AI Guys (like myself) have been trying to get machines to play chess for decades. Personally, I think it’s a misplaced effort, in that it’s allied with the idea that rationale is the missing element in making machines more like natural intelligence. Bull-da-da, I say!
Computers generally play chess by doing massive look ahead to possible future game configurations. People don’t do this. Computers are also programmed to play chess, whereas we mere humans must learn the game. Chess is a very hard game to learn from experience, both for humans and machines.
Backgammon is very different. A machine has learned to play backgammon very well, simply from experience.
Why backgammon, and not chess? I believe it is because lots of game configurations can be accessed even by a novice player in the former, but not the later. In a sense, the randomness of the dice rolls (the noise) in backgammon makes it easier to learn!
To test this theory, I want to playtest what I call noisy chess. It’s like normal chess, but with the addition of a 6-sided die and a spinner:
Game play proceeds the usual way, but the player must roll the die after each move. If the result is a 0 (which happens half the time), the move stands. If the move is a 1 (which happens 1/3 of the time), the player spins. The player reads the spinner on the inner ring in this case (which has 8 slots, corresponding to the 8 squares around the place that the player has just moved to). The player’s piece is moved into the corresponding adjacent square (if it can be placed there, see rules below). If the die result is a 2 (which happens 1/6 of the time), the player spins, and reads from the outer ring of the spinner (which has 16 slots, corresponding to the 16 squares 2 spaces away from the square the player has just moved to). Once again, the player’s piece moves to the corresponding square from the spinner.
Additional rules are that a player’s piece can’t move onto a square where player already has a piece. A player also cannot move into check. If these cases occur due to a spin, the player spins again. One might also want to restrict bishops to appropriate diagonals, but I’m not sure about that.
I think noisy chess would open up lots of possibilities, including proximate attacks. But I theorize that it would make chess much more learnable, both by people and machines.
And I’m going to playtest it, soon as I get a roundtuit. If you do first, please let me know!
Contributed by Rob Smith.
Track with co.mments
