It probably is on a sliding scale, meaning that it adjusts its strengths based on how strong it perceives you.
chess computers
Sort:
adams2001 wrote:
Do chess computers learn?
No.
Well, at least not the ones you played against. I think I heard that Deep Blue's win against Kasparov was partly due to the IBM engineers feeding him with previous Kasparov games.
But I guess the ones on chess.com are just standard software set on "low calculation time" - thus the mistakes, and the possibility to have a lower level. They do not adapt their strength to you. There might be some effect if the server is more or less busy, but I wouldn't bet on it.
hodebisabode wrote:
you just suck that's all.
...said the guy with a rating under 1200 in all categories.
Besides, it is not very constructive.
Irontiger wrote:
adams2001 wrote:
Do chess computers learn?
No.
Well, at least not the ones you played against. I think I heard that Deep Blue's win against Kasparov was partly due to the IBM engineers feeding him with previous Kasparov games.
But I guess the ones on chess.com are just standard software set on "low calculation time" - thus the mistakes, and the possibility to have a lower level. They do not adapt their strength to you. There might be some effect if the server is more or less busy, but I wouldn't bet on it.
Oh he just means chess.com computers? Shredder for one definetly has the ability to use elo to rank you based on how well you play and then readjusts the computers strength to be slightly above or below your strength, based on how you want it to be.
the typical computer engine does not learn. its preprogrammed and will never change. Writing a computer engine goes like this:
1) implement brute force algorithm that just calculate EVERY single move for each player. When reaching the end in a variation add up the piece values to get a "value" of the position. Choose the move leading to the variation that ends up with the highest value for the computer. pretty easy huh?
2) Realize that the computer will be insanely slow and only calculate a few moves ahead. Too easy to beat. It sux.
3) Optimize the code (board representation, piece movement, move generation etc..). Optimize the search algorithm for finding moves. Make the evaluation function more clever taking the trade-off of more heavy computation needed into account.
4) you have made Houdini :D
....
Regarding "computer learning" we have backgammon as a more successful example. Where the computer can learn by playing vs. itself to traing an artificial netural network with the TD-lambda algorithm. It works well with backgammon and it sux with chess (to my knowledge anyway :)
Thats my few cents :)
I know this is a dumb question, but I'll ask it anyway: Do chess computers learn? I've noticed that the first time I play against a certain "book opening" at chess.com, at the 1200 or 1600 level say , i can win a lot of these games, but as i continue to play the same book opening the computer seems to get stronger each game. Is the computer "learning" from each game or if not, what exactly is happening?