Stockfish 221221 vs Kayra 1.1 (400 games)

[dorsz]

Recently, a new interesting engine has appeared, derivative of Stockfish - Kayra. Many users say it is stronger than Stockfish, but the results are contradictory. I had to check it myself. Here is a match between Kayra 1.1 vs Stockfish 221221 it is based on.

Match conditions:
CPU: i5-3570 - 3 cores 3.6GHz used for this match, "modern" compiles used for both engines
Time: 1m + 1s
Tablebases: 6 pieces Syzygy on SSD
Opening test suite: DOTS4 - 200 positions with colors reversed (https://www.open-chess.org/viewtopic.php?f=7&t=3329)

Result:
https://www.mediafire.com/view/xpapwfid ... 1.png/file#
Games in PGN:
https://www.mediafire.com/file/u1jqwrzb ... 1.pgn/file

[dorsz]

Sorry for the mistake, Kayra 1.1 is based rather on Stockfish 221121 not Stockfish 221221.

[sarona]

Thanks for the test and pgns, dorsz. I have had similar results - but at longer TCs (5' + 1"). I have also seen mediocre Kayra results, too, from other testers.

One question: could you tell me the hash size you used in the test? Perhaps 128 or 256 MB?

Thanks also for your continued development of Blue Marlin and Swordfish.

[dorsz]

Hash: 256MB

Out of curiosity, I am doing a match Stockfish 311221 vs Blue Marlin 14.7. The result will be tomorrow.

[dorsz]

Note:
Due to a bug (probably wrong settings of a match or maybe GUI bug?) the match was going fine till 255th game, 256th game was played without a book, from 257th game it started again the first position from DOTS4. This should not have a significant impact on the result of the match, but it does mean that 72 positions from the book were not played at all and some were played twice. I must look into the problem in next match.