Page 1 of 1

Time control comparison between engines

Posted: Wed Jan 01, 2014 10:57 pm
by Rebel
I wrote a (PGN) utility that compares the time control of engines.

Example (1) taken from the tournament & matches section.

Code: Select all

Engine                 Depth       Time   Games     Moves  Average Forfeit
Arminius 2013-08-10    17.07    6:25:18      40      2771    8.34     0
Naraku 1.4 [1c]        17.05    6:16:24      40      2769    8.16     0
Delfi 5.4 [2c]         15.50    5:48:23      40      2693    7.76     0
Chronos 1.9.9 [1c]     15.76    6:58:13      40      3268    7.68     0
Zappa 1.1 [4c]         17.52    5:41:00      40      2668    7.67     2
Djinn 1.008 [1c]       22.62    6:37:22      40      3141    7.59     0
Atlas 3.60 [1c]        18.96    5:23:34      40      2582    7.52     0
Frenzee 3.5.19 [1c]    15.84    5:52:04      40      2897    7.29     0
NirvanaChess 1.3 [1    18.73    6:01:39      40      2985    7.27     0
Hamsters 0.7.1 [4c]    18.96    6:54:46      40      3447    7.22     0
Deuterium 13.1.31.1    18.02    5:46:11      40      2883    7.20     0
Nebula 2.0 [1c]        20.49    6:40:41      40      3355    7.17     0
Naum 2.0 [1c]          14.89    6:34:05      40      3352    7.05     0
Rodent 1.1 [1c]        22.54    6:45:03      40      3461    7.02     0
Bison 9.11 w32 [1c]    15.73    6:06:41      40      3183    6.91     0
Rodent 1.2 [1c]        25.04    6:07:26      40      3206    6.88     0
Alfil 13.1 [4c]        16.75    5:58:46      40      3157    6.82     1
Pro Deo 1.87 [1c]      15.47    6:10:08      40      3475    6.39     0
Pro Deo 1.86 [1c]      15.32    5:48:23      40      3292    6.35     0
Jonny 4 [4c]           18.74    4:49:25      40      2894    6.00     0
Thinker 5.4d Active    0.00    0:00:00      40        15    0.00     0


Time control comparison between engines

Depth     : Average search depth
Time      : Total time engine used
Moves     : Total moves engine played
Average   : Average time per move
Forfeit   : Games engine lost due to time forfeit

List is sorted on Average Time indicating the engine that uses the most time tops.
The list seems to indicate Arminius spends its time best (not saying it has the best time control, that's something else) and mine likely needs an update after 20+ years.

Example (2) taken from a recent CEGT (40/20) download.

Code: Select all

Engine                 Depth       Time   Games     Moves  Average Forfeit
Delphil 3.1 x64 1CP    13.62   20:52:22     100      5454    13.78     0
Naum 4.2 x64 4CPU      19.39   25:31:28     100      6714    13.69     0
Naum 4.2 x64 2CPU      18.49    6:24:08      25      1700    13.56     0
WildCat 5              14.77   21:17:23     100      5689    13.47     0
Stockfish 4.0 x64 4    27.80    6:10:09      20      1658    13.40     0
Naum 4.2 x64 1CPU      18.77   23:04:10     100      6211    13.37     0
Stockfish DD x64 1C    29.47   23:03:47     100      6294    13.19     0
Cheng4 v0.36 x64 1C    22.50   20:05:20     100      5636    12.83     0
Hannibal 1.4a x64 1    23.75   21:08:06     100      6034    12.61     0
Deep Rybka 4 x64 2C    18.02    4:30:48      20      1299    12.51     0
Onno 0.12.0 x64        18.33   19:33:14     100      5656    12.45     0
Gull 2.2 x64 1CPU      21.46   25:51:45     100      7604    12.24     0
Critter 1.6 x64 1CP    19.70   21:12:56     100      6590    11.59     0
Houdini 3 x64 1CPU     29.93   23:01:15     100      7181    11.54     0
Houdini 4.0 x64 1CP    26.95   20:50:30     100      6862    10.93     0
Komodo TCEC x64 1CP    18.00  188:08:56    1000     67569    10.02     0

Time control comparison between engines

Depth     : Average search depth
Time      : Total time engine used
Moves     : Total moves engine played
Average   : Average time per move
Forfeit   : Games engine lost due to time forfeit

List is sorted on Average Time indicating the engine that uses the most time tops.
Oddly Houdini & Komodo leave quite some time on the clock but perhaps this is due to a different time control (40/8) the CEGT guys used. Things are not so clear from the PGN.

Anyway, I will double check the tool for bugs and then release it.

Re: Time control comparison between engines

Posted: Wed Jan 01, 2014 11:04 pm
by bob
Rebel wrote:I wrote a (PGN) utility that compares the time control of engines.

Example (1) taken from the tournament & matches section.

