Search found 90 matches

by Tony P.
Fri Jan 24, 2020 6:57 am
Forum: Computer Chess Club: General Topics
Topic: Another tactical test position
Replies: 1
Views: 53

Re: Another tactical test position

Actually, the whole game was a KID mess :lol: [Event "Computer chess game"] [White "Stockfish_9_x64"] [Black "Stockfish_20011801_x64"] [Result "0-1"] [ECO "E99"] [Opening "King's Indian"] [Variation "Mar del Plata, Benko Attack, 11...Nf6 12.Nd3"] [TimeControl "180+2"] [Termination "normal"] [PlyCoun...
by Tony P.
Fri Jan 24, 2020 6:34 am
Forum: Computer Chess Club: General Topics
Topic: Another tactical test position
Replies: 1
Views: 53

Another tactical test position

7r/p4q1k/1r1P1bpn/2NQp2n/2P1Pp2/1RB2P1b/P3B2P/6RK w - - 12 33 The above position occurred in a casual SF9 vs SF11 blitz on my old 2-core laptop. SF11 thinks White holds, e.g. with 33. Rd1: Stockfish_20011801_x64: 1/2 00:00 254 85k -0.87 33.Nd3 2/2 00:00 493 123k -0.81 33.Nd3 Qxd5 3/3 00:00 693 173k...
by Tony P.
Sun Jan 19, 2020 10:00 pm
Forum: Computer Chess Club: General Topics
Topic: a crying shame (re: self-learning engines)
Replies: 20
Views: 1011

Re: a crying shame (re: self-learning engines)

Strictly speaking, the so-called active learning would only be possible during the training or private testing, not during official competitions, because it supposes that the learner can choose which positions the teacher should label, which can't be forced on the opponent during a competitive game....
by Tony P.
Thu Jan 16, 2020 9:35 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: Board adaptive / tuning evaluation function - no NN/AI
Replies: 18
Views: 824

Re: Board adaptive / tuning evaluation function - no NN/AI

I might end up trying some method from this survey of knowledge graph embeddings/projections (treating positions as KGs) or whichever survey will be the most up-to-date then, but I expect a lot of trial and error :( as the trade-off between eval accuracy and speed seems dramatic in chess.
by Tony P.
Thu Jan 16, 2020 8:07 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: Board adaptive / tuning evaluation function - no NN/AI
Replies: 18
Views: 824

Re: Board adaptive / tuning evaluation function - no NN/AI

They represent the entire game, not just the current state. Mu0 was tested vs A0 with a search budget of 800 (lol) nodes per decision. I doubt Mu0 would be competent at large search depths without modifications, as it only encodes the root and then searches in the latent space. However, if it were ...
by Tony P.
Wed Jan 15, 2020 5:40 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: Board adaptive / tuning evaluation function - no NN/AI
Replies: 18
Views: 824

Re: Board adaptive / tuning evaluation function - no NN/AI

OK, there's no need to read that dissertation - there's so much literature available that I can't be sure which method is optimal, need to use RL techniques to learn to explore the space of RL papers more efficiently :lol: Here's a possibly more interesting paper: Measuring Structural Similarities i...
by Tony P.
Wed Jan 15, 2020 4:28 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: Board adaptive / tuning evaluation function - no NN/AI
Replies: 18
Views: 824

Re: Board adaptive / tuning evaluation function - no NN/AI

Has anyone used a NN for hashing or board state representation/encoding ? It has been done in MuZero *. I like the approach a lot and think Mu0 would grow a lot stronger than Alpha0 if Deepmind hadn't switched it from chess to tasks that interested the team more. Another piece on this topic (albeit...
by Tony P.
Wed Jan 15, 2020 8:29 am
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: Board adaptive / tuning evaluation function - no NN/AI
Replies: 18
Views: 824

Re: Board adaptive / tuning evaluation function - no NN/AI

I'm not sure what you mean by 'AI'. Nowadays, it's standard for strong engines to use at least a bit of machine learning, e.g. Texel tuning . Out of the current engines, Winter is the one that I find the soundest scientifically. It has started using a small NN for eval recently but used to rely on l...
by Tony P.
Tue Jan 14, 2020 8:30 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: Leela avoids taking a pawn early on
Replies: 2
Views: 791

Re: Leela avoids taking a pawn early on

6. Nf3! starts a strong attack that exploits Black's development lag and seems to win at least a piece in all the variations. They're indeed hard for a human to calculate, but modern engines find the move instantly (I've used a 1-year-old SF version on a slow laptop). 1. d4 d5 2. c4 c6 3. Nc3 Bf5?! ...
by Tony P.
Fri Sep 22, 2017 3:32 pm
Forum: Computer Chess Club: General Topics
Topic: Another possible toy for AI programmers: 4-player chess
Replies: 9
Views: 1206

Re: Another possible toy for AI programmers: 4-player chess

This version has substantial bonuses for piece captures, so 'leaning back' is still less lucrative than attacking - in the process of 'exhausting each other', armies will earn points for captures and double / triple checks that might turn out vital in the final tally. For future reference, here are ...