Move strength algorithm
-
I did think that the Move strength algorithm might be extremely complicated to program. But now I think it simply looks at all possible words in all possible places (just an easy enough 'brute force' approach for a computer).
Now here's the part I missed before. Instead of examining all the possible positions (difficult to program?), taking into account doubles/triples on words/letter, I think it could just calculate the optimal maximum score for each word -without taking into account whetherr or not theses maximum score words would actually fit on the particular board.
Now you have a list of words going from zero to some maximum score. This maximum score might be impossible to obtain on the board that the player is using. So what?! You can easily calculate the percentage strengths of these theoretical word fits which would be a good guide for players to gauge how good their words are.
I've come up with this idea because, when playing and checking 'Move Strength' it often feels incorrect. What do you think?
-
@mistertoad said in Move strength algorithm:
I did think that the Move strength algorithm might be extremely complicated to program. But now I think it simply looks at all possible words in all possible places (just an easy enough 'brute force' approach for a computer).
Agree that it would be very complicated to be perfect.
It sometimes indicates that there is a very high-scoring play available, such as a bingo, which just doesn't seem attainable to me.
on the other hand, sometimes it misses a fairly obvious one. in this example, ZA shows as 100% when in fact ZAG would be an improvement. -
Wouldn't it be lovely if one of the developers could pop in and explain a bit more about it? Or even the admins. There have been so many discussions about it, yet complete silence from the Lex team. You'd think they'd be eager to explain the new features.
-
@b-birney said in Move strength algorithm:
@mistertoad said in Move strength algorithm:
I did think that the Move strength algorithm might be extremely complicated to program. But now I think it simply looks at all possible words in all possible places (just an easy enough 'brute force' approach for a computer).
Agree that it would be very complicated to be perfect.
It sometimes indicates that there is a very high-scoring play available, such as a bingo, which just doesn't seem attainable to me.
on the other hand, sometimes it misses a fairly obvious one.Keep in mind that the algorithm knows ALL the possible words; unless you are an expert it knows many more bingos than you do, and it is also trivial for it to find bingos which go through multiple existing (on the board) words, which I for one have trouble seeing.
As to the developers not commenting, I can see why they wouldn't want to-- it's a work in progress and hopefully they have more important work to do than to talk here.
-
It seems to me that this game isn't good at randomizing letter distribution. Bingo words are very often in the first rack, which leads me to think that the game relies on throwing out whole words, not truly random tiles. Very often, certain letters come in clumps, such as UUU (yesterday), or EEEE (too many times to count).
Many players make 2 or 3 bingos in a game, which is just plain not random.
What say you, Lex?
-
@lolamoth said in Move strength algorithm:
Many players make 2 or 3 bingos in a game, which is just plain not random.
Even with perfectly randomized mixing, I wouldn't call that unusual when there are 8 tiles in a rack.
@lolamoth said in Move strength algorithm:
Very often, certain letters come in clumps, such as UUU (yesterday), or EEEE (too many times to count).
This is true. I often find myself fighting with groups of duplicate vowels on V2.
-
@lolamoth said in Move strength algorithm:
It seems to me that this game isn't good at randomizing letter distribution. Bingo words are very often in the first rack, which leads me to think that the game relies on throwing out whole words, not truly random tiles. Very often, certain letters come in clumps, such as UUU (yesterday), or EEEE (too many times to count).
Many players make 2 or 3 bingos in a game, which is just plain not random.
What say you, Lex?
Well one thing I have noticed is that I almost never get bingos in the correct order on my rack. I wonder if they check for bingos and then scramble if there is one!
As for getting multiple bingos per game: I don't think it has anything to do with randomness. There certainly are a lot of cheaters on here but some of us don't cheat and are just strong players!
-
Interesting.
What if I said to you that we are much more likely to get a bingo with the first rack because we are not limited by which hooks are available?
A second thought would be to conjecture that human beings, if asked to create random strings of letters, would intuitively reject the repetition involved in strings such as 'UUU' and 'EEEE'. Machines do not have our discernment in such matters. Having said that, the current pseudo random generators are, well, not truly random at all.
No reason why you or I cannot make a few bingos in a single game, if we are lucky with the draws.
-
@mistertoad
yes, I fairly often find a bingo on the rack but there is nowhere to hook it, but rarely can create one using a hook when there is no bingo on the rack . -
@b-birney said in Move strength algorithm:
@mistertoad
yes, I fairly often find a bingo on the rack but there is nowhere to hook it, but rarely can create one using a hook when there is no bingo on the rack ."White man speaks with forked tongue"
-
I have been playing scrabble for decades—at one stretch, daily—and I can say with certainty that this game's letter distribution is not truly randomized.
My issue isn't about Lex including words I don't know. Of course that's true.
-
I disagree with your premise. The odds against a 7- or 8-letter dictionary word in the first rack are high.
-
I think ZA is a better move here than ZAG, because it limits the opponent's options to just two: zaS or zaG. I might very well go with ZA, holding back the G for my next play, which could be used for gASH or zaG.
-
@lolamoth said in Move strength algorithm:
I have been playing scrabble for decades—at one stretch, daily—and I can say with certainty that this game's letter distribution is not truly randomized.
My issue isn't about Lex including words I don't know. Of course that's true.
"Many players make 2 or 3 bingos in a game, which is just plain not random."
Excuse me, but that makes no sense whatsoever. I have bingoed 3 times in a game playing Scrabble in real life. Doesn't get more random that that!
I don't think anybody is saying Lex is truly random-- as far as I know NO purely code based random generators are truly random.
-
@lolamoth said in Move strength algorithm:
I disagree with your premise. The odds against a 7- or 8-letter dictionary word in the first rack are high.
Thanks for your reply and the interesting link.
You are saying that I claimed the probability of being able to get a 7 or 8 letter dictionary word with first rack are "high"? Firstly, I was not talking about the probability of getting a bingo on the first turn explicitly and secondly I made no claim, only a conjecture, thus:
"What if I said to you that we are much more likely to get a bingo with the first rack because we are not limited by which hooks are available?"
As you can see, I was comparing the probability of getting a bingo on the first turn with the probability of a bingo later on (when hooks would become available.
-
@lolamoth I agree that ZA is a better play; in fact, that is the actual play I made at that point in that actual game.
But I don't think that the Move Strength Indicator algorithm is sophisticated enough to take that into account ... kind of what @mistertoad said in this thread. -
Post 7 of 16