Reckless Tests

[mehmet123]

Reckless vs Obsidain:

Program Elo + - Games Score Av.Op. Draws

1 Obsidian 16.03 avx2 : 3018 9 7 480 52.7 % 3000 92.9 %
2 Reckless 0.80 dev 0106 : 3000 7 9 480 47.3 % 3018 92.9 %


Individual statistics:

1 Obsidian 16.03 avx2 : 3018 480 (+ 30,=446,- 4), 52.7 %

Reckless 0.80 dev 0106 : 480 (+ 30,=446,- 4), 52.7 %

2 Reckless 0.80 dev 0106 : 3000 480 (+ 4,=446,- 30), 47.3 %

Obsidian 16.03 avx2 : 480 (+ 4,=446,- 30), 47.3 %


Game Conditions: Cute Chess Gui, Core-i7 12700h, 1 Cores, 3 min + 2 sec TC, Balsa Opening Book (5 moves), 128 Mb Hash, Ponder Off
https://www.mediafire.com/file/m02yxx1o ... 3.pgn/file


The Reckless chess engine has been showing extraordinary improvement, especially over the last 3 months.
https://github.com/codedeliveryservice/Reckless
According to CEGT 40/20, the elo difference between Reckless 0.7.0 (1 CPU) and Obsidian 15.0 (1 CPU) is -225 Elo. Based on CCRL 40/15, the elo difference between Reckless 0.7.0 (1 CPU) and Obsidian 16.0 (1 CPU) is - 190 Elo.
Out of 480 games, 30 resulted in Reckless losing on time. However, in 28 of those 30 games, Obsidian had a clearly winning position (+6.0 or more). Of the two remaining games, one was a drawn position and the other was a win for Reckless. If we adjust for those two games, the elo difference between the two engines drops slightly from 18 to 17 according to these test conditions.

[mehmet123]

Reckless vs Stockfish:

Program Elo + - Games Score Av.Op. Draws

1 Stockfish 250525 bmi2 : 3030 11 9 450 54.2 % 3000 90.2 %
2 Reckless 0.80 dev 0106 : 3000 9 11 450 45.8 % 3030 90.2 %


Individual statistics:

1 Stockfish 250525 bmi2 : 3030 450 (+ 41,=406,- 3), 54.2 %

Reckless 0.80 dev 0106 : 450 (+ 41,=406,- 3), 54.2 %

2 Reckless 0.80 dev 0106 : 3000 450 (+ 3,=406,- 41), 45.8 %

Stockfish 250525 bmi2 : 450 (+ 3,=406,- 41), 45.8 %


Game Conditions: Cute Chess Gui, Core-i7 12700h, 1 Cores, 3 min + 2 sec TC, Balsa Opening Book (5 moves), 128 Mb Hash, Ponder Off
https://www.mediafire.com/file/vlogmp9m ... 4.pgn/file


In this match, there are also many games that Reckless lost due to time. However, aside from 3 games, the situation was already a clear loss for Reckless (with a +6.0 or more favor of Stockfish). In one game, Stockfish lost on time in a drawn position. If we make these corrections, Stockfish holds a +28 elo advantage over Reckless at these conditions.

[mehmet123]

Reckless vs PlentyChess:

Program Elo + - Games Score Av.Op. Draws

1 PlentyChess 5.0.0 bmi2 : 3010 8 6 430 51.4 % 2395 95.3 %
2 Reckless 0.8.0 dev 0306 : 3000 6 8 430 48.6 % 2405 95.3 %


Individual statistics:

1 PlentyChess 5.0.0 bmi2 : 3010 430 (+ 16,=410,- 4), 51.4 %

Reckless 0.8.0 dev 0306 : 430 (+ 16,=410,- 4), 51.4 %

2 Reckless 0.8.0 dev 0306 : 3000 430 (+ 4,=410,- 16), 48.6 %

PlentyChess 5.0.0 bmi2 : 430 (+ 4,=410,- 16), 48.6 %


Game Conditions: Cute Chess Gui, Core-i7 12700h, 1 Cores, 3 min + 2 sec TC, Balsa Opening Book (5 moves), 256 Mb Hash, Ponder Off
https://www.mediafire.com/file/dsk3uhi5 ... 3.pgn/file

PlentyChess is number 4 engine behind Stockfish, Torch and Obsidian at CEGT (40/20 - Single Versions)
There are 15 time loss game for Reckless. 5 of them is clear win for PlentyChess and rest of them is drawn games.
If we ignore the clear draws lost due to time, Reckless is only 1 or 2 elo behind of PlentyChess according to this test. It would' t be wrong to say that Reckless is now one of the top 6-7 most powerful chess engines.

[mehmet123]

Reckless vs Obsidian:

Program Elo + - Games Score Av.Op. Draws

1 Obsidian dev 16.07 avx2 : 3014 9 7 450 52.0 % 3000 93.8 %
2 Reckless 0.8.0 dev 0906 : 3000 7 9 450 48.0 % 3014 93.8 %

Individual statistics:

1 Obsidian dev 16.07 avx2 : 3014 450 (+ 23,=422,- 5), 52.0 %

Reckless 0.8.0 dev 0906 : 450 (+ 23,=422,- 5), 52.0 %

2 Reckless 0.8.0 dev 0906 : 3000 450 (+ 5,=422,- 23), 48.0 %

Obsidian dev 16.07 avx2 : 450 (+ 5,=422,- 23), 48.0 %


Game Conditions: Cute Chess Gui, Core-i7 12700h, 1 Cores, 3 min + 2 sec TC, Balsa Opening Book (5 moves), 128 Mb Hash, Ponder Off
https://www.mediafire.com/file/us3ndvr6 ... 5.pgn/file


Out of 450 games, 23 resulted in Reckless losing on time. However, in 21 of those 30 games, Obsidian had a clearly winning position (+6.0 or more). Of the two remaining games, one was a drawn position and the other was a win for Reckless. If we adjust for those two games, the elo difference between the two engines drops slightly from 14 to 13 according to these test conditions.