Latest Stockfish 4 dev. 0510 rocks !

[beram]

The latest SF dev 0510 rocks like hell.
It is beating Houdini 3 with an average of over 58 % out of 100 games yet played at TC 3m 2s on my AMD X6 1090T 3200 both, 4CPU, ponder off

With my own testsuite it was 59% with Noomen testsuite 2012 now after 51 games 57,8%

Code: Select all

Stockfish 0510 - Houdini 3, Blitz 3m+2s 
                                
1   Stockfish 051013 64 SSE4.2  +17/=25/-8 59.00%   29.5/50
2   Houdini 3 x64               +8/=25/-17 41.00%   20.5/50

Next engine I want to test it against is Komodo 6 but my download failed (I screwed up) and I havent yet received a reaction on my support question for a new download.
So I will test it against Komodo 5.1r2 instead

Grts and Thx to the Stockfish team

[Modern Times]

Which 5th October one ? There are two.

[gerold]

Could you post a download link for that version if there is a 32 version.

Thanks,
Gerold.

[beram]

The latest, so the second with Timestamp: 1380989572

[beram]

Could you post a download link for that version if there is a 32 version.

Thanks,
Gerold.

see the website - http://abrok.eu/stockfish/ - as usual
or this directlink: http://abrok.eu/stockfish/a0cc15ccbc5dc ... _32bit.exe

[gerold]

Could you post a download link for that version if there is a 32 version.

Thanks,
Gerold.

see the website - http://abrok.eu/stockfish/ - as usual
or this directlink: http://abrok.eu/stockfish/a0cc15ccbc5dc ... _32bit.exe

Thanks,
Gerold.

[beram]

The latest SF dev 0510 rocks like hell.
It is beating Houdini 3 with an average of over 58 % out of 100 games yet played at TC 3m 2s on my AMD X6 1090T 3200 both, 4CPU, ponder off

With my own testsuite it was 59% with Noomen testsuite 2012 now after 51 games 57,8%

Code: Select all

Stockfish 0510 - Houdini 3, Blitz 3m+2s 
                                
1   Stockfish 051013 64 SSE4.2  +17/=25/-8 59.00%   29.5/50
2   Houdini 3 x64               +8/=25/-17 41.00%   20.5/50

Next engine I want to test it against is Komodo 6 but my download failed (I screwed up) and I havent yet received a reaction on my support question for a new download.
So I will test it against Komodo 5.1r2 instead

Grts and Thx to the Stockfish team

Noomen testsuite finished with Houdini winning a few in the last 9 games
So now after 110 games I get over 57% and so this is the greatest win over Houdini 3 sofar
The first win I had of a SF dev over Houdini was of SF dev 3008 with 52%

Code: Select all

Stockfish 0510 - Houdini 3, Blitz 3m+2s  privat testsuite + Noomentestsuite
                                
1   Stockfish 051013 64 SSE4.2  +32/=62/-16 57.27%   63.0/110
2   Houdini 3 x64               +16/=62/-32 42.73%   47.0/110

Earlier results with dev 2809 2909 and the Ipman compile which I hadnt published yet are these:

Code: Select all

Stockfish 2809 - Komodo 5.1r2, Blitz 3m 2s
                                
1   Stockfish 280913 64 SSE4.2  +21/=21/-8 63.00%   31.5/50
2   Komodo 5.1r2 64-bit         +8/=21/-21 37.00%   18.5/50

Stockfish 2809 -Houdini3, Blitz 3m+2s 
                                
1   Houdini 3 x64               +15/=24/-11 54.00%   27.0/50
2   Stockfish 280913 64 SSE4.2  +11/=24/-15 46.00%   23.0/50


Stockfish 2909 - Komodo 5.1r2, Blitz 3m+2s
                                
1   Stockfish 290913 64 SSE4.2  +18/=27/-5 63.00%   31.5/50
2   Komodo 5.1r2 64-bit         +5/=27/-18 37.00%   18.5/50

Stockfish 2909 -Houdini3, Blitz 3m+2s  0
                              
1   Houdini 3 x64               +14/=29/-7 57.00%   28.5/50
2   Stockfish 290913 64 SSE4.2  +7/=29/-14 43.00%   21.5/50

Stockfish 3009_IP - Komodo 5.1r2, 3m+2s
                                
1   Stockfish 300913 64 SSE4.2  +15/=23/-12 53.00%   26.5/50
2   Komodo 5.1r2 64-bit         +12/=23/-15 47.00%   23.5/50

Stockfish 3009_IP - Houdini 3, Blitz 3m 2s
                                
1   Stockfish 300913 64 SSE4.2  +18/=20/-12 56.00%   28.0/50
2   Houdini 3 x64               +12/=20/-18 44.00%   22.0/50

[shrapnel]

Sorry, but I don't agree. The last few SF dev. versions have all been retrograde and are getting weaker and weaker in my opinion.
You're doing Blitz tests, which are not really relevant at assessing the strength of an Engine; only its Time Management efficiency is judged !
Try playing at slightly longer TCs.
You will get very different results !

[M ANSARI]

What changes are there from the sept version?

[Eelco de Groot]

What changes are there from the sept version?

I think Larry tested the Stockfish development version from september 28th and functionally that is still the same Stockfish, it should give exactly the same moves in single thread operation as the one tested by Bram. Only some code clean up and the bench has more positions, so the bench takes a little longer. Can't comment on how it would be different at very long timecontrols, as far as I know there are no data about that. If it's better or worse I have no idea. But compared to sept 28 no difference.

Eelco