Some tests with Stockfish dev.

Discussion of computer chess matches and engine tournaments.

Moderator: Ras

ThatsIt
Posts: 992
Joined: Thu Mar 09, 2006 2:11 pm

Some tests with Stockfish dev.

Post by ThatsIt »

Hi to all !

Just for fun.

Code: Select all

GUI:     Shredder-Classic                       PC: Intel i5-2400 @ 3.1GHz
Level:   Game in 8" + 1" per move               OS: Windows 7 Professional
         (around 2 min. and 46 sec. per game)
Ponder:  off                                        100 starting-positions
HTs:     64MB per engine                             = 200 games per match
Learing: completely disabled
Bases:   3+4                                          All engines x64 1CPU
Stockfish 4.0 (release)

Code: Select all

                                |            | Ø move/|  + - |  Margin |
                                |   +  =  -  |   game |  ELO |  (95%)  |
------------------------------------------------------------------------
vs Houdini 3.0    00.0 - 000.0  | (xx xx xx) |    00  | +- 0 | +xx -xx | still running
------------------------------------------------------------------------
vs Komodo 6.0     00.0 - 000.0  | (xx xx xx) |    00  | +- 0 | +xx -xx | still running
------------------------------------------------------------------------
Total 000.0 out of 000 = xx.x% = +- 0 (+00 -00)
Stockfish 20102013 (stockfish_13102023_x64_modern_sse42)

Code: Select all

                                |            | Ø move/|  + - |  Margin |
                                |   +  =  -  |   game |  ELO |  (95%)  |
------------------------------------------------------------------------
vs Houdini 3.0    92.5 - 107.5  | (53 79 68) |    94  | - 26 | +38 -38 |
------------------------------------------------------------------------
vs Komodo 6.0    101.5 -  98.5  | (53 97 50) |    83  | +  5 | +35 -35 |
------------------------------------------------------------------------
Total 194.0 out of 400 = 48.5% = - 10 (+26 -26)
Stockfish 27102013 (stockfish_13102708_x64_modern_sse42)

Code: Select all

                                |            | Ø move/|  + - |  Margin |
                                |   +  =  -  |   game |  ELO |  (95%)  |
------------------------------------------------------------------------
vs Houdini 3.0   107.0 -  93.0  | (68 78 54) |    93  | + 24 | +38 -38 |
------------------------------------------------------------------------
vs Komodo 6.0    107.0 -  93.0  | (63 88 49) |    89  | + 24 | +36 -36 |
------------------------------------------------------------------------
Total 214.0 out of 400 = 53.5% = + 24 (+26 -26)
More to come ...

Best wishes,
G.S.
Top
Modern Times
Posts: 3807
Joined: Thu Jun 07, 2012 11:02 pm

Re: Some tests with Stockfish dev.

Post by Modern Times »

ThatsIt wrote:
Stockfish 27102013 (stockfish_13102708_x64_modern_sse42)

Code: Select all

                                |            | Ø move/|  + - |  Margin |
                                |   +  =  -  |   game |  ELO |  (95%)  |
------------------------------------------------------------------------
vs Houdini 3.0   107.0 -  93.0  | (68 78 54) |    93  | + 24 | +38 -38 |
------------------------------------------------------------------------
vs Komodo 6.0    107.0 -  93.0  | (63 88 49) |    89  | + 24 | +36 -36 |
------------------------------------------------------------------------
Total 214.0 out of 400 = 53.5% = + 24 (+26 -26)
More to come ...

Best wishes,
G.S.
Stockfish has made amazing progress. I wonder what Houdini 4 will bring.
Top
ThatsIt
Posts: 992
Joined: Thu Mar 09, 2006 2:11 pm

Re: Some tests with Stockfish dev.

Post by ThatsIt »

Update.

Code: Select all

GUI:     Shredder-Classic                       PC: Intel i5-2400 @ 3.1GHz
Level:   Game in 8" + 1" per move               OS: Windows 7 Professional
         (around 2 min. and 46 sec. per game)
Ponder:  off                                        100 starting-positions
HTs:     64MB per engine                             = 200 games per match
Learing: completely disabled
Bases:   3+4                                          All engines x64 1CPU
Stockfish 4.0 (release)

Code: Select all

                                |            | Ø move/|  + - |  Margin |
                                |   +  =  -  |   game |  ELO |  (95%)  |
------------------------------------------------------------------------
vs Houdini 3.0    89.5 - 110.5  | (55 69 76) |    95  | - 37 | +39 -39 |
------------------------------------------------------------------------
vs Komodo 6.0     94.5 - 105.5  | (45 99 56) |    88  | - 19 | +34 -34 |
------------------------------------------------------------------------
Total 184.0 out of 400 = 46.0% = - 28 (+26 -26)
Stockfish 20102013 (stockfish_13102023_x64_modern_sse42)

