The effect of computers on top-level chess


The sixth game in the best-of-twelve World Chess Championship match between champion Magnus Carlsen and challenger Sergey Karjakin petered out in a draw yesterday, leaving the score tied at 3-3. According to the rules, the winner of a game gets one point while a draw nets each player half a point and the first player to reach 6.5 points wins the match. During each game many observers use a computer program called Stockfish to evaluate the moves of the players, and compare them compared with that of the best move by the computer and such programs seem to have become essential training tools for top players. The program app is available for free from the iTunes store and I have downloaded it and tried it out.

Oliver Roeder, who has been reporting on the contest, spoke to Murray Campbell, one of the developers of the Deep Blue chess program at IBM that defeated then champion Garry Kasparov in 1997 and sealed the dominance of computers, about the effect that computers have had on the game and whether computers are draining the beauty out of chess, especially with these two players who have grown up in the computer era.

“Grandmasters that have grown up with most of their training in the computer era play a much more objective style of chess,” Campbell told me. “They’re less willing to dismiss a move because it’s ugly, or doesn’t appeal to their aesthetics.”

Well sure, but does that mean chess’s “aesthetics” have been sacrificed in the process? Has the elegance of a great combination of moves been traded for the the slavish devotion to a computer driven “evaluation”?

“Chess is an art, but it’s more of a sport,” Campbell said. “If you’re interested in winning, then you play the right move, even if it’s an ‘ugly’ move or a ‘computer’ move.” (Air quotes his.)

So is the computer to blame for all these draws over the past week?

“Super-deep preparation can create a draw-ish tendency,” Campbell said. “The white player will try to create a position where the opponent has chances to go wrong. And the black player, if they’ve prepared well enough, will have found the way to navigate through that mess and find the way to the draw. I can certainly think of some 20- or 30-move games that have probably been entirely calculated at home.”

In other words, the computer, Campbell said, accelerates the calculation and preparation that can lead to draws. Thanks, computer.

The role of aesthetics in chess is similar to that in mathematics and may be getting influenced in similar ways. Elegance and beauty is valued in chess play and mathematics and in the past practitioners in each may have been tempted to emphasize it more. Nowadays, the use of computers in generating mathematical proofs and analyzing chess moves has led to the adoption of strategies based on how well they work rather than how pleasing they are. This likely leads to more cautious play. More daring moves in chess that are high risk/high reward are likely to be rejected.

Comments

  1. says

    In the 1980s and 1990s, critics said chess became a game of memorization rather than skill because so many games were played and moves recorded. It’s not surprising that computers have taken something else away from the game.

    Some chess writers have suggested ways of shaking up the game. Sitting down to play without knowing the layout beforehand changes the game entirely (both get the same arrangment). Games would be decided more by ability.

    https://en.wikipedia.org/wiki/List_of_chess_variants#Orthodox_chess_rules

    A player would still study an opponent’s tendencies, but an unknown layout means memorization and study of games is far less important, especially if it’s a layout the player has never played before.

  2. says

    I just had an interesting thought. What if someone didn’t know how to play chess at all, and chinese-roomed the game with mnemonics? While chess has a vast number of moves, very few of them are actually viable in any situation. I took a mnemonics class with Banachek back in ’07 and he had us memorizing decks of playing cards pretty easily by the end of the day. Could someone be a chess-playing idiot that was just a memory cache for a computer?

    It’s just more of the war on authenticity.

  3. John Morales says

    Marcus:

    What if someone didn’t know how to play chess at all, and chinese-roomed the game with mnemonics?

    Nah, doesn’t work that way. Too many combinations are possible for a look-up table approach.

    Things like positioning and territory and tempo are rather important, not to mention taking advantage of opponent’s errors. Strategy aside, the tactical approach is minimax for at least a few moves ahead.

