIB and JAM – that remain words when a B is added to the end. By contrast, there are no three-letter words ending in J that remain words when a B is added to the front. Nor are there any 4-letter words with a J in the third position that remain words when a B is added to the front. Thus, unless one can come up with an alternative for placing the J that matches or exceeds the very high J score yielded by our usual placement of it, it is inevitable that OXYPHENBUTAZONE will be played along the bottom.

OVERIDENTIFYING: As Ryan and I have discussed here, the only attractive alternative that comes to mind is OVERSIMPLIFYING, which replaces IDENT with SIMPL. But that places P on the very center square of the board, which makes it virtually impossible to connect in any useful manner with the second C in MICROTECHNIQUES at the top of the board. (I'm leaving out details in this argument, but I'm confident you've both examined this closely enough to know what I mean.) That leaves OVERIDENTIFYING.

A tougher call is REINJURES. Clearly REINJURED is an alternative, which I've examined and am examining again. The only other alternative I see is READJUSTS, which I haven't ruled out but nevertheless looks unattractive at the moment. The problem with READJUSTS is that neither of the "connector" letters from above (E and T) can form a two-letter word (other than RE at the beginning, but that would require playing an R on the double-word square, depriving one of the 4x multiplier that is the whole purpose of the exercise). Because 2 letters need to be in place before one plays READJUSTS (since READJUSTS is a 9-letter word), this requires 2 separate words descending from OVERIDENTIFYING. That is what was done, of course, in the 3986 "combined score" record game – RAVIGOTE and WOLFLIKE – even though READJUSTS was NOT then played in that game. But two 8-letter "connector" words use up a lot of letters. This is precisely why, as you know, I did away with RAVIGOTE some time ago, leaving me with only WOLFLIKE to descend toward the bottom of the board. That is possible as long as one uses REINJURES (or REINJURED) instead of READJUSTS, since one can play the second RE combination before playing the remaining 7 letters in REINJURES (or REINJURED). This makes it unnecessary to play any word in the RAVIGOTE position. It doesn't necessarily mean one should NOT play a word in the RAVIGOTE position – which does, after all, yield many points – but it does mean one need not do so. I'm still considering a word in the RAVIGOTE position, which would make READJUSTS possible, but my hopes aren't high for that approach at the moment.

Another alternative arguably justifying the use of READJUSTS is one that Ryan discussed here earlier: playing a word in the RAVIGOTE position (but not in the WOLFLIKE position), then playing READJUST for a 2x multiplier, then playing READJUSTS/SPA. But, as Ryan noted, that significantly reduces the value of the J, which drops from an 8x multiplier to a 6x multiplier in this two-step play (setting aside Ryan's clever "double-double J" enhancement, which would be impossible in any even with this two-step approach).

So, bottom line, what I consider to be "locked in" are the three 15-point words, with REINJURES (or REINJURED) close to that status but not quite there. REINJURES (or REINJURED) will be added to my "locked in" list if I can't pretty soon find a justification for READJUSTS, keeping in mind its limitations just described. I haven't ruled that out yet, and my back-to-basics approach will (I hope) keep me patient enough not to rule it out prematurely, but I will say that I am not optimistic at the moment that I'll settle on READJUSTS in favor of REINJURES (or REINJURED).

There are other open questions, of course, in addition to the one I've already mentioned (REINJURED/DITZ, as a possible alternative to REINJURES/SPA) – such as the relative merits, in the lower right corner, of FLAWED, FLAMED and FLAVOR, as Ryan and I have discussed here earlier. Once again, though, I'm hesitant to pick any one of those three alternatives prematurely, since that choice inevitably affects the remaining steps in the analysis. For example, if one chooses FLAMED, there will be a W and a V available for use elsewhere; if one instead chooses FLAVOR, there will instead be an M or a W. I don't think it makes sense to choose among FLAWED, FLAMED and FLAVOR until one has also considered how useful the "freed-up" letters will be elsewhere. I've played around with all three combinations in the lower-right corner long enough to know the pros and cons of each very well, and so I intend to just keep all three in mind as options until I've narrowed the choices elsewhere on the board.

