The Week in Chess

Monday, July 28, 2014

Stockfish 14072703 x64 - Elite Gauntlet Matches - 100 Rounds, 3M+2S

Stockfish 14072703 x64 is a UCI chess engine by +Marco Costalba, +Joona Kiiski, +Tord Romstad released last July 27, 2014.

This is the first rating list publication which features the latest Stockfish chess engine since the release of version 5 two months ago. The reason was because there was practically no ELO improvement with the many patches that were released after version 5. The estimated ELO rating of this latest Stockfish version 14072703 showed only about 5 ELO rating increase which is small.

Stockfish 14072703 scored 58.75% with 125 wins, 55 losses and 220 draws against the top Elite Chess Engines notably Houdini, Komodo and Gull. Version 5 was not taken out in the list to provide the basis for comparison with the rest of the elite chess engines.

Note that this is a longer time control with 3 minutes base + 2 seconds increment and it will not appear in the regular rating list at 1M+1S.

Here is the performance of Stockfish  14072703:
.
Rank Engine Est. ELO Raw ELO Games Score% Points Win Loss Draw
1 Stockfish 14072703 x64 3164.11 48.90 400 58.75 235.0 125 55 220
2 Stockfish 5 x64 3159.42 36.35 100 48.00 48.0 12 16 72
3 Houdini 4 Pro x64 3101.97 0.31 100 43.50 43.5 20 33 47
4 Komodo 7a x64 3096.96 -26.78 100 39.00 39.0 13 35 52
5 Gull 3 x64 3068.01 -58.78 100 34.50 34.5 10 41 49
.
Download the gauntlet PGN games here.

Saturday, July 26, 2014

Gaviota 1.0 x64 - Gauntlet Matches, 100 Rounds, 1M1S

Gaviota 1.0 x64 is a UCI/WInboard chess engine by Miguel Ballicora released last March 9, 2014. This version was previously tested with gauntlet matches against the top chess engines selection but was aborted due to so many losses by time forfeit. It was using the Winboard protocol which was automatically selected by the Arena Chess GUI. By luck I was able to figure out that it could run the UCI Protocol without much time losses. The gauntlet results showed that Gaviota deserves a spot in the Top Chess Engines Selection.

Gaviota scored 31.68% with 338 wins, 1034 losses and 528 draws in the 100 rounds gauntlet matches against the top 19 chess engines selection which earned an ELO rating of 2729.59 and the 19th spot. The ELO rating increase over the last version is a big 100 ELO points.

The Fire 3.0 chess engine was retired from the top selection to reduce the number of Ippolit clones and to make way for new or improving chess engines.

Here is the gauntlet performance of Gaviota:
.
Rank Engine True ELO Raw ELO Games Score% Points Win Loss Draw Change
1 Stockfish 5 x64 3162.54 175.62 100 88.00 88.0 78 2 20 -5.05
2 Houdini 4 Pro x64 3102.97 171.38 100 85.50 85.5 79 8 13 -3.01
3 Komodo 7a x64 3098.06 153.92 100 87.00 87.0 75 1 24 -4.38
4 Gull 3 x64 3070.17 135.11 100 84.50 84.5 73 4 23 -3.78
5 Critter 1.6a x64 3008.32 130.72 100 83.50 83.5 73 6 21 -2.27
6 Equinox 2.02 x64 2966.14 94.55 100 80.00 80.0 68 8 24 -2.12
7 Rybka 4.1 x64 2954.64 58.07 100 75.50 75.5 63 12 25 -2.40
8 Senpai 1.0 x64 2782.09 51.87 100 75.50 75.5 61 10 29 2.91
9 Protector 1.6.0 x64 2840.43 -2.79 100 69.00 69.0 51 13 36 -0.45
10 Spike 1.4 2806.58 -18.51 100 65.50 65.5 53 22 25 -0.10
11 Hannibal 1.4b x64 2827.83 -50.17 100 61.00 61.0 50 28 22 -1.73
12 Deep Hiarcs 14 2814.25 -64.55 100 60.50 60.5 39 18 43 -1.91
13 Naum 4.2 x64 2776.54 -66.43 100 59.00 59.0 46 28 26 -1.46
14 Texel 1.04 x64 2818.58 -67.38 100 59.50 59.5 41 22 37 -2.66
15 DiscoCheck 5.2 x64 2709.07 -93.07 100 55.50 55.5 41 30 29 0.79
16 Sjeng 2010 2747.06 -95.77 100 55.50 55.5 38 27 35 -0.30
17 Shredder 12 x64 2826.62 -108.03 100 53.50 53.5 39 32 29 -3.01
18 Junior 13.8.04 x64 2732.69 -123.48 100 51.50 51.5 36 33 31 -1.11
19 Gaviota 1.0 x64 2729.59 -133.46 1900 31.68 602.0 338 1034 528 2729.59
20 Murka 3 x64 2714.06 -147.60 100 48.00 48.0 30 34 36 -1.13
.
Download the gauntlet PGN games here.

