Stockfish 231213 fine Christmas present

[beram]

Stockfish 231213 seems to take over command in all areas from blitz to LTC
First time that Houdini 4 lost a match in my testmatches has happened today at christmas day.
At SF testsite against master(SFDD) it showed already + 18

As shown below SF has sligthly improved in matches from SFDD, by SF Ipman151213 till today SF 231213.
All under same match conditions AMD X6 1090T 3200 Mhz, 4cpu, Blitz 3m2s, ponder = off 512 Mb hash, 25 positions openingbook, 100 games matches all
No special contempt setting and no Tablebases are used

The amount of draws has increased from SF DD to SF 2312 and is very high considering the blitz time control.
But although blitz (3m2s 4cpu), Houdini 4 searches already at average 20-22 ply in early middlegame. The chess quality of the games is very high

happy Christmas,
thx Stockfish team

Code: Select all

Stockfish DD 4x - Houdini 4, 4x Blitz 3m+2s
                            
1   Houdini 4 x64           +29/=47/-24 52.50%   52.5/100
2   Stockfish DD 64 SSE4.2  +24/=47/-29 47.50%   47.5/100


SF 1512(Ipman)4x - H4 4x, Blitz 3m+2s
                           
1   Houdini 4 x64          +26/=51/-23 51.50%   51.5/100
2   Stockfish 151213IP 64  +23/=51/-26 48.50%   48.5/100


SF 231213 4x - H4 4x, Blitz 3m+2s
                        
1   Stockfish 231213 64  +25/=56/-19 53.00%   53.0/100
2   Houdini 4 x64        +19/=56/-25 47.00%   47.0/100

[ouachita]

SF171213 v. SF231213, 1m+1s, 12/25/13 0
12 cores
~20,000 kn/s

1 Stockfish 171213 64 SSE4.2 +74 +25/=71/-4 60.50% 60.5/100
2 Stockfish 231213 64 SSE4.2 -74 +4/=71/-25 39.50% 39.5/100

Merry Christmas!

[PaulieD]

Bram, Here are some nice victories at even quicker time controls..

This is the first victory at Blitz for the new SF, I ran an interrupted match in DF14, this was a continuous match.

Do we Hail to the new Blitz King so soon!:D

Code: Select all

Program                          Elo    +   -   Games   Score   Av.Op.  Draws

  1 Stockfish 231213               : 3214   51  50   100    54.0 %   3186   46.0 %
  2 Houdini 4 Pro x64              : 3186   50  51   100    46.0 %   3214   46.0 %

SYSTEM---i3 M380|2 Cores@2.53|Win7(64bit)|6GB RAM
FMARK----Speed 4.82/Kns 2313
TB'S------None
PONDER---Off
GUI-------Deep Fritz 14(64 Bit)
HASH-----256 MB
BOOK-----5000 8 move(reversing)
TIME------3 min

Another bullet victory, this time a substantial one. This latest SF is much stronger.

[Milos]

SF171213 v. SF231213, 1m+1s, 12/25/13 0
12 cores
~20,000 kn/s

1 Stockfish 171213 64 SSE4.2 +74 +25/=71/-4 60.50% 60.5/100
2 Stockfish 231213 64 SSE4.2 -74 +4/=71/-25 39.50% 39.5/100

Merry Christmas!

What a wonderful example of what variability really means.
Engine that is 2-3 Elo stronger at the given TC losses by 80Elo.
On 100 games 2SD is 35Elo in this case. 80Elo is more than 4SD i.e. something that you need 16k events to observe, i.e. all matches ever published on talkchess are not enough to observe such anomaly even once.
So either standard model of error margin calculation is really screwed up or something is really wrong with your test system.
I think it's the second .

[ouachita]

So either standard model of error margin calculation is really screwed up or something is really wrong with your test system.
I think it's the second .

Could be the influence of the Sun's trajectory too!?

"Today's heroes are tomorrow's goats; today's goats are tomorrow's heroes."

[mwyoung]

Bram, Here are some nice victories at even quicker time controls..

This is the first victory at Blitz for the new SF, I ran an interrupted match in DF14, this was a continuous match.

Do we Hail to the new Blitz King so soon!:D

Code: Select all

Program                          Elo    +   -   Games   Score   Av.Op.  Draws

  1 Stockfish 231213               : 3214   51  50   100    54.0 %   3186   46.0 %
  2 Houdini 4 Pro x64              : 3186   50  51   100    46.0 %   3214   46.0 %

SYSTEM---i3 M380|2 Cores@2.53|Win7(64bit)|6GB RAM
FMARK----Speed 4.82/Kns 2313
TB'S------None
PONDER---Off
GUI-------Deep Fritz 14(64 Bit)
HASH-----256 MB
BOOK-----5000 8 move(reversing)
TIME------3 min

Another bullet victory, this time a substantial one. This latest SF is much stronger.

I see we are all getting around the same win percentage on 54% or higher for Stockfish when playing Houdini 4.

As for hailing the new blitz king, well the results are the results.

And I have enough games on my private rating list to say yes, Stockfish is the new blitz King. Since it holds the best rating on my ratings list.

And for me that is the results the count the most.

I even held a knock out blitz championship with the strongest blitz engines, and Stockfish won here also. What more do I need the hail in the new King...

[ouachita]

What more do I need the hail in the new King...

We have agreed that's for you to decide.

[mwyoung]

What more do I need the hail in the new King...

We have agreed that's for you to decide.

I did not know I was the Magi for chess engines, the King-maker. I am honored on this Christmas day.

[ouachita]

Of course you are your engine Magi. Who else could be or should be?

[ernest]

Stockfish 231213 seems to take over command in all areas from blitz to LTC

Code: Select all

Stockfish DD 4x - Houdini 4, 4x Blitz 3m+2s
                            
1   Houdini 4 x64           +29/=47/-24 52.50%   52.5/100
2   Stockfish DD 64 SSE4.2  +24/=47/-29 47.50%   47.5/100

SF 231213 4x - H4 4x, Blitz 3m+2s
                        
1   Stockfish 231213 64  +25/=56/-19 53.00%   53.0/100
2   Houdini 4 x64        +19/=56/-25 47.00%   47.0/100

Seems, or perhaps, or maybe so...
The error bar (2SD) for those 100-game matches is around ­±7% !!!