Author: Colin Frayn
Date: 03:57:34 03/10/00
Go up one level in this thread
On March 09, 2000 at 22:11:22, Mark Longridge wrote: >It seems that my book has a few major holes, particularly with the Evans >Gambit and King's Gambit. What is the best way to add new lines to crafty's >book? I've found that a lot of good programs have problems when playing black in the King's gambit accepted. I also got an easy win with ColChess against SSEChess the other day playing the Evans gambit. I think few programs are ready primed to cope with playing black against such strategies, though most (especially Crafty, as I have found) seem more than capable of playing white. >All I did to make my book was medium with all of Kasparov's games and >RJF's games. But I'm thinking adding more human games may not be the best >for crafty. Perhaps a better "computer" book could be devised. Try studying as many games as possible of Paul Morphy. He seems extremely good at playing both these lines (and I presume that's because he knew how to defend them as well). >And I'm still not convinced that some sort of Monte Carlo analysis wouldn't >greatly benefit crafty's endgame. Especially if you used the endgames >typically reached in actual computer verus computer, and computer versus >strong human player. How do you intend to apply Monte Carlo techniques to endgame play? >Idle computer time could be used to generate an analysis of the targetted >end games. Then, when logging onto FICS or whatever, the new endgame >data will improve crafty's game long before the TB's come into effect. So you're suggesting a sort of 'endgame book' rather like endgame table bases? I think you'll find that the number of positions reached in typical endgames becomes prohibitively large. I know that openings are often very similar, or transpose into the same positions, but the same is not true for endgames. To get better endgame play, you really need to work on the static analysis function as there is rarely any chance of detecting variations deep enough otherwise. Cheers, Col
This page took 0.03 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.