I recognize this leaves me far from a solution that will get me to Ryan's 3983, but I think it's both necessary and stimulating. Necessary since, as I wrote above, mere tweaking pretty clearly isn't going to bridge the huge gap from where canad_ian and I are now to 3983; it might get us a few more points, but nowhere near 3983. Stimulating because it feels good not be charging prematurely down some road that I find will yield only a few more points at most.

Any comments you have on this will be appreciated, though I would ask that Ryan not give any comments that would shed light on his 3983-point game. I can only speak for myself, but I think canad_ian agrees that it will be much more satisfying for us to figure out on our own (if ever) how to match that score. In addition, if we do get there on our own, there will be a greater possibility that we can top 3983 rather than simply match it, since it's very likely that Ryan has already optimized the layout he used to get there (or will before we get there, if he hasn't already).

Well said TenzingHillary. I'll need a lot of time and a fresh approach to further my quest of a higher score. Ryan "you da man."

This new approach is bearing fruit. I've got a 3978 game, within shouting distance of 3983. As usual, I see no prospect of squeezing more points out of it, but I've really just begun trying. As I know from Ryan's jump from 3962 to 3966, not all enhancements are immediately obvious.

Excellent! If you beat 3983, I still want some kind of hint though.

Will do, but I've got to get there first. I need not tell you that the last few points are the toughest.

Well, it's possible you've already made some kind of change that would bolster my score, but as we've discussed, it's better for you to advance it as far as possible before hinting it to me. Of course, if you take the lead again, then we'll know you have something better. I guess that push from Ian was all you need to break through the next level.

Ryan,

It's indeed great to have Ian here, though the real push has been your 3983 score. If you hadn't arrived on the scenes, I'd have continued to believe my 3962 was the highest possible.

Incidentally, I'm in the course of playing out my 3978 game, though I've yet to pick the letters necessary for even the first word. I'm hoping that, before I get very far along, I'll have figured out how to boost it higher.

TenzingHillary, how did your game go? It hasn't updated on Facebook as of this post. I've been polishing my existing boards some, but I haven't come up anything really new yet.

I'm starting to think I may be at the point where a more radical innovation is going to be needed again, but I haven't thought of anything else yet. I'll keep optimizing the boards I have to cultivate my inspiration.

Ryan,

I have just gone through over 200 "swaps" trying to get the Y for IDENTIFY (my second word), still with no luck, even though the first 6 letters showed up in about 2 minutes. I've vacillated in any case about playing out a "second place" game, which I've never done, and the longer I wait, the more likely I'll figure out how to add 5 or 6 points. On the other hand, 3978 is impressive enough, and so I'm all but certain I'll settle (at least for now) for a very respectable second place if I can't think of some clever enhancement in the next day or two. Certainly on Saturday, if I haven't played it out by then.

It tooke me 30, 26, and 35 swaps to play my first three turns (each was a 7 letter word).

Each was a 7 tile play, rather.

Ryan and Canad_ian,

I will mention one change that may be of interest to you, though it's possible Ryan's 3983 game used the same change: My game does not use REINJURES, but does use one of the two alternatives we've been discussing lately. I have a hunch that Ryan may have made the same change, but that's mere speculation.

I'll also mention that, at the very end, I'm left with 3 "free letters" to do whatever I want with -- either WAS or WOS -- but I need to come up with 55 points from those three letters to tie Ryan's 3983, or 56 points to top that score. So far, all I can come up with is 50. While I don't see any prospects at the moment for increasing that total, I'm mindful of the valuable lesson Ryan taught me when he managed to squeeze 4 more points out of my then-record board, boosting the record from 3962 to 3966. That lesson was that it may be necessary to make some more comprehensive substitutions of letters at the end -- not merely looking for a better spot for one or two tiles.

I'm hoping some brilliant idea comes to mind before I'm too far along in the game play to take advantage of it. Even if I have played out the game too far by then, I'm sure I'll just start over if something better comes to mind. Absent that stroke of brilliance, though, I'll just play out the 3978 game. Wish I'd thought of it a few weeks ago -- just proves the old adage: Necessity is the mother of invention. "Necessity" in the form of Ryan's 3983-point game.

Hmmm.... for some reason I was expecting you to say you had an "L" as well. Let me see what I do with this information.

