Maelstrom 3.1.1 - new version chess engine

Maelstrom - UCI-compliant Golang chess engine in development from scratch.

Bugfix - 3-fold repetition issue

Full Changelog: v3.1.0...v3.1.1

Estimated ELO: ~3070

vs. Maelstrom v3.1.0

STC:

Score of Maelstrom v3.1.1 vs Maelstrom v3.1.0: 557 - 522 - 921 [0.509] 2000

... Maelstrom v3.1.1 playing White: 317 - 225 - 458 [0.546] 1000

... Maelstrom v3.1.1 playing Black: 240 - 297 - 463 [0.471] 1000

... White vs Black: 614 - 465 - 921 [0.537] 2000

Elo difference: 6.1 +/- 11.2, LOS: 85.7 %, DrawRatio: 46.1 %

SPRT: llr 0.55 (19.0%), lbound -2.25, ubound 2.89

LTC:

Score of Maelstrom v3.1.1 vs Maelstrom v3.1.0: 209 - 189 - 426 [0.512] 824

... Maelstrom v3.1.1 playing White: 122 - 83 - 208 [0.547] 413

... Maelstrom v3.1.1 playing Black: 87 - 106 - 218 [0.477] 411

... White vs Black: 228 - 170 - 426 [0.535] 824

Elo difference: 8.4 +/- 16.5, LOS: 84.2 %, DrawRatio: 51.7 %

SPRT: llr 0.419 (14.5%), lbound -2.25, ubound 2.89

vs. Stash 27 (LTC)

Score of Maelstrom v3.1.1 vs Stash 27: 763 - 601 - 472 [0.544] 1836

... Maelstrom v3.1.1 playing White: 423 - 275 - 220 [0.581] 918

... Maelstrom v3.1.1 playing Black: 340 - 326 - 252 [0.508] 918

... White vs Black: 749 - 615 - 472 [0.536] 1836

Elo difference: 30.7 +/- 13.7, LOS: 100.0 %, DrawRatio: 25.7 %

Changelist:

Fix threefold repetition issue found in #7 where engine would go into threefold when mate was found by only performing TT cutoffs at non-PV nodes

Fix UCI reporting statistics to report total time and nodes, not time/nodes per depth

Change mate score to return a ply-based score rather than WIN_VAL and convert it to mate depth when reporting UCI statistics

Instead of stopping search immediately when mate is found, only stop when mate found is in <= 3 moves

Maelstrom 3.1.1 download

Chess Engines Diary