Search found 300 matches

by Pio
Sat Jul 24, 2021 10:22 am
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: Static exchange evaluation with promotion
Replies: 23
Views: 2724

Re: Static exchange evaluation with promotion

The "speculative" thing in wiki entry never made sense to me. Obviously you don't want to arbitrarily ignore the gains to be had from the last capture. I can't come up with a reason why you would? I'm honestly baffled that this is even a discussion. After the entire exchange you'd end up with black...
by Pio
Tue Jul 13, 2021 9:16 am
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: Just an untested idea
Replies: 10
Views: 3810

Re: Just an untested idea

A learning rule usually has the property that, after the parameters have been correctly learned, it should not change them, at least in expectation. I don't think what you are proposing has that feature. Can you give some intuition for why you think this should work? Because the elo it gave me in t...
by Pio
Sun Jul 11, 2021 8:20 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: Efficiently Generated Legal moves question
Replies: 45
Views: 8560

Re: Efficiently Generated Legal moves question

My conclusion on this is that the possibly main cause is the slowness of C# in . net framework 4. You are jumping to conclusions way too quickly. C# and .NET is not to blame for your engine performing at 1/10th the speed it should. My engine is written in C# using magic bitboards. It searches 3.7 m...
by Pio
Sat Jul 10, 2021 8:23 am
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: Efficiently Generated Legal moves question
Replies: 45
Views: 8560

Re: Efficiently Generated Legal moves question

My perft is only about 8,000,000 leaves per second On my hardware, my engine has roughly 20M NPS for perft in the initial position, and 1.8M NPS for regular search. The move generator logic is the same as yours, with make/unmake, and the "in check" detection is without bitboards. Instead, with dedi...
by Pio
Fri Jul 09, 2021 11:35 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: Efficiently Generated Legal moves question
Replies: 45
Views: 8560

Re: Efficiently Generated Legal moves question

My perft is only about 8,000,000 leaves per second On my hardware, my engine has roughly 20M NPS for perft in the initial position, and 1.8M NPS for regular search. The move generator logic is the same as yours, with make/unmake, and the "in check" detection is without bitboards. Instead, with dedi...
by Pio
Thu Jul 08, 2021 8:15 am
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: I generated a tree diagram of a minimax search versus an alpha beta search. Interesting to see how much gets pruned.
Replies: 13
Views: 3696

Re: I generated a tree diagram of a minimax search versus an alpha beta search. Interesting to see how much gets pruned.

So how do you explain that the best move in the root was searched first? You have 20 moves in the root, so without ordering that alone would be a 1-in-20 probability. And on most other moves (except for two, as Chris correctly pointed out) you pick a cut-move on the first try (also out of 20?). Thi...
by Pio
Tue Jul 06, 2021 5:11 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: Futility reductions
Replies: 11
Views: 2042

Re: Futility reductions

What would happen if you do the same as you do for ply 0 but for higher depths do reduced depth searches? What I had in mind was for those really bad positions with the same margins as you have do: Depth 0 do the same as you do Depth 1 return QS Depth 2 and 3 return depth 1 search plus QS Depth 4 a...
by Pio
Tue Jul 06, 2021 4:32 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: Futility reductions
Replies: 11
Views: 2042

Re: Futility reductions

It would be very interesting to see what results you would get if you changed your futility thresholds to: 200 283 346 400 447 490 529 I tried these, unfortunately they don't seem to work in Nalwald. Before, it won 56.6% against an earlier version, with these values only 54.5%. Though I am surprise...
by Pio
Tue Jul 06, 2021 11:28 am
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: Futility reductions
Replies: 11
Views: 2042

Re: Futility reductions

The idea is to generalize futility pruning to all nodes (not only depth == 1, ..., depth == constant). The amount of reduction is dependent on how much staticEvaluation + see(move) differs from alpha (see for quiet and tactical moves). In Nalwald I calculate the reductions like this ( value is alph...
by Pio
Tue Jul 06, 2021 7:53 am
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: PV-move ordering necessary if you have TT-move ordering?
Replies: 51
Views: 6584

Re: PV-move ordering necessary if you have TT-move ordering?

You must decide for yourself which approach conforms best to your measure of simplicity: 1) Putting extra code in Search() to follow the PV from the triangular array in the first branch of the tree, next to probing the TT and using the TT move. 2) Stuffing the PV from the triangular array back into...