Tuned Futility Equation
authorprotonspring <mike@whiteley.org>
Wed, 21 Aug 2019 01:52:18 +0000 (19:52 -0600)
committerSt├ęphane Nicolet <stephanenicoletsuriphone@gmail.com>
Wed, 21 Aug 2019 07:34:46 +0000 (09:34 +0200)
commit18279b24fc76bb6eaf6ac01f3032b1b90da5dabb
tree372f2cd00de8dc9a362833320c458683a1f2c459
parentd4dca9187e83dde29be8d76ca50ff53d14199ce9
Tuned Futility Equation

@Vizvezdenec array suggested that alternate values may be better than current
master (see pull request #2270 ). I tuned some linear equations to more closely
represent his values and it passed. These futility values seem quite sensitive,
so perhaps additional Elo improvements can be found here.

STC
LLR: 2.95 (-2.94,2.94) [0.50,4.50]
Total: 12257 W: 2820 L: 2595 D: 6842
http://tests.stockfishchess.org/tests/view/5d5b2f360ebc5925cf1111ac

LTC
LLR: 2.96 (-2.94,2.94) [0.00,3.50]
Total: 20273 W: 3497 L: 3264 D: 13512
http://tests.stockfishchess.org/tests/view/5d5c0d250ebc5925cf111ac3

Closes https://github.com/official-stockfish/Stockfish/pull/2272

------------------------------------------
How to continue from there ?

a) we can try a simpler version for the futility margin, this would
   be a simplification :
    margin = 188 * (depth - improving)

b) on the other direction, we can try a complexification by trying
   again to gain Elo with an complete array of futility values.

------------------------------------------

Bench: 4330402
src/search.cpp