Author: martin fierz
Date: 17:14:19 08/30/02
Go up one level in this thread
>>I believe that more time and better hardware can help >>programs to avoid more traps and to expose holes >>in the book so I expect the difference to be smaller at >>120/40 time control with faster hardware. > >From my experiments, I expect the difference to be larger at longer time >controls. At very fast time control (e.g. G/1 or G/2 minutes) there is very >little repeatable difference. At G/30 and above, the difference becomes very >clear. > >I have no idea why it turns out this way, because I should expect that the book >would matter more when you cannot think about the right move very long, since >the book moves should be pretty good, at least. my experiments with book/non-book in checkers (usually 288 games per match) say the same: at long time controls, the difference is larger, if i look at the ratio of wins/losses. in checkers its hard to define what is better - would you rather lose a 288 game match with 40-20 or with 10-0? i'd say losing 10-0 is worse, although the difference in wins is smaller. with fast time controls, what happens is that the error frequency of both programs is high, so even if one has a book and regularly achieves comfortable positions, it will lose games because it makes errors. at long time controls, checkers programs play close to perfect, and then the program with a good book nearly never loses a game, and the win/loss ratio goes way up. aloha martin
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.