TalkChess.com

// associate best score with best move if (alpha != old_alpha) best_move = best_so_far; Is best_move a global? If yes, then shouldn't we only be setting this at the root of the search? Otherwise this might get set on moves at nearly the leaf of the tree, and should often be suggesting illegal moves.

- Tried to write it functional. So recursion instead of loops. Recursion instead of iteration is pretty bad in languages without tail-call recursion. And, honestly, tail-call recursion is a hack to allow functional languages to have better performance than they should. - yes for debugging nested ex...

Instead of calling .Aggregate(), you can call .Sum() to simplify. Even though the performance for perft isn't the biggest concern, there are 2 issues with using LinQ IMO: It creates additional objects, which need to be GCd. You lose some control over the sequence in which the code is executed. It ca...

It doesn't look like FirstChess will evaluate a draw for repetition or 50 move rule, so it will actually blindly stumble into those without realizing they terminate the game. The only draw FirstChess can see is stalemate, which I see on line 623. For contempt, there are 2 things to take into account...

Which is your setting about resign and draw? How much CPU can be saved with it? I think it can be somewhat dangerous to use automatic adjudication in engine testing tournaments. This is for 2 reasons: there are some positions that engines typically see as a win but are actually drawn, and otherwise...

So if I understand the matter correctly, here's a conclusion: If every PV node and cut node has perfect ordering (i.e. the PV move or a cut move is evaluated first), then you will see that every expected cut node will be a cut node, and every expected all node will be an all node. However, if a cut ...

You should first convert your scores to win probabilities first so the big scores won’t dominate. It is the mean squares of the win probabilities you want to minimise. Yeah, texel tuning does this via a sigmoid function. It isn't only to get the probabilities but the absolute values will have a dif...

Pio wrote: ↑

Fri Jul 10, 2020 1:43 pm

You should first convert your scores to win probabilities first so the big scores won’t dominate. It is the mean squares of the win probabilities you want to minimise.

Yeah, texel tuning does this via a sigmoid function.

The explanation of average scores of many similar positions does partially answer my question, but there's another aspect that I don't think I expressed well. When tuning something (an eval in chess, or anything else), your error function will determine what things are most important to your tuning ...

I have a theoretical question on Texel's tuning method. If I understand it correctly, the function takes a set of positions, and minimizes the mean squared error of the positions. The error in a position is the difference between the (adjusted) evaluation and the result of the game. Clearly this is ...