Stockfish Results: Don't Understand_Can't Begin To Explain!

[geots]

Let me be very clear. I cannot explain these results, and they have happened repeatedly over the course of time. No way is this anywhere near the first comparison. But I can for sure tell you that it has absolutely little or nothing to do with the compiles. The speeds are fairly close, plus when it comes to compiling- Jim_Ablett_has_no_peers! So that is not the place to look. Anyway, this is not one of those results "that just happens" from time to time. As I said, I have no answer.

In the 2nd match- Stockfish is "version 300413", by Ipman/Jean-Paul. And he has done a good job with it. It is the version that scored 57% ag. Critter 1.6a and 48% ag. Houdini 3 in TomCass' 6-core testing. He obviously has some insight that even Marco and Gary cannot match.

Lastly, the difference is usually slightly larger than this one- and you don't get that from simple param changes.




Alienware AURORA_R4
Intel i7 w/6 True Cores
Fritz 11 gui
1CPU/64bit
128MB hash
Bases=NONE
Ponder_Learning=OFF
Perfect 2012b.ctg w/12-move limit
40/3 Repeating (Benched to adapt to 40/4)
Match=100 games

Code: Select all

SP8-i7, 3'/40+3'/40+3'/40  5/12/2013  


Strelka 5.6 x64               +53    +30/-15/=55   57.50%   57.5/100  
Stockfish 3 JA 64bit SSE4.2   -53    +15/-30/=55   42.50%   42.5/100

Code: Select all

SP8-i7, 3m/40+3m/40+3m/40  5/13/2013  

                                       
Strelka 5.6 x64               +3    +22/=57/-21   50.50%   50.5/100
Stockfish 3 64bit SSE4.2 IP   -3    +21/=57/-22   49.50%   49.5/100

Now you know as much as I do-

george

[Martin Thoresen]

Hi George,

Shouldn't you know by now that 100 games is like a lottery?

[geots]

Hi George,

Shouldn't you know by now that 100 games is like a lottery?

Yes, if I had not run exactly 23 other 100 game matches with them ag. other common opponents, and repeated some of these 2 and 3 times. 50 elo is actually as close as it has been.


Best,