Closed eduherminio closed 4 weeks ago
See also https://github.com/lynx-chess/Lynx/pull/917, the bucketization attempt of single TaperedEvaluationTerm.
TaperedEvaluationTerm
Tuned at 37k epoch
Test | eval/psqt-TaperedEvaluationTermArray-custom23buckets Elo | -2.23 +- 3.41 (95%) SPRT | 8.0+0.08s Threads=1 Hash=32MB LLR | -2.29 (-2.25, 2.89) [0.00, 3.00] Games | 19816: +5690 -5817 =8309 Penta | [654, 2413, 3864, 2360, 617] https://openbench.lynx-chess.com/test/637/
After enabling mobility values addition to the piece values, still kinda neutral
Test | eval/psqt-TaperedEvaluationTermArray-custom23buckets Elo | -1.01 +- 4.05 (95%) SPRT | 8.0+0.08s Threads=1 Hash=32MB LLR | -2.25 (-2.25, 2.89) [0.00, 5.00] Games | 13822: +3928 -3968 =5926 Penta | [421, 1693, 2744, 1611, 442] https://openbench.lynx-chess.com/test/640/
See also https://github.com/lynx-chess/Lynx/pull/917, the bucketization attempt of single
TaperedEvaluationTerm
.Tuned at 37k epoch
After enabling mobility values addition to the piece values, still kinda neutral