Author: Roberto Nerici
Date: 15:03:44 07/19/05
Go up one level in this thread
>Yes, go is a great game and a great answer. Even relatively weak humans still >trounce computers at go. Another answer might be backgammon. The top backgammon >programs are stronger than all but the very strongest humans, with whom they are >about on par; but due to backgammon's element of chance, brute force calculation >does not work. Backgammon programs use neural nets. To me, the most interesting difference between backgammon and chess, in terms of comparing human and computer play, is that in backgammon the programs are relatively good at strategic/pattern play, and relatively weak at tactics*. This is the inverse of the case in chess. This comes out of the fact that the best computer players are neural net based, as Robin pointed out. It might also be worth elaborating on why searching (brute force) doesn't work so well in backgammon. There are two reasons: 1. A typical backgammon position has around 20 possible moves (compared to 30+ in chess) but you also have 21 possible dice rolls, so a full ply has a branching factor or about 400. 2. Any of those 21 dice rolls can come up and therefore many of possible moves you could make would be good for at least one of the rolls. AB pruning doesn't handle this, so you really do have to look at all 400 possible outcomes. Regards Roberto/. * I'll make a caveat here that the neural net players are quite good at tacticaly play too in many positions; it is just they have holes that the best humans don't.
This page took 0 seconds to execute
Last modified: Thu, 15 Apr 21 08:11:13 -0700
Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.