Code: Select all

Engine                 Depth       Time   Games     Moves  Average Forfeit
Arminius 2013-08-10    17.07    6:25:18      40      2771    8.34     0
Naraku 1.4 [1c]        17.05    6:16:24      40      2769    8.16     0
Delfi 5.4 [2c]         15.50    5:48:23      40      2693    7.76     0
Chronos 1.9.9 [1c]     15.76    6:58:13      40      3268    7.68     0
Zappa 1.1 [4c]         17.52    5:41:00      40      2668    7.67     2
Djinn 1.008 [1c]       22.62    6:37:22      40      3141    7.59     0
Atlas 3.60 [1c]        18.96    5:23:34      40      2582    7.52     0
Frenzee 3.5.19 [1c]    15.84    5:52:04      40      2897    7.29     0
NirvanaChess 1.3 [1    18.73    6:01:39      40      2985    7.27     0
Hamsters 0.7.1 [4c]    18.96    6:54:46      40      3447    7.22     0
Deuterium 13.1.31.1    18.02    5:46:11      40      2883    7.20     0
Nebula 2.0 [1c]        20.49    6:40:41      40      3355    7.17     0
Naum 2.0 [1c]          14.89    6:34:05      40      3352    7.05     0
Rodent 1.1 [1c]        22.54    6:45:03      40      3461    7.02     0
Bison 9.11 w32 [1c]    15.73    6:06:41      40      3183    6.91     0
Rodent 1.2 [1c]        25.04    6:07:26      40      3206    6.88     0
Alfil 13.1 [4c]        16.75    5:58:46      40      3157    6.82     1
Pro Deo 1.87 [1c]      15.47    6:10:08      40      3475    6.39     0
Pro Deo 1.86 [1c]      15.32    5:48:23      40      3292    6.35     0
Jonny 4 [4c]           18.74    4:49:25      40      2894    6.00     0
Thinker 5.4d Active    0.00    0:00:00      40        15    0.00     0


Time control comparison between engines

Depth     : Average search depth
Time      : Total time engine used
Moves     : Total moves engine played
Average   : Average time per move
Forfeit   : Games engine lost due to time forfeit

List is sorted on Average Time indicating the engine that uses the most time tops.
The list seems to indicate Arminius spends its time best (not saying it has the best time control, that's something else) and mine likely needs an update after 20+ years.

Example (2) taken from a recent CEGT (40/20) download.

Code: Select all

Engine                 Depth       Time   Games     Moves  Average Forfeit
Delphil 3.1 x64 1CP    13.62   20:52:22     100      5454    13.78     0
Naum 4.2 x64 4CPU      19.39   25:31:28     100      6714    13.69     0
Naum 4.2 x64 2CPU      18.49    6:24:08      25      1700    13.56     0
WildCat 5              14.77   21:17:23     100      5689    13.47     0
Stockfish 4.0 x64 4    27.80    6:10:09      20      1658    13.40     0
Naum 4.2 x64 1CPU      18.77   23:04:10     100      6211    13.37     0
Stockfish DD x64 1C    29.47   23:03:47     100      6294    13.19     0
Cheng4 v0.36 x64 1C    22.50   20:05:20     100      5636    12.83     0
Hannibal 1.4a x64 1    23.75   21:08:06     100      6034    12.61     0
Deep Rybka 4 x64 2C    18.02    4:30:48      20      1299    12.51     0
Onno 0.12.0 x64        18.33   19:33:14     100      5656    12.45     0
Gull 2.2 x64 1CPU      21.46   25:51:45     100      7604    12.24     0
Critter 1.6 x64 1CP    19.70   21:12:56     100      6590    11.59     0
Houdini 3 x64 1CPU     29.93   23:01:15     100      7181    11.54     0
Houdini 4.0 x64 1CP    26.95   20:50:30     100      6862    10.93     0
Komodo TCEC x64 1CP    18.00  188:08:56    1000     67569    10.02     0

Time control comparison between engines

Depth     : Average search depth
Time      : Total time engine used
Moves     : Total moves engine played
Average   : Average time per move
Forfeit   : Games engine lost due to time forfeit

List is sorted on Average Time indicating the engine that uses the most time tops.
Oddly Houdini & Komodo leave quite some time on the clock but perhaps this is due to a different time control (40/8) the CEGT guys used. Things are not so clear from the PGN.

Anyway, I will double check the tool for bugs and then release it.
How does pondering fit into this? For example, a program ponders for 2 minutes, the opponent thinks for that 2 minutes and moves, and then the program searches for another minute. Does that count for 3 minutes or 1? It only used one on the clock...

Re: Time control comparison between engines

Posted: Wed Jan 01, 2014 11:25 pm
by Rebel
The util just takes the time given in the PGN. Currently it supports 2 GUI's.

Code: Select all

[Cutechess-cli] 10. exf6 {book} gxf6 {book} 11. O-O {+0.19/10 2.5s} Nc6 {-0.54/13 0.62s}

[Arena]  9. dxe5 Ng4 10. cxd5 cxd5 11. Nf3 {+0.44/14 2} Nc6 {-0.80/16 2}
Cutechess-cli has the time in centi-seconds, Arena in whole seconds. Therefore Cutechess-cli is much more accurate.

Regarding the tool, use it wisely. But whether the PB is "on" or "off" leaving a lot of time on the clock is spilling all the hard efforts you invested speeding up the thing.