EN-Tournament - Tournament mode engine test started!
This test makes it possible to estimate the tournament playing strength of an engine within one day. From various sources, I have selected 100 short variants of the most important openings, which also occur in the games on the well-known servers (Lichess, InfinityChess, PlayChess.com). I've even left out some of the most played lines, because 99% of the games end in a draw. At the last InfinityChess Engine-Masters tournament, only 12 variations were allowed, where White had an advantage of +0.80 each (Stockfish analysis). This advantage was enough to win 4 games out of 120 played (I didn't include losses due to operator error and disconnects). This is the trend today in computer chess on the servers! But as long as there are such tournaments there, as well as such engine game rooms where all possible opening books are allowed in freestyle mode, I will test the engines for such games!
For Position test: Download ENET-2023 Test (PGN & CBH, 1 10 Positions)
I test with 60s/position on 14 threads
(ENET-2024 Test: Coming soon at end 2023)
For ENG Matches: Download 100 short variants (CBH and PGN format) as "EN-Tournament":
Pixeldrain & Filehorst.de
My settings in tournament mode:
Threads - Either Ryzen 3900X or Ryzen 2700, 6 threads per engine.
GUI Fritz 18 - Offline I use Fritz 18 and play in ENG-ENG mode.
Hashtables - 128 MB (Bullet) to 1024 MB (Blitz 5 Min).
Ponder ON (Permanent Brain) - Ponder ON is also played on all 3 servers.
3456men Syzygy - All types of databases can be used on all servers.
Time control - 2 to 5 minutes, 0s to 3s bonus time, offline I use 0.1s bonus.
MoveOverhead - The engine should not lose on time forfeit - Offline I use 100ms for Stockfish.
Opening Book - Offline I use EN-Tournament.cbh with color swap, 200 games in total.
06. May 2023: Stockfish 16 (use 1536 NNUE) vs Stockfish dev 010723 (use 2048 NNUE)
Hardware Ryzen 2700, hash 512 MB, time control 5m + 0.1s
Final result: +5 =191 -4 as seen by Stockfish 16 --> eng-eng live picture.
SF 16 uses a smaller NNUE network otherwise the engines are identical.
Download all 200 games in PGN:
https://pixeldrain.com/u/SXEHBWHP
07. May 2023: Stockfish 16 (use 1536 NNUE) vs Little Beast 7 sl (use 2048 NNUE)
Hardware Ryzen 2700, hash 512 MB, time control 5m + 0.1s
Final result: +3 =191 -6 as seen by Stockfish 16.
Download all 200 games in PGN:
https://pixeldrain.com/u/Y4MdG2YZ
25. July 2023: Network Test - Dimension 2560 vs Dimension 2048
I did another test today. Already in my first test, the new unofficial dimension 2560 won against the SF dev standard network with dimension 2048. My private engine Smile XPro is not identical to SF dev. It is a modified SF version, and adapted to the new NNUE Dimension 2560.
SF dev used network "nn-c38c3d8d3920.nnue" (Dimension 2028)
Smile XPro used network "nn-1ceb1ade2048.nnue" (Dimension 2560)
After 36 games, Smile XPro was able to achieve a +4 lead. So it is clear to me that the new NNUE net with dimension 2560 plays better than SF dev when played with enough time. Bullet games often end randomly, in 10 minute games the engines take more time (sometimes even more than 60s per move). Ponder ON is very important to me, because we play with it on all servers.
The games was played on Ryzen 2700 with the Fritz 18 GUI.
6 Threads/Engine
Ponder ON
Timecontrol 10m + 0.1s
Hash 512 MB/Engine
Book EN-Tournament with color swap
Syzygy all 3456men
Result:
+2 =28 -6 as seen by Stockfish dev 160723
All games as PGN:
https://pixeldrain.com/u/nZeChHBk
09. Sep 2023: Short Bullet-Test Sun Light vs Stockfish dev 20230903:
Ryzen 2700
GUI Fritz 18
Timecontrol 60s+0.6s
6 Threads/Engine, Hash 128 MB, all 3456men Syzygy, Ponder ON
Book EN-Tournament with color swap
Result:
+4 =42 -4 as seen by Stockfish dev 20230903
All games as PGN:
https://pixeldrain.com/u/HLDHzmzW
05. Oct 2023: Stockfish dev 20230928 vs Light 6.0:
Hardware Ryzen 2700
GUI Fritz 18, Hash 512 MB/Engine, 6 threads/engine, Ponder ON, all 3456men Syzygy.
Timecontrol 5m + 0.1s
Book EN-Select, 50 variants with color swap
+6 =84 -10 as seen by Stockfish dev 20230928
All games as PGN:
https://pixeldrain.com/u/wrYvcmmZ
03. Dec 2023: Little Beast 12.1 (private Engine) vs Eman 9.80 (private Engine).
GUI Fritz 18
Ryzen 3900X
6 Threads/Engine (about 5400 kns/Engine)
Hash 512 MB/Engine
Ponder ON
Timecontrol 5m + 0.1s
Syzygy all 3456men
Default Settings
Book EN-Select 100 games with color swap
Result:
+5 =87 -8 as seen by Eman 9.80
All games as PGN:
https://pixeldrain.com/u/yVucmyP3
After I implemented some newest Stockfish patches until Dec 02 in Little Beast (called now Little Beast 12.2): In a quick test of 24 games under the same conditions as the first test, Eman 9.80 losed again (+3 = 16 -5).
Download PGN:
https://pixeldrain.com/u/jb2FmL8p
Copyright by Eduard Nemeth. All rights reserved.