Stefan Pohl Computer Chess

Home of famous UHO openings and EAS Ratinglist

Lc0 or other GPU-Neural Nets versus Stockfish 15.1 testing

Playing conditions:

Hardware: Ryzen 7 6800H 2.6GHz Notebook, RTX 3060 GPU, Windows 11 64bit, 32GB RAM

Cuda version installed: Cuda 11.7

Speed: Stockfish 15.1 plays with 14 Threads (=7 cores) and reaches 10 MN/s in the middlegame. Lc0 minibatchsize parameter is set to the best value for each netsize, depending on Lc0's benchmark with backendbench --clippy.

Hash: 2 GB Hash for Stockfish 15.1 / (NNCacheSize 1000000 or 8192 RamLimitMb for Lc0)

GUI: Cutechess-cli (GUI ends game, when a 5-piece endgame is on the board)

Tablebases: None for engines, 5 Syzygy for cutechess-cli

Openings: UHO_2022_6mvs_+120_+129.pgn. Download my UHO 2022 openings here

Ponder, Large Memory Pages & learning: Off

Thinking time: 2min+2sec for Lc0 and 1min+1sec for Stockfish 15.1: I measured nps on my system and compared these values with the TCEC: My CPU is way too fast, compared with Lc0 running on my RTX 3060 GPU, so it makes sense to set the thinking-time of Stockfish to only 50% of the thinking-time of Lc0. For compensating the fast CPU and the fact, that in TCEC Lc0 benefits from fast hardware and long thinking-time (both is better for Lc0, not for Stockfish)

One testrun takes around nearly 5 days. Average game-duration: 6min 45sec

Each Lc0 / Neural Net plays 1000 games vs. Stockfish 15.1 with my UHO 2022 openings

Learn more about Lc0 (getting started in a GUI, links to net-downloads, FAQs, development-informations and the Leela-Blog) here

Latest update: 2024/03/18: Lc0 0.31dev BT4-6315000 (progress to BT4-6147500, but still weaker than BT4-6077500)

Download all played games (games of the old test-setups, too): here

Program Elo + - Games Score Av.Op. Draws

1 Stockfish 15.1 avx2 : 0 3 3 25000 56.9% -49 49.5% 2 Lc0 0.31dev T3-2815 : -14 15 15 1000 48.0% 0 47.7% 3 Lc0 0.31dev BT4-6077500 : -14 16 16 1000 48.0% 0 48.7% 4 Lc0 0.31dev BT4-6315000 : -20 15 15 1000 47.1% 0 51.4% 5 Lc0 0.31dev BT4-5757500 : -21 15 15 1000 47.0% 0 52.1% 6 Lc0 0.31dev TCEC 25 SuFi : -21 15 15 1000 47.0% 0 49.8% 7 Lc0 0.31dev TCEC 25 : -22 15 15 1000 46.9% 0 52.3% 8 Lc0 0.31dev BT4-6147500 : -25 15 15 1000 46.5% 0 49.7% 9 Lc0 0.31dev BT4-5000 : -33 15 15 1000 45.3% 0 49.4% 10 Lc0 0.31dev BT3-2860 : -35 15 15 1000 45.0% 0 50.6% 11 Lc0 0.31dev BT4-3400 : -37 15 15 1000 44.8% 0 49.6% 12 Lc0 0.31dev 817477 : -38 15 15 1000 44.6% 0 48.2% 13 Lc0 0.30dev T1-4000 : -39 14 14 1000 44.5% 0 49.8% 14 Lc0 0.31dev 817886 : -39 15 15 1000 44.4% 0 50.4% 15 Lc0 0.30dev 811107 : -41 15 15 1000 44.1% 0 46.1% 16 Lc0 0.30dev TCEC 24 : -42 16 16 1000 44.1% 0 51.0% 17 Lc0 0.30rc1 T1-4000 : -44 15 15 1000 43.7% 0 49.8% 18 Lc0 0.30dev BT2-4510 : -45 16 16 1000 43.5% 0 47.5% 19 Lc0 0.30dev T1-30875 : -45 15 15 1000 43.5% 0 47.5% 20 Lc0 0.30.0 815863 : -73 16 16 1000 39.8% 0 47.8% 21 Lc0 0.30rc2 814174 : -80 15 15 1000 38.8% 0 51.0% 22 Lc0 0.30dev 813207 : -84 15 15 1000 38.3% 0 49.6% 23 Lc0 0.30dev TCEC 20 : -90 15 15 1000 37.5% 0 50.5% 24 Lc0 0.30dev T1-2432500 : -94 16 16 1000 36.9% 0 47.2% 25 Lc0 0.30dev TCEC 22 : -95 15 15 1000 36.8% 0 49.4% 26 Lc0 0.30dev TCEC 18 : -133 16 16 1000 31.9% 0 50.5%

