Search found 566 matches

by mvk
Thu Jan 14, 2016 6:49 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: test positions for texel tuning
Replies: 32
Views: 7014

Re: test positions for texel tuning

It may also be possible to do standard gradient descent even if the function is not differentiable, by numerically estimating direction of gradients (using Newton's method to estimate partial derivatives). This may perform better than ad-hoc methods, and is very well studied. Gradient descent looks...
by mvk
Mon Dec 28, 2015 3:53 pm
Forum: Computer Chess Club: General Topics
Topic: Stockfish 7 beta 1
Replies: 29
Views: 13586

Re: Stockfish 7 beta 1

It's not of high priority, because position is highly artificial, but we'll fix it at some point... That point is now :-) I have open a pull request with the fix: https://github.com/official-stockfish/Stockfish/pull/546 That is certainly faster than expected. I quickly tested your branch and I can ...
by mvk
Mon Dec 28, 2015 11:49 am
Forum: Computer Chess Club: General Topics
Topic: Stockfish 7 beta 1
Replies: 29
Views: 13586

Re: Stockfish 7 beta 1

The "artificial" position is from a CCRL 40/4 game. (GrayMatter SVN1604 vs POS 1.20, date 2012-07-16). POS 1.2 is rated 230 elo weaker than RAM 2.0, which is a RANDOM MOVER ! So you're looking at a quality of chess worse than random moves! The fact that these patzers cannot even mate the opponent w...
by mvk
Mon Dec 28, 2015 10:06 am
Forum: Computer Chess Club: General Topics
Topic: Stockfish 7 beta 1
Replies: 29
Views: 13586

Re: Stockfish 7 beta 1

Issue #2 is just normal behaviour. Issue #1 is a real issue though. I've recorded it in our issue tracker: https://github.com/official-stockfish/Stockfish/issues/544 It's not of high priority, because position is highly artificial, but we'll fix it at some point... Thanks for reporting. Thanks for ...
by mvk
Mon Dec 28, 2015 8:22 am
Forum: Computer Chess Club: General Topics
Topic: Weird stockfish behaviour at low depths
Replies: 7
Views: 2654

Re: Weird stockfish behaviour at low depths

Even weirder, entering an infinite loop: position fen 7k/QQQQR3/2B5/4KN1Q/3QQ3/8/8/4R3 b - - 0 1 go depth 1 (To my surprise, this position comes from a CCRL game, not from Dann...) For this position the latest SF (051215) does not give infinite eval: position fen 7k/QQQQR3/2B5/4KN1Q/3QQ3/8/8/4R3 b ...
by mvk
Mon Dec 28, 2015 8:17 am
Forum: Computer Chess Club: General Topics
Topic: Stockfish 7 beta 1
Replies: 29
Views: 13586

Re: Stockfish 7 beta 1

Stockfish 7 beta 1 is out. We would kindly ask rating list authors and other interested parties to do test runs with this version and report any encountered issues in this thread. Thank you. There two bugs seem unsolved in version 7 beta 1. http://talkchess.com/forum/viewtopic.php?p=650311#650311 $...
by mvk
Thu Dec 10, 2015 8:19 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: Null Move in Quiescent search
Replies: 26
Views: 6099

Re: Null Move in Quiescent search

In PDS with 2 moments, variation/sigma always goes down with deeper search, even in the fully uncorrelated back propagation model, as long as the node is not singular (in which case sigma stays the same). At least, it won't pathologically go up when expanding the tree. For correlated back propagatio...
by mvk
Wed Dec 09, 2015 6:07 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: Null Move in Quiescent search
Replies: 26
Views: 6099

Re: Null Move in Quiescent search

It's interesting to read the first chapter of that thesis. It confirms Nau's observation that minimax search is often pathological. The deeper you search the worse you play, even with a good static evaluation. Why minimax search works so well in chess like games remains somewhat of a mystery (not s...
by mvk
Wed Dec 09, 2015 4:45 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: Null Move in Quiescent search
Replies: 26
Views: 6099

Re: Null Move in Quiescent search

It's interesting to read the first chapter of that thesis. It confirms Nau's observation that minimax search is often pathological. The deeper you search the worse you play, even with a good static evaluation. Why minimax search works so well in chess like games remains somewhat of a mystery (not s...
by mvk
Wed Dec 09, 2015 10:27 am
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: Null Move in Quiescent search
Replies: 26
Views: 6099

Re: Null Move in Quiescent search

It's interesting to read the first chapter of that thesis. It confirms Nau's observation that minimax search is often pathological. The deeper you search the worse you play, even with a good static evaluation. Why minimax search works so well in chess like games remains somewhat of a mystery (not s...