Author: Tony Hedlund
Date: 08:27:01 10/20/97
Go up one level in this thread
On October 19, 1997 at 23:40:08, Keith Ian Price wrote: >On October 18, 1997 at 08:13:03, Tony Hedlund wrote: > > >> 31% >>E1 (164)E12 (89) 29% total 46-54 >> 39% >>Queenindian seem to be the name of the game for black.. >> > >I checked 13700 QI games between human masters, and while white's >winning percentage is close, the draws and black wins are much >different. Your sample size is small, but I wonder what makes it >different for computer-computer games. It would be good if everyone who >tests two programs against each other at 40/2 could submit the results >to a central area where everyone could download the results. I've taken the games from such an area. Our database at: http://userweb.interact.se/~tonyh/welcome.htm. >With a larger sample to work with, some trends could be identified that might >help the chess program authors to handle their opening books better. I >don't mean killer books, but just lines that seem to do well against all >computers as black or white. > >Data from 13700 human games: > >White wins 34.66% >Draws 43.00% >Black wins 22.34% > >This fits in with the general perception that the Queen's Indian is >drawish. > >The main line variation was even more so (5154 games): > >White 31.67% >Drawn 46.71% >Black 21.62% > >I think that SSDF should perhaps require submission of PGNs from all >testers to validate the data, and to create data as described above. It's the general idea. But most of the testers don't have the skill the time or the interest do to that. >Thanks for your interesting post, > >kp Thank you for the interesting respond! Tony
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.