Ryan,

I'm probably making too much of your observation, but I looked at my board to figure out why you thought I might have an L too. I did use an L in the upper right corner, to make ILL coming down from the top. Just a speculation here, but I'm guessing that, if you had an extra L, you played a word there that doesn't require a third letter for any of the words formed in the second row. At the moment (without a word list in hand) only one such word comes to mind: REEDITS. I didn't play that there (or anywhere), though I certainly considered it and still am. The fact that you ended up with an extra L suggests to me that you may have.

No need to comment, especially if doing so would give me a hint. I'm just thinking out loud here.

Ryan,

As is probably obvious from my mention of REEDIT, my game uses CRANKLING rather than CAROUSING.

I wasn't saying I had an extra L, I was just saying that if you had an extra W, S, and A/O, I would have expected you to have one. ILL explains it.

Ryan,

Now I'm even more impressed that you figured out I should have an extra L (but for ILL). I'm beginning to think we used the same approach, though, if that's true, obviously you've made 5 points more from it. I'll keep trying, but so far no breakthrough. I may end up with just 3978 for now (if I ever draw the Y I need for IDENTIFY, that is).

Ryan and Canad_ian,

No brilliant ideas came to me, and so I played out my 3978 game. As I mentioned, I'd been reluctant to play out a game that didn't put me in first place, but there were no competitors like you two out there when I settled upon that practice. You guys have made it much more interesting (TOO interesting, in the case of Ryan's 3983 game!), and so I'm proud to post any score that's higher than what I've done before.

It may be that I'm simply overlooking some enhancement in the layout that yielded 3978, but I'm going to yank myself back to my disciplined approach and try a different variation, and then maybe a third if the next variation doesn't yield a higher game. Once I've experimented with each of those variations (and any others that come to mind along the way), I'll try to figure out which of them deserves a closer look.

Incidentally, that was a very odd game to play out. It took me (I'd guess) about 300 swaps to get the Y I needed for IDENTIFY. Later, when I was missing the E and Y necessary to make OXYPHENBUTAZONE, both letters showed up the first time I drew to get them. Other "waiting times" were more or less average, but it's nonetheless remarkable how much the "luck of the draw" can vary. It's hard to get used to the inescapable fact that, if you have, say, 2 chances in 79 of picking the letter you need and that letter doesn't show up 50 times in a row, your odds remain just as low, 2 in 79, as they were on the very first try.

That's why, of course, we lay out our games ahead of time -- and why I never look forward to actually playing one out.

Congrats TenzingHillary!

Since you were being rotated between second and third place, being tied, and now you're in a solid second, it was an advancement, in a way, at least in my opinion.

I've got a few new boards to work on as well, as I had previously ruled out some words that do seem to work. Everything looks like it's going to fall a few points short of 3983 though, but I'll keep crunching the numbers.

If you come up with something higher, I still want to eventually know a little more about your 3978 board, as I don't have a layout that scores exactly that.

Ryan,

I used REINJURED. I hope that helps.

Interesting commentary regarding highest possible scrabble game. I am actually working towards a similar challenge with an added twist. Scrabble for iPad/iPhone has recently added gamecenter, allowing people to keep track of their highest score and compare to other people. The game allows you to play yourself, however, the added challenge is that this game only allows 5 straight swaps, as your 6th automatically ends the game as a forfeit. This is super annoying. I've been following all the steps above, and the highest score I've obtained is 3736, putting me in 3rd place out of 300,000 or so. It's a much more complicated game because now you have to have "safety net" tiles stored up in advance, so that you can place a single letter somewhere and continue swapping till you get that 7 tile max word. It's probably raised my blood pressure several points (anyone?)

johnsbuffalo,

I don't have Scrabble for iPad/iPhone, and the maximum-pass rule would certainly prevent very high games. Why don't you try your hand at Facebook Scrabble?

Johnsbuffalo,

Definitely an interesting, more challenging task. I'm not sure how I feel about it though, as it makes luck more important (or our challenge makes it less important, however you want to frame it). I may jump into that at some point, after I'm done here. I would think you could make most of the same plays, but not get most of the 50 point bonuses, so it would involve building the words one to a few letters at a time, based upon what you had. It would be a neat challenge to see how to modify the plays to increase their likeliness of being drawn quickly, while reducing the points as little as possible. I'm sure there were plenty of "more likely to be drawn" plays that we've ruled out because they score one or more points less than our strategies.

And I second TenzingHillary's challenge to you to try it on Facebook - how much more could you score if you didn't have that restriction?