Author: Shaun Brewer
Date: 03:31:30 08/28/02
Opening book test results: generated books! All tests using crafty on two Athlon Tbird 850mz's at 15m + 15s 2600Plus is a pgn database where both players are rated above 2600 Database is the Lunsen database C = 2600Plus.pgn 50 1 50 (winner) D = 2600Plus.pgn 50 2 50 CraftyC : 209 (+ 43,=139,- 27), 53.8 % G = 2600Plus.pgn 50 1 25 B = 2600Plus.pgn 50 2 50 (winner) CraftyB : 200 (+101,= 77,- 22), 69.8 % B = Database.pgn 50 1 50 G = Database.pgn 50 2 50 (winner) CraftyG : 168 (+ 29,=129,- 10), 55.7 % E = 2600Plus.pgn 50 1 50 F = Database.pgn 50 2 50 (winner) CraftyF : 222 (+ 75,=136,- 11), 64.4 % Accidental repeat test stopped at 61 games B = 2600Plus.pgn 50 2 50 G = 2600Plus.pgn 50 1 50 (winner) CraftyG : 61 (+ 17,= 41,- 3), 61.5 % PGN's available if anyone is interested. I am currently running tests with a modified Database.pgn I took all the wins + losses (ie draws removed) All 2400+ All 2500+ All 2600+ and concatenated the pgns I have generated a book based on this using 50 3 50 the idea being this book will contain all the 2600 lines all the wins and loss lines for 2500 games etc. e.g. an attempt to bias the book towards the lines played by higher ranked players and reduce the number of lines which draw. This is being matched against Database.pgn 50 2 50 Current score 61 (+ 7,= 47,- 7), 50.0 % - too few games to draw any conclusions yet. I also plan to experiment with adding computer v computer games SSDF etc. I will also be testing against existing crafty books. My aim is to prove a method of generating a reasonable (good hopefully) book that anyone can copy. Shaun
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.