Novel Chess
November 7, 2009 4:49 PM Subscribe
Reading to the Endgame: Algorithmic translation of classic nineteenth century novels into chessboard slugfests. Select the opponents from a list of fifty-five novels in five languages, and watch each text maneuver across the battlefield.
Another problem that occurs to me, and perhaps a project for the next NaNoWriMo:
Given a fixed pair of white and black move spirals, how can I optimise my novel to play the best possible game of chess? Because if I can't write a Heartbreaking Work of Staggering Genius, then my novel should at least be able to beat it at chess.
posted by kaibutsu at 5:09 PM on November 7, 2009 [3 favorites]
Given a fixed pair of white and black move spirals, how can I optimise my novel to play the best possible game of chess? Because if I can't write a Heartbreaking Work of Staggering Genius, then my novel should at least be able to beat it at chess.
posted by kaibutsu at 5:09 PM on November 7, 2009 [3 favorites]
what
posted by mhjb at 5:25 PM on November 7, 2009 [4 favorites]
posted by mhjb at 5:25 PM on November 7, 2009 [4 favorites]
Call of the Wild (W) vs. The Wonderful Wizard of Oz (B): Stalemate (W)
W. Oz (W) vs. C. Wild (B): Checkmate (B)
Oliver Twist (W) vs. Madame Bovary (B): Checkmate (B)
M. Bovary (W) vs. O. Twist (B): Draw
Then I got bored.
posted by Mister Moofoo at 5:31 PM on November 7, 2009
W. Oz (W) vs. C. Wild (B): Checkmate (B)
Oliver Twist (W) vs. Madame Bovary (B): Checkmate (B)
M. Bovary (W) vs. O. Twist (B): Draw
Then I got bored.
posted by Mister Moofoo at 5:31 PM on November 7, 2009
I don't know who this Ethan Frome guy is, but I watched him play with himself.
posted by cjorgensen at 5:38 PM on November 7, 2009
posted by cjorgensen at 5:38 PM on November 7, 2009
Books suck at chess.
posted by shadow vector at 6:14 PM on November 7, 2009 [5 favorites]
posted by shadow vector at 6:14 PM on November 7, 2009 [5 favorites]
Since they were using completely arbitrary strings of letters to determine the moves, and the moves had no strategy at all, the novels really had nothing to do with anything.
Those books are still classics today precisely because they are greater than the sum of their words, which is all the more apparent when you try to randomize them into a meaningless chess game.
posted by infinitefloatingbrains at 6:25 PM on November 7, 2009 [1 favorite]
Those books are still classics today precisely because they are greater than the sum of their words, which is all the more apparent when you try to randomize them into a meaningless chess game.
posted by infinitefloatingbrains at 6:25 PM on November 7, 2009 [1 favorite]
See, here I was hoping that they'd analyzed the strategies of the protagonists and antagonists of novels and converted them into chess strategies, pitting Protagonist A from book one against Antagonist B from book two, and then seeing who would have triumphed. That would have been really interesting, to me.
But, you know, strip-mining the books for random letter combinations is good, too...
/English major
posted by ilana at 7:19 PM on November 7, 2009 [3 favorites]
But, you know, strip-mining the books for random letter combinations is good, too...
/English major
posted by ilana at 7:19 PM on November 7, 2009 [3 favorites]
Given a fixed pair of white and black move spirals, how can I optimise my novel to play the best possible game of chess?
i'd go with this set of moves.
18-4
40-60
38-8
60-30
anybody have a way to count tuples in a text? If I could do that, I'd actually write something.
posted by empath at 7:26 PM on November 7, 2009
i'd go with this set of moves.
18-4
40-60
38-8
60-30
anybody have a way to count tuples in a text? If I could do that, I'd actually write something.
posted by empath at 7:26 PM on November 7, 2009
The Bible vs. The God Delusion. I'd pay money to see that one.
posted by pahool at 7:33 PM on November 7, 2009
posted by pahool at 7:33 PM on November 7, 2009
Using bigrams of letters, of course, means that your novel's playing style, insomuch as it has one at all, is mostly at the mercy of the novel's vocabulary—unless the letter patterns were arranged deliberately, which is usually not the case. Do not bring up the 'Bible Code' at this point. I may fling poo. Authorial style seems to have a lot to do with syntactic habits, though, so I'm curious what would happen if their tuples of choice were bigrams of syntactic tags, a la Hirst & Feiguina.
Possibly authors who use a comparatively narrower range of syntactic devices really would play different chess from those more versatile. And if you assigned a single spiral, based on the two texts' combined frequencies, then absolute differences in syntactic habits would lead to different idiosyncrasies in play (whereas by-player spirals mean it doesn't matter which syntactic patterns you favor, only how much you favor them).
