deepcrayon
/
stockfish
1
0
Fork 0
Code Releases Activity

Update WDL model for NNUE 

This updates the WDL model based on the LTC statistics in June this year (10M games),
so from pre-NNUE to NNUE based results.

(for old results see, https://github.com/official-stockfish/Stockfish/pull/2778)

As before the fit by the model to the data is quite good.

closes https://github.com/official-stockfish/Stockfish/pull/3582

No functional change
pull/3595/head
Joost VandeVondele 2021-06-23 07:23:21 +02:00
parent e47b74457e
commit dc4983327d
1 changed files with 3 additions and 3 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

6
src/uci.cpp
View File

@ -207,13 +207,13 @@ namespace {
// Coefficients of a 3rd order polynomial fit based on fishtest data
// for two parameters needed to transform eval to the argument of a
// logistic function.
double as[] = {-8.24404295, 64.23892342, -95.73056462, 153.86478679};
double bs[] = {-3.37154371, 28.44489198, -56.67657741, 72.05858751};
double as[] = {-3.68389304, 30.07065921, -60.52878723, 149.53378557};
double bs[] = {-2.0181857, 15.85685038, -29.83452023, 47.59078827};
double a = (((as[0] * m + as[1]) * m + as[2]) * m) + as[3];
double b = (((bs[0] * m + bs[1]) * m + bs[2]) * m) + bs[3];
// Transform eval to centipawns with limited range
double x = std::clamp(double(100 * v) / PawnValueEg, -1000.0, 1000.0);
double x = std::clamp(double(100 * v) / PawnValueEg, -2000.0, 2000.0);
// Return win rate in per mille (rounded to nearest)
return int(0.5 + 1000 / (1 + std::exp((a - x) / b)));