Author: Uri Blass
Date: 14:48:23 06/13/04
Go up one level in this thread
On June 13, 2004 at 17:00:38, Tord Romstad wrote: >On June 13, 2004 at 16:19:58, Uri Blass wrote: > >>I think to look at fruit's source code to learn about it's evaluation. >> >>Fruit nunn based rating increased from 2496 to 2642 based on the following link >>http://www.husvankempen.de/nunn/ratings/ratingsnunn2.htm >> >>I understand that it is only thanks to better evaluation when it even has no >>king safety evaluation. >> >>Did you look at Fruit1.5's source code to learn about it's evaluation. > >No, I haven't done that, and at the moment I don't have any plans to do so. >But from I have seen of Fruit's code, the most valuable things to learn are not >evaluation terms or search tricks, but rather robust, high-quality C/C++ >coding style. This is much more important (and much more difficult!) than >most people seem to believe. A stable and bug-free engine is one of the >most important keys to success. I cannot ignore the fact that based on the comments of fabian the only change from fruit1.0 to fruit1.5 was in the evaluation and it seems to give it a big improvement. I believe that at least for me it should be good to learn about fruit's pawn structure evaluation and mobility's evaluation. I believe that fruit has a relative advantage in this stuff relative to movei. Maybe for you things are different but I did not put much knowledge in the evaluation because I am afraid from bugs. Of course deciding to copy the evaluation is not a good idea because movei may evaluate the same thing indirectly in different stuff so I plan first to test the same knowledge with smaller scores. Considering the fact that Fabien did not implement king safety and did not start improving the search and considering the fact that results suggest that it is not more than 50 elo weaker than crafty I predict that in the next year fruit is going to become the best free program with open source for one processor. > >>I still did not look much at it but I guess that the fastest way for improvement >>should be to try to learn from it. > >I'm not sure, but at any rate I am not looking for the fastest way to >improvement. >One of the joys of being an amateur is that I can always concentrate on whatever >I >think is most fun at the moment, without considering whether it is the best way >to improve. > >Tord Things are different for me and before WCCC I want to get the biggest improvement that I can get. I am also not sure if implementing fruit's knowledge is the best way for me and if I succeed in doing it but I will probably try it. I delay rewriting the search code for after WCCC(I plan to try also some search changes but only changes that are simple for me to write without bugs when they are about adding few lines to the code). Uri
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.