Search found 211 matches

by asanjuan
Wed Nov 16, 2011 11:20 am
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: A newbie question about testing
Replies: 11
Views: 1607

Re: A newbie question about testing

Thanks Robert for your comments. I believe that your epd database is wide enough to give confident results.
This is the explanation i was looking for.
I was thinking of building a similiar db by myself, but your work is even better.
Thanks everybody.
by asanjuan
Wed Nov 16, 2011 9:19 am
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: Want to know what savings to expect with pruning methods...
Replies: 23
Views: 3778

Re: Numbers

Yes: starting position and 16Mb for TT, 2 killer moves. I think that's all. But just a suggestion: don't try to get my numbers, i think it's better to compare the gains in %, for example: using killers ->XX% less nodes, using null move: XX% and so on. This is because the number of nodes are very dep...
by asanjuan
Tue Nov 15, 2011 2:00 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: A newbie question about testing
Replies: 11
Views: 1607

Re: A newbie question about testing

Andrés, ¿am i suposed to delete repeated games after using a book? This is what you told me once, but is extactly what i want to avoid.
I'd like to know more opinions.
by asanjuan
Tue Nov 15, 2011 1:25 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: Want to know what savings to expect with pruning methods...
Replies: 23
Views: 3778

Re: Numbers

Here are my numbers for a 8 ply search with quiescence, using iterative deepening. (Not the best numbers, but i'm still trying to improve ;D): Plain alfa-beta without any intelligeng move ordering: 104.935.779 nodes PVS, adding killers and history heuristics for move ordering: 3.452.753 nodes adding...
by asanjuan
Mon Nov 14, 2011 4:04 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: A newbie question about testing
Replies: 11
Views: 1607

Re: A newbie question about testing

Thanks a lot.
My engine uses UCI, i'm going to download them now, both LittleBlitzer and the epd set.
by asanjuan
Mon Nov 14, 2011 2:53 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: A newbie question about testing
Replies: 11
Views: 1607

A newbie question about testing

I've read things about LOS, error margin for elo, and things like that in this forum.. and at the end all is reduced to play a great number of games between your engine vs older versions or vs other sparring engines. My question is: If i want to test my engine A against an older version B, and want ...
by asanjuan
Mon Nov 14, 2011 10:57 am
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: Want to know what savings to expect with pruning methods...
Replies: 23
Views: 3778

Re: Want to know what savings to expect with pruning methods

50 Mnodes for 8 ply sounds too much for me. Maybe you are not counting the nodes very well. Maybe your conter++ is in the wrong place (only a suggestion). I recently done some tests with my engine, and at the end, I had to rewrite the entire search, doing everything from the begining, starting with ...
by asanjuan
Mon Sep 12, 2011 3:23 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: question about SEE
Replies: 11
Views: 3346

Re: question about SEE

Sounds more like there is something wrong with your PVS implementation. well, is quite probable :wink: . Maybe i have to re-write my search functions and try without reductions to see if there is something wrong. The search function, and specially PVS, is simple in concept, but has some tricks. I c...
by asanjuan
Mon Sep 12, 2011 10:11 am
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: question about SEE
Replies: 11
Views: 3346

Re: question about SEE

I promised to keep you updated, so here it is: SEE was not the problem. My experiments showed that using SEE is better than MVV/LVA in quiescence, having the same results, but lesser nodes. I solved my problem with node explosion changing a simple "return beta". In my zero-window search i was return...
by asanjuan
Tue Sep 06, 2011 2:59 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: question about SEE
Replies: 11
Views: 3346

Re: question about SEE

Thanks for your comments.
I didn't realized that there were so many contributors, and that you can answer so fast!
I'll try with the condition material_gain >=0, implement a SEE routine and also revising move ordering.
This can take me some days, but I promise to keep you updated.
Thankyou again.