Elo-Mania, 60 +15, 4cpu, ponder ON

[Hugo]

Hi all

since a few days I have started two enginematches with Houdini 3 Std
Opponents are Stockfish 2.3.1 and Rybka 4.1 .
Conditions:

engine matches 4cpu running on Intel(R) Xeon(R) CPU E5-2687W 0 @ 3.40GHz (16 cpu/ 32 threads)

timecontrol = 60min +15sec
all engines use 4 cpu, 64 bit and sse4 if possible,
ponder = ON, no hyperthreading
4096 MB hashtables
engines use no tablebases
DeepFritz 13 gui with CB-EndgameTurbo3 (Nalimov, 3-4-5 some 6)
Grandmasterbook limit 5 moves
same opening will be played with reversed colors
50 games per match

No live games. Crosstables and ReplayZone /PGN updated daily.

regards, Clemens keck



http://www.clemens-keck.de/base.htm

[Jouni]

Interestingly Rybka is giving really good challenge:

1 Houdini 3 x64 STD x4 ½01½½½½01½½11½1½½½10½1½01½½½½0½½110½½ 20.5 / 37

2 Deep Rybka 4.1 SSE42 x64 x4 ½10½½½½10½½00½0½½½01½0½10½½½½1½½001½½ 16.5 / 37

Rybka is "scaling" better ?!

[Uri Blass]

Interestingly Rybka is giving really good challenge:

1 Houdini 3 x64 STD x4 ½01½½½½01½½11½1½½½10½1½01½½½½0½½110½½ 20.5 / 37

2 Deep Rybka 4.1 SSE42 x64 x4 ½10½½½½10½½00½0½½½01½0½10½½½½1½½001½½ 16.5 / 37

Rybka is "scaling" better ?!

I think that you need to play houdini 1 cpu against Rybka4 cpu at different time control to check the theory that rybka scales better.

If houdini 1 cpu can beat Rybka4 cpu at 4+1 ponder on match but lose at 60+15 then it is going to be an evidence that rybka is scaling better.

Note that houdini 4 cpu had problems against Rybka in the CEGT 40/20(meaning less than 60% against the opponent only when the opponent was Rybka)
It happened both against Rybka4.1 4 cpu and surprisingly also against Rybka4 2 cpu when Houdini scoresd only minimal win against it +5 =9 -4.

http://www.husvankempen.de/nunn/40_40%2 ... ons/1.html

[beram]

Interestingly Rybka is giving really good challenge:

1 Houdini 3 x64 STD x4 ½01½½½½01½½11½1½½½10½1½01½½½½0½½110½½ 20.5 / 37

2 Deep Rybka 4.1 SSE42 x64 x4 ½10½½½½10½½00½0½½½01½0½10½½½½1½½001½½ 16.5 / 37

Rybka is "scaling" better ?!

I think that you need to play houdini 1 cpu against Rybka4 cpu at different time control to check the theory that rybka scales better.

If houdini 1 cpu can beat Rybka4 cpu at 4+1 ponder on match but lose at 60+15 then it is going to be an evidence that rybka is scaling better.

Note that houdini 4 cpu had problems against Rybka in the CEGT 40/20(meaning less than 60% against the opponent only when the opponent was Rybka)
It happened both against Rybka4.1 4 cpu and surprisingly also against Rybka4 2 cpu when Houdini scoresd only minimal win against it +5 =9 -4.

http://www.husvankempen.de/nunn/40_40%2 ... ons/1.html

Despite all the speculations halfway, the matches are finished with Houdini 3 winning percentage pretty much in line with CEGT 40/20, in fact even better.
58% score against Rybka 4.1 (CEGT was 56%) and 64 % against Stockfish 2.3. (CEGT also 64%)

grts Bram

Code: Select all

ELO-MANIA-1 : Houdini 3 STD x4  vs Stockfish 2.3.1 x4; Blitz 60m+15s
  
1	Houdini 3 x64 STD x4	 	         ½1½½1110½11½½01½½11½½½½½1½111011½½½½1½1½½½1½½½½½0½	32.0 / 50	+100 
2	Stockfish 2.3.1 JA 64bit SSE4	 	½0½½0001½00½½10½½00½½½½½0½000100½½½½0½0½½½0½½½½½1½	18.0 / 50

Code: Select all

ELO-MANIA-2 : Houdini 3 STD x4  vs Deep Rybka 4.1 x4; Blitz 60m+15s  

1	Houdini 3 x64 STD x4	       	 ½01½½½½01½½11½1½½½10½1½01½½½½0½½110½½½11½½½½1½½½1½	29.0 / 50	+56 
2	Deep Rybka 4.1 SSE42 x64 x4	 	½10½½½½10½½00½0½½½01½0½10½½½½1½½001½½½00½½½½0½½½0½	21.0 / 50