Discussion of anything and everything relating to chess playing software and machines.
Moderators: hgm , Rebel , chrisw
Uri Blass
Posts: 10282 Joined: Thu Mar 09, 2006 12:37 am
Location: Tel-Aviv Israel
Post
by Uri Blass » Wed Sep 06, 2017 8:06 pm
Laskos wrote: A small rating list of top 10 engines at Blitz 360''+ 3.6''. The error margins as shown in Ordo are really 1.7-1.8 times smaller using pentanomial variance. Syzygy 6-men in use for engines supporting them.
Code: Select all
Program Elo + - Games Score Av.Op. Draws
1 Komodo 11.2.2 64-bit : 3156 40 40 90 67.8 % 2983 53.3 %
2 Deep Shredder 13 x64 : 3091 42 42 90 60.6 % 2990 54.4 %
3 Stockfish 260817 64 BMI2 : 3082 40 40 90 59.4 % 2991 47.8 %
4 Gull 3 x64 : 3057 40 40 90 56.7 % 2994 44.4 %
5 Houdini 5.01 Pro x64-popc : 3014 42 42 90 51.7 % 2998 52.2 %
6 Fire 5 x64 : 2971 42 42 90 46.7 % 3003 46.7 %
7 Fritz 15 : 2952 40 40 90 44.4 % 3005 62.2 %
8 Andscacs 0.91b : 2918 44 44 90 40.6 % 3009 56.7 %
9 Fizbo 1.9 : 2909 42 42 90 39.4 % 3010 47.8 %
10 Booot 6.2_x64 : 2849 42 42 90 32.8 % 3017 56.7 %
Games : 450 (finished)
White Wins : 6 (1.3 %)
Black Wins : 209 (46.4 %)
Draws : 235 (52.2 %)
Unfinished : 0
White Score : 27.4 %
Black Score : 72.6 %
Observe the underperformance of Stockfish in this disbalance, although it used Syzygy 6-men from SSD.
I think that you observe underperformance only because you compare stockfish with Komodo or shredder.
It seems to me that the difference of stockfish from houdini is higher
3 Stockfish 260817 64 BMI2 : 3082 40 40 90 59.4 % 2991
5 Houdini 5.01 Pro x64-popc : 3014 42 42 90 51.7 % 2998 52.2 %
Laskos
Posts: 10948 Joined: Wed Jul 26, 2006 10:21 pm
Full name: Kai Laskos
Post
by Laskos » Wed Sep 06, 2017 8:46 pm
Fern, do you have something like a link for me to somehow track it?
Btw, I hope your health is better, and even better with a glass of wine.
Laskos
Posts: 10948 Joined: Wed Jul 26, 2006 10:21 pm
Full name: Kai Laskos
Post
by Laskos » Wed Sep 06, 2017 8:48 pm
Uri Blass wrote: Laskos wrote: A small rating list of top 10 engines at Blitz 360''+ 3.6''. The error margins as shown in Ordo are really 1.7-1.8 times smaller using pentanomial variance. Syzygy 6-men in use for engines supporting them.
Code: Select all
Program Elo + - Games Score Av.Op. Draws
1 Komodo 11.2.2 64-bit : 3156 40 40 90 67.8 % 2983 53.3 %
2 Deep Shredder 13 x64 : 3091 42 42 90 60.6 % 2990 54.4 %
3 Stockfish 260817 64 BMI2 : 3082 40 40 90 59.4 % 2991 47.8 %
4 Gull 3 x64 : 3057 40 40 90 56.7 % 2994 44.4 %
5 Houdini 5.01 Pro x64-popc : 3014 42 42 90 51.7 % 2998 52.2 %
6 Fire 5 x64 : 2971 42 42 90 46.7 % 3003 46.7 %
7 Fritz 15 : 2952 40 40 90 44.4 % 3005 62.2 %
8 Andscacs 0.91b : 2918 44 44 90 40.6 % 3009 56.7 %
9 Fizbo 1.9 : 2909 42 42 90 39.4 % 3010 47.8 %
10 Booot 6.2_x64 : 2849 42 42 90 32.8 % 3017 56.7 %
Games : 450 (finished)
White Wins : 6 (1.3 %)
Black Wins : 209 (46.4 %)
Draws : 235 (52.2 %)
Unfinished : 0
White Score : 27.4 %
Black Score : 72.6 %
Observe the underperformance of Stockfish in this disbalance, although it used Syzygy 6-men from SSD.
I think that you observe underperformance only because you compare stockfish with Komodo or shredder.
