Why is Rybka 3 w32 (1-cpu) so dominant in Blitz play? Is there ANY engine that can defeat it presently?
>Why is Rybka 3 w32 (1-cpu) so dominant in Blitz play?
I wasn't aware that it was, do you mean against humans or engines? If so, which engines?
>Is there ANY engine that can defeat it presently?
Uhm sure, as any chess engine rating site would show. To give some examples:
Rybka 4.1
Houdini 1.5 and above (latest version 4.0)
Stockfish 2.0 and above (latest version at time of writing 6.0)
Komodo 3.0 and above (latest version at time of writing, 9.3)
http://www.computerchess.org.uk/ccrl/404/cgi/compare_engines.cgi?class=32-bit+Single-CPU+engines&print=Rating+list&print=Results+table&print=LOS+table&table_size=12&cross_tables_for_best_versions_only=1
It would appear that Rybka 3 is still on top. Perhaps I could have been a little more specific in saying 40/4 play.
http://www.husvankempen.de/nunn/40_4_Ratinglist/40_4_AllVersion/rangliste.html
The following are 40/4, w32, 1 CPU:
#41 Stockfish 6.0 3177
#42 Komodo 9.2 3174
#59 Houdini 4.0 3115
#64 Stockfish 5.0 3108
#78 Stockfish DD 3073
#97 Gull 3.0 3030
#100 Komodo 7.0a 3028
#104 Stockfish 4.0 3023
#122 Houdini 1.5 3011
#127 Komodo TCEC 3003
#137 Strelka 5.0 2994
#138 Critter 1.6 2991
#142 Critter 1.4 2989
#147 Critter 1.2 2981
#150 Komodo 6.0 2978
#151 Equinox 3.30 2978
#167 Stockfish 3.0 2962
#180 Komodo CCT 2947
#184 Stockfish 2.2.2 2941
#187 Stockfish 2.3 2934
#192 Stockfish 2.1.1 2930
#194 Critter 1.0 2928
#195 Stockfish 1.7.1 2927
#196 Stockfish 2.2.1 2926
#199 Rybka 4.0 2922
#204 Sting SF 2916
#206 Stockfish 2.0 2914
#208 Komodo 5.0r1 2912
#211 Stockfish 1.9 2908
#227 BlackMamba 2.0 2892
#228 Chiron 2.0 2892
#236 Komodo 3.0 2883
#239 Texel 1.05 2873
#241 Naum 4.6 2872
#243 Stockfish 1.8 2869
#245 Critter 0.90 2869
#254 Stockfish 1.71 2858
#255 Rybka 3.0 2858
The list has 1,663 entries, so R3 is listed above another 1,408 entries of various engines, engine versions and core counts, nonetheless Rybka 3.0 w32 1 CPU is -319 Elo compared to Stockfish 6.0 w32 1 CPU on this list.
Thats not what I was talking about. The link I provided is dated November 28th, 2015, so I don't think it is outdated at all. And, the link I provided only compares 1 cpu 32 bit with other 1 cpu 32 bits.
Am I missing something here?
36 Stockfish DD x64 4CPU 3190 13 13 3050 71.6% 2999 37.0%
37 Houdini 4.0 x64 2CPU 3189 16 16 2000 57.6% 3122 38.6%
38 Fritz 15 x64 12CPU 3185 36 36 200 49.8% 3187 59.5%
39 Komodo TCEC x64 4CPU 3184 18 18 1000 66.3% 3059 40.5%
40 Gull 3.0 x64 4CPU 3183 12 12 4650 60.7% 3084 41.7%
41 Stockfish 6.0 w32 1CPU 3177 13 13 2200 73.5% 2984 39.6%
42 Komodo 9.2 w32 1CPU 3174 19 19 1100 69.5% 3021 41.1%
43 Houdini 3.0 x64 2CPU 3167 14 14 1800 72.9% 2980 33.8%
44 Gull 2.8 beta x64 4CPU 3164 19 19 1000 63.0% 3064 42.3%
45 Stockfish 5.0 x64 1CPU 3159 10 10 4050 60.1% 3077 43.5%
46 Houdini 2.0c x64 4CPU 3158 13 13 3200 78.7% 2901 28.2%
47 Komodo 7.0a x64 2CPU 3154 17 17 1050 57.4% 3098 44.1%
48 Stockfish 4.0 x64 4CPU 3144 15 15 2500 72.3% 2951 36.5%
49 Komodo 8.0 x64 1CPU 3143 11 11 3150 58.6% 3074 43.6%
50 Komodo 6.0 x64 4CPU 3140 13 13 3000 68.1% 2984 38.7%
51 Gull 3.0 x64 2CPU 3140 14 14 2350 48.7% 3147 47.4%
52 Houdini 1.5 x64 4CPU
>>Correct me if I am mistaken, but the link you provided shows all versions compared to each other. So a 1 cpu 32 bit is being compared to a 12 cpu 32 bit.
You can click on the link of each engine to see which games it played.
>>Thats not what I was talking about. The link I provided is dated November 28th, 2015, so I don't think it is outdated at all.
That list is absolutely outdated. Just look at the versions of the programs. Rybka 3 is the latest version of Rybka listed, 13 is the latest version of Fritz, 2.01 is the latest on Stockfish, and Komodo 1.3 is the latest of Komodo. Rybka 3 is over 7 years old. I think the date at the top of that page just reports when any page on the site was last updated.
What I think is happening here is that CCRL simply stopped testing 32-bit engines around 5 years ago, so when you search for 32-bit only you're effectively setting a latest date of around then. Well it's a bit more complicated than that, they appeared to have stopped testing 32-bit versions of an engine if a 64-bit version was available, and did not test any new 32-bit engines after that. Maybe Graham can clarify.
>>And, the link I provided only compares 1 cpu 32 bit with other 1 cpu 32 bits.
That's what's listed, the engines actually play against all different ones, including 64 bit engines, and engines running on more than one core. Again, you can click on an entry to see which games it played, but there aren't going to be much for some of them because like I said I think they stopped testing 32-bit engines a while ago.
>>Am I missing something here?
Also it's kind of a moot point that the 32-bit engines on 1 core are being tested against other 64-bit engines on multiple cores and all the combinations therein. Elo is Elo, the higher rated engine is surely stronger here given the number of games and number of competitors.
> What I think is happening here is that CCRL simply stopped testing 32-bit engines around 5 years ago, so when you search for 32-bit only you're effectively setting a latest date of around then. Well it's a bit more complicated than that, they appeared to have stopped testing 32-bit versions of an engine if a 64-bit version was available, and did not test any new 32-bit engines after that. Maybe Graham can clarify.
Correct.
We're all testing the 64-bit versions of engines if available.
The only 32-bit engines still being tested are those that do not have 64-bit versions.
Powered by mwForum 2.27.4 © 1999-2012 Markus Wichitill