Search found 17373 matches

by hgm
Sat Jan 19, 2019 8:59 pm
Forum: Computer Chess Club: General Topics
Topic: Fairymax is using 25% of cpu
Replies: 6
Views: 262

Re: Fairymax is using 25% of cpu

Now is your complaint that 25% is too high or too low?

25% seems normal usage, for a single-threaded engine running on a two-core (= 4 hyperthreads) system. Fairy-Max shouldn't use any CPU when not on move; it would just be waiting for input.
by hgm
Fri Jan 18, 2019 4:36 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: Generate EGTB with graphics cards?
Replies: 29
Views: 1548

Re: Generate EGTB with graphics cards?

Perhaps KHEK is too small to be representative: with 25KB the entire EGT would easily fit in the level-2 cache, which is very fast. Could you also give me the building times for KRCKR? This has 59M positions. (Half of that when using symmetry. Does your generator make use of this symmetry?) This is ...
by hgm
Fri Jan 18, 2019 12:40 pm
Forum: Computer Chess Club: General Topics
Topic: Polyglot extension tool
Replies: 43
Views: 1615

Re: Polyglot extension tool

There is a tool that can convert bin books to pgn called Cerebellum book CONVERTOR You would have to be careful with that, as converting Polyglot books to PGN is not generally possible. There is no way to convert the position keys in the book back to positions. So you would be dependent on followin...
by hgm
Thu Jan 17, 2019 12:03 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: Generate EGTB with graphics cards?
Replies: 29
Views: 1548

Re: Generate EGTB with graphics cards?

Ah, you nearly fooled me! :wink: But KHEK only has 25K positions, with 2-fold symmetry (9*9*7*90/2), instead of 2M for KBNK with 8-fold symmetry (64*64*64*64/8). That is 80 times smaller. So it is not so fast at all: in stead of 2x faster than Marcel's "PrettyFast" generator it is 40 times slower. A...
by hgm
Thu Jan 17, 2019 7:22 am
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: Generate EGTB with graphics cards?
Replies: 29
Views: 1548

Re: Generate EGTB with graphics cards?

Just to make absolutely sure there is no miscommunication: with KBNK you mean the Chess end-game. Not Xiangqi KHEK, right?
by hgm
Wed Jan 16, 2019 8:10 pm
Forum: Computer Chess Club: General Topics
Topic: Can somebody explain what makes Leela as strong as Komodo?
Replies: 50
Views: 2150

Re: Can somebody explain what makes Leela as strong as Komodo?

I did that for pawn structure patterns in earlier versions of Schooner. However, lack of sufficient training data gave unreliable results. It seemed to help a bit - but not enough to make a 100 elo of difference. Mostly I set intuitive values for "patterns" from experience. Are not stored position ...
by hgm
Wed Jan 16, 2019 6:13 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: On-line engine blitz tourney January
Replies: 5
Views: 240

On-line engine blitz tourney January

The monthly on-line engine blitz tourney for January will take place on:

Saturday January 19, 3pm Boston time (21:00 Amsterdam time)

To connect:

winboard -zp -ics -icshost winboard.nl -icshelper timeseal -fcp ENGINE.exe -fd ENGINEFOLDER -autoKibitz

(for UCI engines, add -fUCI)
by hgm
Wed Jan 16, 2019 3:50 pm
Forum: Computer Chess Club: General Topics
Topic: Analyze game for dummies
Replies: 3
Views: 176

Re: Analyze game for dummies

It is a de-facto standard to report scores in this format from the mover's point of view. It seems easy enough to spot blunders by ignoring the sign altogether. E.g. where the absolute value jumps from 1.75 to 4.44 there was probably a blunder.. Of course the recommended way is not by staring at the...
by hgm
Wed Jan 16, 2019 7:48 am
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: Staged move generation and killers
Replies: 24
Views: 1824

Re: Staged move generation and killers

The difference is that he opens the window, so that he gets exact scores that he can sort on, rather than an upper bound that is meaningless for other purposes than singling out the best move.
by hgm
Wed Jan 16, 2019 7:45 am
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: MCTS: How to deal with extreme imbalances
Replies: 6
Views: 293

Re: MCTS: How to deal with extreme imbalances

Yeah, I already raised that point in a cute-chess issue discussion as well. The purpose of the heuristic evaluation is not only to reflect winning probability, but also to reflect progress. It is very bad to let it saturate too easily. Centi-Pawn scoring doesn't suffer from such saturation, and rema...