It seems to me that the difference of stockfish from houdini is higher
3 Stockfish 260817 64 BMI2 : 3082 40 40 90 59.4 % 2991
5 Houdini 5.01 Pro x64-popc : 3014 42 42 90 51.7 % 2998 52.2 %
I am not sure, but I guess that all in all, compared to performance from regular openings, Stockfish underperforms, and Houdini underperforms even more.
Laskos
Posts: 10948 Joined: Wed Jul 26, 2006 10:21 pm
Full name: Kai Laskos
Post
by Laskos » Thu Sep 07, 2017 12:50 pm
Laskos wrote: Fern, do you have something like a link for me to somehow track it?
Btw, I hope your health is better, and even better with a glass of wine.
Got it this morning to the door, many thanks! Already saw some new to me Greek and Roman "orgies"
Laskos
Posts: 10948 Joined: Wed Jul 26, 2006 10:21 pm
Full name: Kai Laskos
Post
by Laskos » Wed Oct 04, 2017 12:41 pm
Laskos wrote: Uri Blass wrote: Laskos wrote: A small rating list of top 10 engines at Blitz 360''+ 3.6''. The error margins as shown in Ordo are really 1.7-1.8 times smaller using pentanomial variance. Syzygy 6-men in use for engines supporting them.
Code: Select all
Program Elo + - Games Score Av.Op. Draws
1 Komodo 11.2.2 64-bit : 3156 40 40 90 67.8 % 2983 53.3 %
2 Deep Shredder 13 x64 : 3091 42 42 90 60.6 % 2990 54.4 %
3 Stockfish 260817 64 BMI2 : 3082 40 40 90 59.4 % 2991 47.8 %
4 Gull 3 x64 : 3057 40 40 90 56.7 % 2994 44.4 %
5 Houdini 5.01 Pro x64-popc : 3014 42 42 90 51.7 % 2998 52.2 %
6 Fire 5 x64 : 2971 42 42 90 46.7 % 3003 46.7 %
7 Fritz 15 : 2952 40 40 90 44.4 % 3005 62.2 %
8 Andscacs 0.91b : 2918 44 44 90 40.6 % 3009 56.7 %
9 Fizbo 1.9 : 2909 42 42 90 39.4 % 3010 47.8 %
10 Booot 6.2_x64 : 2849 42 42 90 32.8 % 3017 56.7 %
Games : 450 (finished)
White Wins : 6 (1.3 %)
Black Wins : 209 (46.4 %)
Draws : 235 (52.2 %)
Unfinished : 0
White Score : 27.4 %
Black Score : 72.6 %
Observe the underperformance of Stockfish in this disbalance, although it used Syzygy 6-men from SSD.
I think that you observe underperformance only because you compare stockfish with Komodo or shredder.
It seems to me that the difference of stockfish from houdini is higher
3 Stockfish 260817 64 BMI2 : 3082 40 40 90 59.4 % 2991
5 Houdini 5.01 Pro x64-popc : 3014 42 42 90 51.7 % 2998 52.2 %
I am not sure, but I guess that all in all, compared to performance from regular openings, Stockfish underperforms, and Houdini underperforms even more.
It seems the things have changed with Houdini 6, though I tested at short time control:
Code: Select all
# PLAYER : RATING ERROR POINTS PLAYED (%) CFS(next)
1 Houdini 6.02 Pro x64-pext : 3214.3 25.2 348.0 480 72.5 99
2 Brainfish 021017 64 BMI2 : 3171.3 23.0 327.0 480 68.1 65
3 Komodo 11.2.2 64-bit : 3165.3 22.5 324.0 480 67.5 100
4 Gull 3 x64 : 3112.9 23.5 297.0 480 61.9 59
5 Deep Shredder 13 x64 : 3109.1 22.8 295.0 480 61.5 100
6 Andscacs 0.92 : 3048.4 22.2 262.5 480 54.7 78
7 Fire 6.1 x64 popcnt : 3035.5 22.6 255.5 480 53.2 55
8 Fritz 15 : 3033.6 21.8 254.5 480 53.0 100
9 Hakkapeliitta TCEC : 2958.7 22.0 214.0 480 44.6 67
10 Booot 6.2_x64 : 2951.2 22.2 210.0 480 43.8 100
11 Texel 1.07tuning : 2875.4 23.9 170.5 480 35.5 100
12 Fizbo 1.9 : 2742.7 27.1 109.0 480 22.7 100
13 Hannibal 1.7 x64 : 2581.5 35.2 53.0 480 11.0 ---
Brainfish is just a very fast compile of Stockfish, no Cerebellum book or this stuff.