Games : 25000 (finished)

White Wins : 12531 (50.1 %) Black Wins : 93 (0.4 %) Draws : 12376 (49.5 %)

Below the gamebase recalculated with my Gamepairs Rescorer Batch-Tool. Realizing Vondele's (Stockfish maintainer) idea: "Thinking uniquely in game pairs makes sense with the biased openings used these days. While pentanomial makes sense it is a bit complicated so we could simplify and score game pairs only (not games) as W-L-D (a traditional score of 2-0, or 1.5-0.5 is just a W)."

# PLAYER : RATING ERROR PLAYED W D L (%) CFS(%) 1 Stockfish 15.1 avx2 : 0 ---- 12500 4859 6217 1424 63.7 100 2 Lc0 0.31dev T3-2815 : -27 21 500 89 283 128 46.1 57 3 Lc0 0.31dev BT4-6077500 : -30 21 500 95 267 138 45.7 70 4 Lc0 0.31dev BT4-6315000 : -39 22 500 80 285 135 44.5 59 5 Lc0 0.31dev BT4-5757500 : -42 22 500 83 274 143 44.0 53 6 Lc0 0.31dev TCEC 25 SuFi : -44 24 500 78 282 140 43.8 52 7 Lc0 0.31dev TCEC 25 : -44 22 500 85 267 148 43.7 62 8 Lc0 0.31dev BT4-6147500 : -49 22 500 80 270 150 43.0 85 9 Lc0 0.31dev BT4-5000 : -66 22 500 66 275 159 40.7 61 10 Lc0 0.31dev BT3-2860 : -70 21 500 72 257 171 40.1 61 11 Lc0 0.31dev BT4-3400 : -75 22 500 64 267 169 39.5 57 12 Lc0 0.31dev 817477 : -78 22 500 61 269 170 39.1 54 13 Lc0 0.30dev T1-4000 : -79 23 500 62 265 173 38.9 57 14 Lc0 0.31dev 817886 : -82 22 500 53 279 168 38.5 52 15 Lc0 0.30dev 811107 : -83 22 500 53 278 169 38.4 61 16 Lc0 0.30dev TCEC 24 : -87 22 500 56 266 178 37.8 57 17 Lc0 0.30rc1 T1-4000 : -90 22 500 62 250 188 37.4 55 18 Lc0 0.30dev T1-30875 : -93 23 500 60 251 189 37.1 54 19 Lc0 0.30dev BT2-4510 : -94 22 500 60 249 191 36.9 100 20 Lc0 0.30.0 815863 : -151 25 500 34 229 237 29.7 84 21 Lc0 0.30rc2 814174 : -168 24 500 28 221 251 27.7 67 22 Lc0 0.30dev 813207 : -176 25 500 21 226 253 26.8 78 23 Lc0 0.30dev TCEC 20 : -190 24 500 25 203 272 25.3 76 24 Lc0 0.30dev T1-2432500 : -202 25 500 20 200 280 24.0 58 25 Lc0 0.30dev TCEC 22 : -206 26 500 25 186 289 23.6 100 26 Lc0 0.30dev TCEC 18 : -315 31 500 12 118 370 14.2 ---

------------------------------------------------------------------- --- Number of all Gamepairs : 12500 --- Number of drawn Gamepairs overall: 6217 (= 49.74%) --- Number of 1:1 drawn Gamepairs : 3148 (= 25.18%) --- Number of 2-draws drawn Gamepairs: 3069 (= 24.55%) -------------------------------------------------------------------

You can download my Gamepairs Rescorer Tool right here

Datenschutzerklärung