Chess Engines Diary

Chess engine for Android - Kobra 2.0 Rating CEDR=??? Kobra 2.0 for Android download Chess engine for Android - PZChessBot dev-20250421T23 JA Rating CEDR=3574 PZChessBot dev-20250421T23 JA  for Android download Chess engine for Android - Seawall 20250322 JA Rating CEDR=3163 Seawall 20250322 JA for Android download Chess engine for Android - Shallow Guess 0.1.0 JA Rating CEDR=??? Shallow Guess 0.1.0 JA for Android download Chess engine for Android - Stash 37 JA Rating CEDR=3574 Stash 37 JA for Android download Chess engine for Android - Stockfish 17.1 Rating CEDR=3163 Stockfish 17.1 for Android download Chess engine for Android - Superultra 2.1 JA Rating CEDR=3574 Superultra 2.1 JA for Android download Chess engine for Android - Timecat 1.45.0 JA Rating CEDR=3163 Timecat 1.45.0 JA for Android download

Superultra (NNUE) Rating Chess Engines Diary CEDR=3488 Superultra is a UCI compliant chess engine that uses an alpha-beta search framework, various heuristics and optimization techniques, and an efficiently updatable neural network (NNUE) in order to analyze chess positions and calculate moves. v.2.1: Major Changes: Significantly reduced required stack size Changed all types away from (u)int_fastN_t Style Elo (STC 6+0.06) Score of Superultra-2.1 vs Superultra-2.0: 805 - 676 - 1292  [0.523] 2773 ...      Superultra-2.1 playing White: 455 - 282 - 650  [0.562] 1387 ...      Superultra-2.1 playing Black: 350 - 394 - 642  [0.484] 1386 ...      White vs Black: 849 - 632 - 1292  [0.539] 2773 Elo difference: 16.2 +/- 9.4, LOS: 100.0 %, DrawRatio: 46.6 % SPRT: llr 2.95 (100.0%), lbound -2.94, ubound 2.94 - H1 was accepted Elo (LTC 30+0.1) Score of Superultra-2.1 vs Superultra-2.0: 1346 - 1210 - 2837  [0.513] 5393 ...  ...

Superultra (NNUE) Rating Chess Engines Diary CEDR=3488 Superultra is a UCI compliant chess engine that uses an alpha-beta search framework, various heuristics and optimization techniques, and an efficiently updatable neural network (NNUE) in order to analyze chess positions and calculate moves. v.2.1: Major Changes: Significantly reduced required stack size Changed all types away from (u)int_fastN_t Style Elo (STC 6+0.06) Score of Superultra-2.1 vs Superultra-2.0: 805 - 676 - 1292  [0.523] 2773 ...      Superultra-2.1 playing White: 455 - 282 - 650  [0.562] 1387 ...      Superultra-2.1 playing Black: 350 - 394 - 642  [0.484] 1386 ...      White vs Black: 849 - 632 - 1292  [0.539] 2773 Elo difference: 16.2 +/- 9.4, LOS: 100.0 %, DrawRatio: 46.6 % SPRT: llr 2.95 (100.0%), lbound -2.94, ubound 2.94 - H1 was accepted Elo (LTC 30+0.1) Score of Superultra-2.1 vs Superultra-2.0: 1346 - 1210 - 2837  [0.513] 5393 ...  ...

Superultra (NNUE) Rating CEDR=3488 Superultra is a UCI compliant chess engine that uses an alpha-beta search framework, various heuristics and optimization techniques, and an efficiently updatable neural network (NNUE) in order to analyze chess positions and calculate moves. v.2.0: Major Changes: Revamped NNUE Changed hidden layer size from 384 to 512 Changed king bucket count from 6 to 10 (mirrored) Created 8 output weight buckets that are selected based on the number of pieces left More training data from the same dataset More relaxed quantization Improvements Introduced cut-nodes Optimized QS SEE and Probcut Rewrote time manager score fluctuation multiplier Decay old histories New features Ponder Support for various UCI Search specifications such as depth limit and node limit New Makefile Elo (STC 6+0.06) Score of V2.0 vs V1.0: 189 - 88 - 284  [0.590] 561 ...      V2.0 playing White: 112 - 40 - 130  [0.628] 282 ...      V2.0 playing Black: 77...

Superultra (NNUE) Superultra is a UCI compliant chess engine that uses an alpha-beta search framework, various heuristics and optimization techniques, and an efficiently updatable neural network (NNUE) in order to analyze chess positions and calculate moves. Superultra v1.0 download