Of course, it's an open question whether and how much the distance between two styles as measured by Hirst and Feiguina's method correlates with the distance as evaluated subjectively by readers; there's no guarantee that human psychology and the machine learning algorithm are gonna pick up on the same things. But the fact that their system does so well on authorship discrimination leads me to hope that it may be so.
empath, counting tuples is essentially the job of a Markov chain model. If you have some Perl-fu, there's useful code in this Markov generator by Kernighan and Pike—essentially the first half of the program.
posted by eritain at 11:36 PM on November 7, 2009
Possibly authors who use a comparatively narrower range of syntactic devices really would play different chess from those more versatile. And if you assigned a single spiral, based on the two texts' combined frequencies, then absolute differences in syntactic habits would lead to different idiosyncrasies in play (whereas by-player spirals mean it doesn't matter which syntactic patterns you favor, only how much you favor them).
Of course, it's an open question whether and how much the distance between two styles as measured by Hirst and Feiguina's method correlates with the distance as evaluated subjectively by readers; there's no guarantee that human psychology and the machine learning algorithm are gonna pick up on the same things. But the fact that their system does so well on authorship discrimination leads me to hope that it may be so.
empath, counting tuples is essentially the job of a Markov chain model. If you have some Perl-fu, there's useful code in this Markov generator by Kernighan and Pike—essentially the first half of the program.
posted by eritain at 11:36 PM on November 7, 2009
B\ah, this is massively stupid as near as I can see.
For one thing, the algorithm seems knows nothing about the game of chess except what moves are legal. For someone who "speaks" chess, this is like creating a sentence by picking words at random out of the dictionary.
For another, it's hard to tell but it appears as if the works of the author are boiled down into a number that's used to see a random number generator - and that the seed number is created by simply composing all the characters in the work in question.
What this means is that the moves have nothing whatsoever to do with the content of the work in question - that if one character in the text were changed, you'd get completely different play from this machine.
Perhaps I do not have the details exactly right - but what I am sure is that if there were real work put in there to extract high-level information from the text they would have told us about it.
This does not illuminate the game of chess, the authors' works, or the interaction between chess and literature. It is essentially a fake.
posted by lupus_yonderboy at 8:40 AM on November 8, 2009
For one thing, the algorithm seems knows nothing about the game of chess except what moves are legal. For someone who "speaks" chess, this is like creating a sentence by picking words at random out of the dictionary.
For another, it's hard to tell but it appears as if the works of the author are boiled down into a number that's used to see a random number generator - and that the seed number is created by simply composing all the characters in the work in question.
What this means is that the moves have nothing whatsoever to do with the content of the work in question - that if one character in the text were changed, you'd get completely different play from this machine.
Perhaps I do not have the details exactly right - but what I am sure is that if there were real work put in there to extract high-level information from the text they would have told us about it.
This does not illuminate the game of chess, the authors' works, or the interaction between chess and literature. It is essentially a fake.
posted by lupus_yonderboy at 8:40 AM on November 8, 2009
Hmm, the last page came up and of course I now have to recant what I wrote.
So they're searching the text for tuples matching a move. So that's actually quite neat. While it isn't entirely "chess-y", there is an actual correspondence between the text and the game - changing a letter wouldn't change the game much, textual edits would somewhat correspond to "game edits".
So I take it back... it is clever, and does say something about the text.
Note to self: patience before posting!
posted by lupus_yonderboy at 8:45 AM on November 8, 2009
So they're searching the text for tuples matching a move. So that's actually quite neat. While it isn't entirely "chess-y", there is an actual correspondence between the text and the game - changing a letter wouldn't change the game much, textual edits would somewhat correspond to "game edits".
So I take it back... it is clever, and does say something about the text.
Note to self: patience before posting!
posted by lupus_yonderboy at 8:45 AM on November 8, 2009
Well, at least Huckleberry Finn beat out The Call of the Wild. I never liked that book. (Well, okay, it was a stalemate, but Huckleberry Finn had a queen and a rook, whilst The Call of the Wild was left with just the king, so, y'know, in any sort of normal game, Huck woulda won.)
posted by The Great Big Mulp at 11:24 AM on November 8, 2009
posted by The Great Big Mulp at 11:24 AM on November 8, 2009
I really want to see how well this book plays...
posted by You Can't Tip a Buick at 4:52 PM on November 8, 2009
posted by You Can't Tip a Buick at 4:52 PM on November 8, 2009
« Older I found some evidence! | It's what it says on the tin. Newer »
This thread has been archived and is closed to new comments
I'm curious, though. They found that their scheme for novel-chess resulted in more checkmates than random chess. I'm curious what would happen if they randomised their 'spirals' for the white and black moves; are some spirals more likely to produce checkmates than others?
posted by kaibutsu at 5:07 PM on November 7, 2009