Scrabble Tactics

[David Williams]

There's been a bit lately about the tactical element of Scrabble differentiating it from Countdown. The following is in today's Times.

At the start of a game you have AEIDRTW. What is your best play?

[Kai Laddiman]

At the start of a game you have AEIDRTW. What is your best play?

Play dead.

[Charlie Reams]

Would be very tempting to change the W. ADEIRT gives bonuses from lots of possible draws, and by going second you give yourself more tiles to play through, boasting your chance of an 8. But you'd have to be quite confident to resist WAITED for 28.

[Daniel O'Dowd]

Would be very tempting to change the W. ADEIRT gives bonuses from lots of possible draws, and by going second you give yourself more tiles to play through, boasting your chance of an 8. But you'd have to be quite confident to resist WAITED for 28.

Fishing like that may always go wrong, and it gives your opponent a strong suspicion of what you are doing, offering a chance to change his own letters, or close down the board with some unhookable 3/4 letter word or a word with awkward consonants, after which if you have a bingo, you may not be able to place it, and a seven to make eight isn't so likely either. It's a nice idea, but it isn't quite aggressive enough. Similarly, WAITED for 28 is nice, but you're reducing your rack to a crapshoot.

from ADEIRTW, I'd probably play something like WARD for 16, keeping EIT.

[Craig Beevers]

Would be very tempting to change the W. ADEIRT gives bonuses from lots of possible draws, and by going second you give yourself more tiles to play through, boasting your chance of an 8. But you'd have to be quite confident to resist WAITED for 28.

Fishing like that may always go wrong, and it gives your opponent a strong suspicion of what you are doing, offering a chance to change his own letters, or close down the board with some unhookable 3/4 letter word or a word with awkward consonants, after which if you have a bingo, you may not be able to place it, and a seven to make eight isn't so likely either. It's a nice idea, but it isn't quite aggressive enough. Similarly, WAITED for 28 is nice, but you're reducing your rack to a crapshoot.

from ADEIRTW, I'd probably play something like WARD for 16, keeping EIT.

Keeping EIT isn't much better than keeping D. Vowel heaviness is the main reason. You want to generally keep 1 more consonant than vowel. ERT would be a far stronger than EIT, EIRT would be stronger still because you're closer to a bonus and reduce the random element which is enough to compensate for the slightly vowel heavy leave. Don't overvalue the 3-4 letter leaves too much though. Unless they involve an S or blank they're generally just worth a handful of points. For me there aren't enough 7s possible to warrant sacrificing 28 points to change one letter. The likely reply is a change or some crappy floaters.

Quackle superleave figures if anyone really wants them...

D 0.43
EIT 1.87
ERT 8.11
EIRT 9.49
ADEIRT 22.13

Of course these are just averaged out.

[JimBentley]

Play WATER for 24, leaving ID? I've no idea, my tactics in Scrabble are awful. My real answer, however, is "whatever Craig's about to say".

Edit: Oh bollocks, not fast enough.

[Daniel O'Dowd]

Keeping EIT isn't much better than keeping D. Vowel heaviness is the main reason. You want to generally keep 1 more consonant than vowel. ERT would be a far stronger than EIT, EIRT would be stronger still because you're closer to a bonus and reduce the random element which is enough to compensate for the slightly vowel heavy leave. Don't overvalue the 3-4 letter leaves too much though. Unless they involve an S or blank they're generally just worth a handful of points. For me there aren't enough 7s possible to warrant sacrificing 28 points to change one letter. The likely reply is a change or some crappy floaters.

Quackle superleave figures if anyone really wants them...

D 0.43
EIT 1.87
ERT 8.11
EIRT 9.49
ADEIRT 22.13

Of course these are just averaged out.

Hmm, I did think about WAD leaving EIRT, but I was slightly fearful that this would be easier to close down, which is an odd thing to say because I always played very openly anyways (I stopped playing a good year and a half ago with a rating about 1050 and a game average of 400.) It does have a back hook though of course, I just wish I'd not gone to chess first, or I'd have still been Scrabbling and getting higher up, but I've gotten so good with the chess that I haven't enough time for both in one lifetime.

[Jon O'Neill]

(I stopped playing a good year and a half ago with a rating about 1050 and a game average of 400.)

Impressive, Craig is only 191!

[David Williams]

Can't find it online, but The Times agrees with Charlie. Change the W.

You have a 57% chance of picking a letter that gives you a seven letter word, and, if your opponent doesn't swap as well, there are 81 different ways of making an eight letter word. Mind you, I've never come across Alan Simmons who writes the article, so if Craig disagrees, I'm with Craig.

[Ben Wilson]

Can't find it online, but The Times agrees with Charlie. Change the W.

You have a 57% chance of picking a letter that gives you a seven letter word, and, if your opponent doesn't swap as well, there are 81 different ways of making an eight letter word. Mind you, I've never come across Alan Simmons who writes the article, so if Craig disagrees, I'm with Craig.

Ironically, Allan Simmons beat Craig in the final of last year's National Scrabble Championship.

[David Williams]

Whoops. I'm still with Craig, though.

[Martin Smith]

My instinctive move would be DAW rather than WAD, because it leaves a slightly lower chance of your opponent being able to play a 7 (or 8) letter word. Keeping the r gives you -er, -ier and re-. WADI might be an option, if it's a valid play. My concern about changing one tile is that your opponent will realise that you have a very good rack, and will instinctively play a blocker.

[Gavin Chipper]

Quackle superleave figures if anyone really wants them...

D 0.43
EIT 1.87
ERT 8.11
EIRT 9.49
ADEIRT 22.13

Of course these are just averaged out.

What does all that mean?

Also from what I understood when Martin Gardner was posting some stuff once, Quackle is a sort of Apterous Rex for Scrabble. So what would Quackle do?

[Charlie Reams]

Quackle superleave figures if anyone really wants them...

D 0.43
EIT 1.87
ERT 8.11
EIRT 9.49
ADEIRT 22.13

Of course these are just averaged out.

What does all that mean?

Also from what I understood when Martin Gardner was posting some stuff once, Quackle is a sort of Apterous Rex for Scrabble. So what would Quackle do?

Super-leaves are Quackle's estimate of what would how many points each set of tiles is likely to be worth if retained for a future turn. Obviously those values above should be offset by the points you'd get for playing the remaining ones. I did a little sim and Quackle seemed to like WAIRED for 28 and several other moves better than the suggested exchange, and as highlighted above it will tend to overvalue the exchange anyway because it can't account for the suspicious signal you're sending to your opponent, so it looks like taking the points is probably the best option after all.

[Charlie Reams]

Okay, I ran it properly this time (50k iterations at 3 ply) and it liked WAIDE best, and Exchange W second (about 0.4 points difference). Most other plays that score 24 or more are almost as good.

[Martin Smith]

That wouldn't've been my choice even if I knew it, but I can see the logic in keeping RT. Obviously it's back to the drawing board for me.

[Gavin Chipper]

How complicated are Scrabble tactics generally? I mean if you, instead of trying stuff out and seeing if you win games over a long period, just see what Quackle does in various situations, would that be a fast-track method of being a good tactician or would it still take ages?

[Craig Beevers]

You would get a rough feel pretty quickly just by playing and seeing the evaluations from generate choices.