    (I used to play competitively when I was in high school, too)

  4. says

    John Morales@#2:
    Too many combinations are possible for a look-up table approach

    OK, that makes sense. And, I suppose since its asymmetrical you can’t cut down the search-space using mirroring.
    I don’t know what “too many combinations” means but someone with mnemonics skills – especially since they’d have lots of time – could probably memorize hundreds of thousands of games.

    I just did a little googling and I can see that after 3 moves there are millions of possible games, though I don’t expect anyone skilled would play the vast majority of those games. But, yeah, you’d be into the billions pretty quickly even if you pruned the search tree pretty aggressively.

    I gather it’s now a big thing to try to detect chess players that are acting as sock puppets for a computer. I wonder if it’s going to get to the point of a colonoscopy before you go play in a faraday cage..

    Sometimes, the advent of digital innovation can free up an art-form, as it has done for photography. I’ll guess that as long as lots of money and nationalist pride are riding on chess, there will be cheating; it won’t be liberated as an art-form until the computers and the cheaters suck all the money out of it.

  5. consciousness razor says

    I don’t know what “too many combinations” means but someone with mnemonics skills – especially since they’d have lots of time – could probably memorize hundreds of thousands of games.

    Hundreds of thousands seems to be pushing it. But as you apparently found out, that’s hardly better than zero anyway, compared to the number of possible legal games.

    People do analyze endgames in a similar way, however. When there are a small number of pieces remaining on the board (in any configuration), there are endgame tablebases telling you exactly what to do to get the best possible result. These are relatively easy to study and understand (if not complete memorize), so you can play endgames very precisely. The game is perfectly and completely solved for those positions, and it would be solved at the opening if we ever got to tablebases for all 32 pieces, at which point we would have no reason to play chess anymore.

    We’ll almost certainly never be there, but the point is that you don’t start at the opening and work from there — what people are actually doing is (ever so slowly) backtracking from the end. Some openings would remain which (with perfect play) lead to a win, so in real games you’d only be memorizing some dozens of perfect moves (in one branch) to get to that. And if your opponent plays less-than-perfect moves, you can theoretically win even more easily (if you memorized those branches as well, or just know how to play well without rote memorization).

    I wonder if it’s going to get to the point of a colonoscopy before you go play in a faraday cage..

    I don’t know how online chess works, but in live tournaments where it really matters, cell phones and other such devices are banned. I’m not aware of any incidents of people cheating like that.

    You can use computers (along with trainers/coaches) as much as you want when you prepare before the tournament. Nobody does or should think of that as cheating. Something you often see is that a player (generally both players) had some nice preparation at the start, forgets it, gets confused about move order, or their opponent plays unusual or even suboptimal moves simply to take them out of their preparation. (If you can figure out how to punish your opponent for playing suboptimally, then they obviously weren’t helping themselves, but you may not figure that out.)

    Anyway, it’s fairly quickly back to being a matter of who can calculate the best at the table. Normal games with higher-level players don’t end when their preparation leads to a checkmate or resignation — that rarely ever happens. They make a blunder here or there (as everyone does), but they are still playing another human and can often get out of trouble if they are resourceful. Computer have shown that people were not (and never will be) perfect at calculating the best moves every turn. They’ve shown that there are ways of defending/attacking in certain positions which were rather dogmatically assumed to be bad, so they’re approached differently now, because people have become more familiar with some computeresque moves that may turn a losing position into a draw or a draw into a win.

    I think it’s good that they’re playing better chess. They are much better at defending, of course, which shouldn’t by itself strike you as a problem. Super-boring/pointless draws can generally be avoided with better tournament rules which incentivize more interesting strategies, not by refraining from playing whatever is (as a computer ought to agree) the best move in any particular position. You could for instance simply make it so that a single win is worth more than two draws — that’s something tournaments do, so players are less interested in the prospect of a draw (especially when playing black). There’s no reason why the results need to be counted as they traditionally are, with a draw equal to a half-point. Tournaments also play around with different time controls and various other features like that, without changing anything fundamental about the rules of the game itself.

Leave a Reply

Your email address will not be published. Required fields are marked *