Smooth improving
authorStefano Cardanobile <stefano.cardanobile@gmail.com>
Thu, 14 Oct 2021 20:26:42 +0000 (22:26 +0200)
committerStéphane Nicolet <cassio@free.fr>
Fri, 15 Oct 2021 12:57:01 +0000 (14:57 +0200)
commit4231d99ab408674115623f42f7ff89f3f189ca23
treeecf71855d29402cd82c7ab137d734f6954fe90a8
parent580698e5e57f40dcba52b92a7f0c7a0e9ab09437
Smooth improving

Smooth dependency on improvement margin in null move search.

STC
LLR: 2.93 (-2.94,2.94) <-0.50,2.50>
Total: 17384 W: 4468 L: 4272 D: 8644
Ptnml(0-2): 42, 1919, 4592, 2079, 60
https://tests.stockfishchess.org/tests/view/61689b8a1e5f6627cc1c0fdc

LTC
LLR: 2.94 (-2.94,2.94) <0.50,3.50>
Total: 45648 W: 11525 L: 11243 D: 22880
Ptnml(0-2): 26, 4731, 13036, 4997, 34
https://tests.stockfishchess.org/tests/view/6168a12c1e5f6627cc1c0fe3

It would be interesting to test if the other pruning/reduction heuristics
in master which are using the improving variable (ie the sign of improvement)
could benefit from a smooth function of the improvement value (or maybe a
Relu of the improvement value).

closes https://github.com/official-stockfish/Stockfish/pull/3740

Bench: 4916775
src/search.cpp