Author: Sandro Necchi
Date: 10:37:06 10/11/05
Go up one level in this thread
On October 11, 2005 at 12:52:59, Dagh Nielsen wrote: >Very interesting, Sandro. > >I don't have Schredder, but it has inspired me to try and fiddle with the >parameters of Fruit 2.2 :-) Shall I than ask Fabien to compensate my finding work?:-) > >As it happens, some of the weaknesses I feel Fruit 2.2 default have are too >little penalty on bad pawn structures, not enough respect for passed pawns, and >not enough respect for king safety either. Also, sometimes it overestimates the >chances of bishop/knight + pawn vs. rook, but I find it difficult to adress this >issue without making Fruit on the same time overestimate Rook + pawn vs. two >minor pieces. I am trying now to make pawns worth less to solve this. And then I >make pawn structure, passed pawns and king safety more important exactly as in >your Gustav setting. Well, but to do this with setting is not so easy as the best would be to improve the program specifically. However considering positions in percentage this is possible with settings improvements... > >I am curious to find out how much improvement is possible. If, as I suspect, >Fruit's evaluation functions are generally just less sophisticated than >Schredder's, it MAY not help much to put more focus on pawn structure, for >instance. Well, the point is that changing more paramethers may they may interfer to each other, so a good balance is not so easy... > >I play on Playchess.com with my Fruit 2.2 engine, using my own book lines, so >that also has to be taking into account, results may not be universal. I would >be interested to find out how much book choices and engine settings depend on >each other, I mean, if it is possible to make a book and one engine setting that >is a killer, but the same engine setting would fail with other books. Yes, this is very true and the best is a book specifically suited to a given program/setting. > >In general, I find all this an interesting field of "research", it requires >creativity, good judgement and hard work, and results can be measured with the >cold ELO numbers :-) Yes, and some luck too! Sandro
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.