Page 1 of 1

Gandalf 6.0 in test suite "WM-Test"

PostPosted: 23 Nov 2004, 13:57
by Manfred Meiler
Hello,

some days ago Gandalf author Steen Suurballe reported in german CSS forum very good results of his new Gandalf 6.0 engine in the test suite "Weltmeister-Test" (WM-Test): even better than "Fritz 8 Bilbao", the best of 290 tested engines in WM-Test so far.
The test suite "WM-Test" includes 100 difficult test positions from games of the various (human) chess world champions - 38 positions in king attack, 36 positions in positional playing and 26 positions in endgame. Download of the 100 test positions (also in pgn and epd) and an excel sheet with my detailed results of 260 (versions of) engines at http://www.computerschach.de/test/index.htm.

Now I ran the WM-Test with Gandalf 6.0 on my "WM-Test machine" (AMD Athlon Thunderbird 1400 Mhz) - as UCI engine under Arena with 256 mb hash ... and the results of Gandalf in this test suite were really great:
The best solve score (79% solved positions, F8-Bilbao 77%) and only because of the slower mean solution time Gandalf scored a little bit worse in WM-Test as F8-Bilbao.

See the summary of my WM-Test results of Gandalf 6.0 and 4 previous Gandalf versions below, for comparison also the results of Fritz 8 Bilbao:

Code: Select all
    WM-Test               F8-Bilbao  |    G. 6.0       G. 5.1       G. 5.0       G. 4.32h        G. 4.32f
AMD TB 1400 MHz           CB/256MB   |   UCI/256MB     CP/200MB     CP/200MB     UCI/104MB       WB/104MB
-------------------------------------+--------------------------------------------------------------------
solved K-pos. (38)          34       |     33            23            22            22            22     
solve score (K)             89%      |     87%           61%           58%           58%           58%       
rating king attack          2.760    |     2.744         2.667         2.659         2.661         2.662
.                                    |
solved P-pos. (36)          25       |     27            22            18            20            17
solve score (P)             69%      |     75%           61%           50%           56%           47%   
rating positional play      2.699    |     2.705         2.662         2.638         2.655         2.630
.                                    |       
solved E-Pos. (26)          18       |     19            15            14            12            12
solve score (E)             69%      |     73%           58%           54%           46%           46%   
rating endgame              2.700    |     2.706         2.664         2.653         2.633         2.628
.                                    |
? solve time min. (total)   2,23     |     4,27          5,15          4,49          3,83          4,37
solved pos. (total)         77       |     79            60            54            54            51%   
rating WM-Test              2.722    |     2.720         2.665         2.650         2.652         2.642
-------------------------------------+--------------------------------------------------------------------


I don't know yet whether Gandalf 6.0 is approximately able to confirm these great test suite results also in "practical" chess playing - but they are maybe an indication.

Best regards,
Manfred

Correction

PostPosted: 23 Nov 2004, 14:26
by Manfred Meiler
Sorry for the unreadable chart in my forst post - I make a new try, this time without my results of Gandalf 5.0 in the chart below:

Some days ago Gandalf author Steen Suurballe reported in german CSS forum very good results of his new Gandalf 6.0 engine in the test suite "Weltmeister-Test" (WM-Test): even better than "Fritz 8 Bilbao", the best of 290 tested engines in WM-Test so far.
The test suite "WM-Test" includes 100 difficult test positions from games of the various (human) chess world champions - 38 positions in king attack, 36 positions in positional playing and 26 positions in endgame. Download of the 100 test positions (also in pgn and epd) and an excel sheet with my detailed results of 260 (versions of) engines at http://www.computerschach.de/test/index.htm.

Now I ran the WM-Test with Gandalf 6.0 on my "WM-Test machine" (AMD Athlon Thunderbird 1400 Mhz) - as UCI engine under Arena with 256 mb hash ... and the results of Gandalf in this test suite were really great:
The best solve score (79% solved positions, F8-Bilbao 77%) and only because of the slower mean solution time Gandalf scored a little bit worse in WM-Test as F8-Bilbao.

See the summary of my WM-Test results of Gandalf 6.0 and 4 previous Gandalf versions below, for comparison also the results of Fritz 8 Bilbao:

Code: Select all
    WM-Test               F8-Bilbao  |    G. 6.0       G. 5.1       G. 4.32h        G. 4.32f
AMD TB 1400 MHz           CB/256MB   |   UCI/256MB     CP/200MB     UCI/104MB       WB/104MB
-------------------------------------+------------------------------------------------------
solved K-pos. (38)          34       |     33            23            22            22     
solve score (K)             89%      |     87%           61%           58%           58%       
rating king attack          2.760    |     2.744         2.667         2.661         2.662
.                                    |
solved P-pos. (36)          25       |     27            22            20            17
solve score (P)             69%      |     75%           61%           56%           47%   
rating positional play      2.699    |     2.705         2.662         2.655         2.630
.                                    |       
solved E-pos. (26)          18       |     19            15            12            12
solve score (E)             69%      |     73%           58%           46%           46%   
rating endgame              2.700    |     2.706         2.664         2.633         2.628
.                                    |
? solve time min. (total)   2,23     |     4,27          5,15          3,83          4,37
solved pos. (total)         77       |     79            60            54            51%   
rating WM-Test              2.722    |     2.720         2.665         2.652         2.642
-------------------------------------+------------------------------------------------------


I don't know yet whether Gandalf 6.0 is approximately able to confirm these great test suite results also in "practical" chess playing - but they are maybe an indication.

Best regards,
Manfred