Author: pavel
Date: 09:30:15 04/14/05
Go up one level in this thread
On April 14, 2005 at 12:20:53, pavel wrote: >On April 14, 2005 at 12:02:07, Steve Maughan wrote: > >>Pavel, >> >>>In discussion with Dann. I wanted to try out to see how many games it would >>>take to achieve 1 ELO error margin. While I do agree with Dann that it would >>>take 100,000s games to achieve this. I would like to see this in practice. >> >>You're thirst for accuracy is laudable but I fear pointless. Why stop at +/-1 >>ELO at 95% certainty? Why not +/-1 ELO at 99.99% certainty. IMO people can >>become obsessed with the error bars and forget the excepted outcome. >> >>Just my 2 cents, >> >>Steve > >Actually I am doing it just for fun. Reaching the targetted error margin is one >of the goals, but playing few hundred thousands games between two engines, >however pointless, is something big in itself. > >Another thing I would like to see is that how important "position learning" can >be. I have always been facinated by chess programs learning from their games. >AFAIK Yace has one of the best positional learning function among free engine. >While Aristarch only has book learning. > >Clearly Aristarch is the stronger engine among this two (as the results show in >many other tournaments), I would like to see if YACE positional learning will >help it narrow the gap in the long run. The thought may be naive, considering >that it probably won't reach the same position more than often and actually win >those games in those positions because of learning, but I think it's worth a >shot. > >This post is reply of all three posts (Steve, Kolss and Uri). > >Cheers, >pavs One thing about YACE learning feature that I dislike is that it doesn't overwrite the same position when importing it. For instance my YACE learnpos.bin has "28169" positions. But all of them are not unique position. If I try to import 30 positions from learnpos.yac 5 times it will add 150 new positions, so it will add the same positions 5 times. I don't know if it does the same thing during a game. If it learns from a position it reached and learned before, does it overwrite it with a new score or update it with a new score? pavs
This page took 0.01 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.