Search found 211 matches

by asanjuan
Mon Jan 14, 2013 12:10 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: All and Cut nodes
Replies: 69
Views: 15585

Re: All and Cut nodes

I do the folowing: I assign the expected node type depending on the position in the tree, as is defined in the wiki (and other papers): - expected ALL nodes: LMR - expected CUT nodes: IID - both ALL and CUT nodes: null-move pruning, usually solves the classification when an ALL node is really a CUT ...
by asanjuan
Mon Jan 14, 2013 8:57 am
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: Need opponents inside the 2300-2400 elo range
Replies: 15
Views: 2564

Re: Need opponents inside the 2300-2400 elo range

I do not know which engines in that Elo range are stable at that time control. However, you could try using a shorter time control for those stable but too strong engines. Miguel Ballicora and I use an idea promoted by Don Dailey to test Gaviota. We use stronger engines with shorter time controls t...
by asanjuan
Mon Jan 14, 2013 8:46 am
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: Need opponents inside the 2300-2400 elo range
Replies: 15
Views: 2564

Re: Need opponents inside the 2300-2400 elo range

Stockfish, Crafty, and Arasan (among others) have a variable-strength feature. You can try adjusting the strength level until you get a reasonable strength opponent. I really don't trust in the strenght parameter. I prefer engines that are doing their maximum to have trustable results. I prefer to ...
by asanjuan
Mon Jan 14, 2013 8:34 am
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: Need opponents inside the 2300-2400 elo range
Replies: 15
Views: 2564

Re: Need opponents inside the 2300-2400 elo range

Maybe too strong for rhetoric, but i'll try.
I haven't understood what you mean when you say:
be compatible with both epd or pgn for starting positions.
Some engines doesn't support the position command (or the equivalent in winboard), so it's difficult to test with tem.
by asanjuan
Mon Jan 14, 2013 8:29 am
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: Need opponents inside the 2300-2400 elo range
Replies: 15
Views: 2564

Re: Need opponents inside the 2300-2400 elo range

Good to know that there are so many "dirties". I'll try them.
Thank you Andrés.
PD: In fact I'm searching engines as stable as eveann. ;)
by asanjuan
Fri Jan 11, 2013 9:55 am
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: Need opponents inside the 2300-2400 elo range
Replies: 15
Views: 2564

Need opponents inside the 2300-2400 elo range

Hello, I need some engines to test Rhetoric. They must meet some conditions: - be stable at super fast time control (40/5+0.05) using cutechess-cli. Engines producing time losses will be discarded. - CCRL elo >= 2300 - CCRL elo <= 2400 - be compatible with both epd or pgn for starting positions. I o...
by asanjuan
Fri Oct 05, 2012 8:13 am
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: LOS calculation: Does the same result is always the same?
Replies: 30
Views: 5153

Re: LOS calculation: Does the same result is always the same

LOS is only based on statistical error. The only way this could be incorrect is that something causes systematic error. For example some other process is making heavy use of CPU from time to time. This kind of random systematic error of course favours the weaker engine. Unfortunately there is no st...
by asanjuan
Fri Aug 10, 2012 8:21 am
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: How effective is move ordering from TT?
Replies: 217
Views: 40486

Re: How effective is move ordering from TT?

the killer move heuristic isn't very useful, it only gives a small improvement over a proper implementation of History ordering. History and Killer are a bit redundant in a way, so if you have to choose one or the other it should be History (much more general and scalable) really??? Then your tests...