TOP 20:
# PLAYER : RATING ERROR POINTS PLAYED (%) 1 Stockfish 15 : 3694 46 801.0 1020 78.5% 2 Shashchess 24.0 : 3679 53 318.0 408 77.9% 3 Berserk 9.0 : 3515 50 235.0 408 57.6% 4 Koivisto 8.0 : 3471 47 401.0 680 59.0% 5 Slowchess 2.8 : 3441 49 253.0 408 62.0% 6 Rubichess 2021 NNUE : 3388 59 83.0 204 40.7% 7 Seer 2.5 : 3360 55 116.0 272 42.6% 8 Rebel 15.1 : 3340 52 138.0 408 33.8% 9 Minic 3.30 : 3322 50 128.5 408 31.5% 10 Arasan 23.3 NNUE : 3291 52 138.0 408 33.8% 11 Fire 8.1 : 3291 41 343.0 680 50.4% 12 Igel 3.05 : 3282 38 816.0 1564 52.2% 13 Black Marlin 7 : 3274 53 104.0 408 25.5% 14 Pedone 3.1 : 3274 39 608.0 1020 59.6% 15 Komodo 12 : 3262 38 754.5 1384 54.5% 16 Ethereal 13 classic : 3258 43 267.0 476 56.1% 17 Xiphos 0.6 : 3188 42 227.5 476 47.8% 18 Rofchade 2.3 : 3142 44 220.0 476 46.2% 19 Booot 6.5 : 3123 45 163.5 476 34.3% 20 Laser 1.7 : 3115 36 504.5 976 51.7%
Click here to download the complete rating list:
The rating list is updated to 18/09/2022, it is calculated with Ordo and it is bases on the following assumptions.
- The Elo score of Rebel 6 UCI version is fixed at 2470, therefore it acts as an “anchor” for all the other engines. I calculated this score as the average between results obtained by the engine againsts humans players in the ’90, with the sources I was able to recover in Internet (Talkchess.com, Rebel14.nl, forums…).
- Besides Rebel 6, Fritz Brains in Bahrain has been also used, at a speed of 3500 kN/s, with its rating fixed at 2800 elo.
- Time for each match has been fixed to 40 moves/120 minutes repeated, calibrated on a Pentium 90 processing power. The processing power has been emulated, after estimation by using benchmarks with real P90 results. Accordingly, on modern PC the effective match time was fixed to 40 moves/125” or 40/130” depending on the PC used for the test, with expected results similar to blitz matches. For some chess engines, such as for example Chess Titans, it was not possible to define the same match time; for this reason, I have shown for this engine, the characteristic of the PC used.
- The opening suite consists of only 34 positions, which are repeated for each engine (each engine plays each opening one as white, and one as black). For older engines, which do not uses standard UCI of XBoard interfaced (such as Battlechess for example), I’ve reproduced by hand the moves as indicated by the programs. The time for this engines has been calibrated on the processing power of a Pentium 90 as well. For all this reasons, the number of match of these engines are few, and based on random selection of the 34 opening positions. The opening suite is downloadable here.
- All the tested chess engines are freely downloadable from Intenet, except for the programs I’ve bought myself years ago, and are all limited to 1 core.
- Only the best version of Stockfish is shown in the “Top 10”, and some of the strongest derivatives (such as Shashchess for example). In the full rating list I’ve included them all (past SF versions, derivatives, clones…).
- The main goal of this rating list, made for my exclusive enjoyment, was to estimate the level of play of modern engines, with respect to the older ones, but also to evaluate the actual playing strenght of many engines I loved in my youth (the already cited Battleches, for example).