Code: Select all

                                |            | Ø move/|  + - |  Margin |
                                |   +  =  -  |   game |  ELO |  (95%)  |
------------------------------------------------------------------------
vs Houdini 3.0    92.5 - 107.5  | (53 79 68) |    94  | - 26 | +38 -38 |
------------------------------------------------------------------------
vs Komodo 6.0    101.5 -  98.5  | (53 97 50) |    83  | +  5 | +35 -35 |
------------------------------------------------------------------------
Total 194.0 out of 400 = 48.5% = - 10 (+26 -26)
Stockfish 27102013 (stockfish_13102708_x64_modern_sse42)

Code: Select all

                                |            | Ø move/|  + - |  Margin |
                                |   +  =  -  |   game |  ELO |  (95%)  |
------------------------------------------------------------------------
vs Houdini 3.0   107.0 -  93.0  | (68 78 54) |    93  | + 24 | +38 -38 |
------------------------------------------------------------------------
vs Komodo 6.0    107.0 -  93.0  | (63 88 49) |    89  | + 24 | +36 -36 |
------------------------------------------------------------------------
Total 214.0 out of 400 = 53.5% = + 24 (+26 -26)
More to come ...

Best wishes,
G.S.
Top
Maharadja
Posts: 78
Joined: Thu Dec 24, 2009 1:22 pm

Re: Some tests with Stockfish dev.

Post by Maharadja »

ThatsIt wrote:Update.

Best wishes,
G.S.

Hi,

What's the update? looks like the first post.
Top
ThatsIt
Posts: 992
Joined: Thu Mar 09, 2006 2:11 pm

Re: Some tests with Stockfish dev.

Post by ThatsIt »

Maharadja wrote: Hi,
What's the update? looks like the first post.
The results of Stockfish 4.0 (release).

Best wishes,
G.S.
Top
lkaufman
Posts: 6284
Joined: Sun Jan 10, 2010 6:15 am
Location: Maryland USA
Full name: Larry Kaufman

Re: Some tests with Stockfish dev.

Post by lkaufman »

I wonder why you choose to test with a base to increment ratio of only 8 to 1, when everyone else uses a ratio of 100 or even more? Using a small ratio is very inefficient, it means that almost as much time is spent per move playing out 200 move drawn endings as in complex middlegames. Also, no one uses ratios like this in development, so you are testing something that no one ever tried to optimize. Maybe the results wouldn't differ by much, but if you want games of this length the "LightSpeed" time control of 45" plus half a second increment would produce much higher quality games in about the same time, and would give a much better indication of relative strength at bullet chess levels.

Best, Larry
Top
ThatsIt
Posts: 992
Joined: Thu Mar 09, 2006 2:11 pm

Re: Some tests with Stockfish dev.

Post by ThatsIt »

lkaufman wrote:I wonder why you choose to test with a base to increment ratio of only 8 to 1, when everyone else uses a ratio of 100 or even more? Using a small ratio is very inefficient, it means that almost as much time is spent per move playing out 200 move drawn endings as in complex middlegames. Also, no one uses ratios like this in development, so you are testing something that no one ever tried to optimize. Maybe the results wouldn't differ by much, but if you want games of this length the "LightSpeed" time control of 45" plus half a second increment would produce much higher quality games in about the same time, and would give a much better indication of relative strength at bullet chess levels.
Best, Larry
Hi Larry !

2 reasons:
1st: just for fun
and
2nd: to do any different.

Just for fun is the main reason of course.

btw.:

Code: Select all

the longest games until now:
Stockfish 211013 vs Komodo 6.0  = 257 moves (draw)
Stockfish 271013 vs Houdini 3.0 = 248 moves (draw)
Stockfish 271013 vs Houdini 3.0 = 243 moves (draw)
Houdini 3.0 vs Stockfish 211013 = 242 moves (draw)
Stockfish 211013 vs Houdini 3.0 = 240 moves (draw)

Code: Select all

the longest games with a win/lost until now:
Houdini 3.0 vs Stockfish 4.0    = 222 moves (0-1)
Stockfish 211013 vs Komodo 6.0  = 183 moves (1-0)
Houdini 3.0 vs Stockfish 271013 = 182 moves (0-1)
Stockfish 4.0 vs Houdini 3.0    = 174 moves (1-0)
Komodo 6.0 vs Stockfish 4.0     = 170 moves (1-0)
No time losses at all so far after 1400 ultra bullet games.

Best wishes,
G.S.
Top
Maharadja
Posts: 78
Joined: Thu Dec 24, 2009 1:22 pm

Re: Some tests with Stockfish dev.

Post by Maharadja »

ThatsIt wrote:
Maharadja wrote: Hi,
What's the update? looks like the first post.
The results of Stockfish 4.0 (release).

Best wishes,
G.S.
hi,

thx!!! I need new glasses :shock:
Top

Return to “Computer Chess Club: Tournaments and Matches”