Owl Computer Chess Engines Rating List - 07/27/2014

The Owl Computer Chess Engines Rating List released, 07/27/2014.

View the full rating list here.

Thursday, July 24, 2014

Mars 2.2 x64 - Gauntlet Matches, 100 Rounds 1M1S

Mars 2.2 x64 by Trap is a UCI chess engine by Trap released last July 20, 2014.

After so many version releases in a span of a month, the Mars chess engine finally has improvement worthy of ELO rating publication.

Mars 2.2 scored 52.5% with 410 wins, 320 losses and 1070 draws against the selection of Top Ippolit chess engines. It posted an ELO rating of 2990.78 which is 8 ELO higher than version 1.5 and the number 5 spot in the rankings that displaced its father clone, Fire 3.0.
.
Rank Engine True ELO Raw ELO Games Score% Points Win Loss Draw Change
1 Houdini 4 Pro x64 3105.98 160.88 100 71.50 71.5 47 4 49 -0.93
2 Strelka 6 3060.54 109.07 100 63.50 63.5 37 10 53 -0.88
3 Robodini 1.1 x64 3048.73 38.47 100 53.00 53.0 26 20 54 -2.18
4 Critter 1.6a x64 3010.59 19.58 100 50.50 50.5 24 23 53 -1.61
5 Mars 2.2 x64 2990.78 16.09 1800 52.50 945.0 410 320 1070 2990.78
6 Fire 3.0 x64 2983.53 8.88 100 49.00 49.0 19 21 60 -1.40
7 PanChess 00.537 x64 2966.17 -5.39 100 47.00 47.0 18 24 58 -1.36
8 Mars 1.5 x64 2982.37 -7.11 100 46.50 46.5 12 19 69 -2.08
9 LEOpard 0.7c x64 2940.30 -11.37 100 46.00 46.0 16 24 60 -1.04
10 Firenzina 2.4 xTreme x64 2971.38 -14.04 100 45.50 45.5 12 21 67 -1.67
11 Vitruvius 1.11C x64 2935.49 -18.94 100 45.00 45.0 14 24 62 -0.92
12 Bouquet 1.8 x64 2975.21 -21.47 100 44.50 44.5 15 26 59 -1.85
13 Igorrit 0.086v9 x64 2959.80 -26.98 100 43.50 43.5 10 23 67 -1.62
14 Saros eXp R5 x64 2971.04 -28.07 100 43.50 43.5 11 24 65 -2.40
15 RobboLito 0.21Q x64 2957.32 -37.13 100 42.50 42.5 15 30 55 -1.70
16 Tactico Power 2011 x64 2945.56 -37.31 100 42.00 42.0 11 27 62 -1.51
17 Ivanhoe 46h x64 2948.72 -40.00 100 42.00 42.0 14 30 56 -1.59
18 Black Mamba 2.0 x64 2920.10 -45.94 100 40.50 40.5 9 28 63 -0.94
19 Akkad 0.52b x64 2885.77 -59.22 100 39.00 39.0 10 32 58 -0.67

.
Download the gauntlet PGN games here.

Owl Computer Chess Engines Rating List - 07/24/2014

The Owl Computer Chess Engines Rating List released, 07/24/2014.

View the full rating list here.

Monday, July 21, 2014

Stockfish 14072002 x64 vs. Stockfish 5 x64 - Test 3M2S, 100 Rounds x 4

It's been a long time since Stockfish has been released and there is no good news for it since then. The last test match I made was more than a month ago and it showed a regression with my test environment. There was a continuous test of the Stockfish development versions at different settings, computers, number of CPU cores, time controls, etc. The results showed no improvement or regression in some patches like the latest patch with contempt = 20.

To give some idea on the progress of Stockfish since version 5, I decided to make a match with the latest version 14072002 released yesterday, July 20 at longer time control of 3 minutes base + 2 seconds increment for 100 rounds with 4 tournament instances running simultaneously in the same computer which was also used in previous matches.

In the total of 400 games, Stockfish 14072002 defeated Stockfish 5 with a score of 202 -198 which is a 4 point advantage.  This translates to 1 point advantage to every 100 games played which is just too stastistically very  small to be considered likely to be superior. Both contestants won two 100 round matches each but the later version was lucky to score higher in some parts.
 
Here is the summary of the match in version 14072002 perspective:

Stockfish 14072002 x64 vs. Stockfish 5 x64
Test Match 100R x 4, 3 Min + 2 Secs


Description Games Points Win Loss Draw
Part 1 100 51.5 13 10 77
Part 2 100 49.5 8 9 83
Part 3 100 52.5 13 8 79
Part 4 100 48.5 9 12 79
Total 400 202.0 43 39 318

Download the PGN games played here.

Chessdom News