Search found 455 matches

by Robert Pope
Fri Jul 19, 2019 4:57 pm
Forum: Computer Chess Club: General Topics
Topic: Math question
Replies: 22
Views: 842

Re: Math question.

Given the big difference of played games of the 'Big Four' (43.18%, 36.48%, 10.57% and 8.08%), I obtain plausible weights IMHO although the math is not so simple. You can tweak the parameters z and a . z ~ 1.96 or z = 2 are pretty standard, but a is more subjective and susceptible of being tested. ...
by Robert Pope
Wed Jul 10, 2019 7:36 pm
Forum: Computer Chess Club: General Topics
Topic: Whatever is current - Amazing Leela
Replies: 111
Views: 6785

Re: Whatever is current - Amazing Leela

Haha, I was the same guy that a few weeks ago was saying that there's no such thing as a "true score". But it clearly does, and private books prove it, as they aim, and succeed, at having the best possible "true score" once out of book, and if it's high enough, you get a "book win" (the game was de...
by Robert Pope
Mon Jul 01, 2019 1:29 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: tuning for the uninformed
Replies: 64
Views: 13380

Re: tuning for the uninformed

Hi Folkert, The sigmoid function is used to map the score of the evaluation (which is f.i. in the range of -100 to +100) to a value in the range of 0 to 1. A score of -infinity results in a value of 0, a score of +infinity in a value of 1. A score of 0 results in a value of the sigmoid of 0.5. So i...
by Robert Pope
Thu Jun 27, 2019 7:14 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: tuning for the uninformed
Replies: 64
Views: 13380

Re: tuning for the uninformed

K has to be positive or sigmoid goes the wrong direction, so that's a critical problem.

I would suggest picking a dozen positions and checking your calculation to get K. You should be able to do that by hand and see where it is going wrong.
by Robert Pope
Thu Jun 27, 2019 2:06 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: tuning for the uninformed
Replies: 64
Views: 13380

Re: tuning for the uninformed

It looks to me like you have a scale mismatch: value_from_fen is in [0,1] calculateSigmoid(eval_score) is in [-1,1]. Are you sure? Because this is what the wiki says about it: https://www.chessprogramming.org/images/1/18/TexelTuneMathE.jpg "Ri is the result of the game corresponding to position i; ...
by Robert Pope
Wed Jun 26, 2019 5:18 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: tuning for the uninformed
Replies: 64
Views: 13380

Re: tuning for the uninformed

Something I forgot to ask: for totalError += pow(value_from_fen - calculateSigmoid(eval_score), 2); value_from_fen = 1.0 for 1-0, 0.0 for 0-1 and so on but eval_score, should it be from the point of view of the fen-string? or from white? or...? Obviously the sigmoid function (and therefore also eva...
by Robert Pope
Wed Jun 26, 2019 3:19 pm
Forum: Computer Chess Club: General Topics
Topic: evaluation that usually go up when you are winning
Replies: 14
Views: 1064

Re: evaluation that usually go up when you are winning

Eelco de Groot wrote:
Wed Jun 26, 2019 2:23 pm

After 14 full moves followed by a search of 53 plies (+ elastic :) ) you would expect to have a fairly deep horizon but the score is still only 13.65.
But it's continually increasing with depth, so Uri should be happy. :)
by Robert Pope
Wed Jun 26, 2019 2:05 pm
Forum: Computer Chess Club: General Topics
Topic: evaluation that usually go up when you are winning
Replies: 14
Views: 1064

Re: evaluation that usually go up when you are winning

Evaluation staying planted at +13 only means the engine is unable to find a clear-cut game continuation where white forces the win of more material. Which is normal because.. well, it's almost the starting position and the search tree is still immense:) Good evaluation should go up not only based o...
by Robert Pope
Mon Jun 24, 2019 9:03 pm
Forum: Computer Chess Club: General Topics
Topic: LC0-Leela0 Rating Controversy..Whats The Strength?
Replies: 7
Views: 955

Re: LC0-Leela0 Rating Controversy..Whats The Strength?

What are these statements? Lc0 T40 on an RTX 2070 GPU is stronger head to head than SF_dev on 8 i7 or Ryzen cores, from not too outrageous openings. What FIDE rating that means, you guess. I think it's impossible to _measure_ the FIDE rating of Lc0 T40 RTX 2070 in a pool of human FIDE players, as m...
by Robert Pope
Mon Jun 24, 2019 3:00 pm
Forum: Computer Chess Club: General Topics
Topic: evaluation that usually go up when you are winning
Replies: 14
Views: 1064

Re: evaluation that usually go up when you are winning

Increasing evaluations is due to the search discovering new tactical combinations that win more and more material. Aside from the missing queen, this is the opening position. The search is spending so much of its tree just developing pieces that tactical combinations just don't show up immediately. ...