Houdini 2.0b Pro x64 First Test

[mwyoung]

I7 Q 840
5 man TB
1024 HT
TC 5min 3inc
Generic book to 12 moves.

34 of 60 games played.

Houdini 2.0b Pro x64 - Critter 1.2 64-bit SSE4 ...7.5 - 4.5 ...+4/=7/-1 ...62.50%
Houdini 2.0b Pro x64 - Deep Rybka 4.1 SSE42 x64 ...8.0 - 3.0 ...+5/=6/-0 ...72.73%
Houdini 2.0b Pro x64 - Stockfish 2.1.1 JA 64bit ...7.5 - 3.5 ...+6/=3/-2 ...68.18%

[Dr.Wael Deeb]

Impressive results I have to admit
Dr.D

[mwyoung]

I7 Q 840
5 man TB
1024 HT
TC 5min 3inc
Generic book to 12 moves.

34 of 60 games played.

Houdini 2.0b Pro x64 - Critter 1.2 64-bit SSE4 ...7.5 - 4.5 ...+4/=7/-1 ...62.50%
Houdini 2.0b Pro x64 - Deep Rybka 4.1 SSE42 x64 ...8.0 - 3.0 ...+5/=6/-0 ...72.73%
Houdini 2.0b Pro x64 - Stockfish 2.1.1 JA 64bit ...7.5 - 3.5 ...+6/=3/-2 ...68.18%

60 of 60 games played.

Houdini 2.0b Pro x64 - Critter 1.2 64-bit SSE4 ...12.0 - 8.0 ...+5/=14/-1 ...60.00%
Houdini 2.0b Pro x64 - Deep Rybka 4.1 SSE42 x64 ...14.5 - 5.5 ...+9/=11/-0 ...72.50%
Houdini 2.0b Pro x64 - Stockfish 2.1.1 JA 64bit ...13.0 - 7.0 ...+10/=6/-4 ...65.00%




First Rating of Houdini 2.0b

1 Houdini 2.0b Pro x64 3381
2 Houdini 2.0 x64 3352
3 Houdini 2.0 Pro x64 3345
4 Houdini 1.5a x64 3322

[ernest]

First Rating of Houdini 2.0b

OMG!!!

3381 +-60 !!!

[Anton]

Surely a thread about a Robbolito clone belongs in the cloning forum.

btw, would you be interested in a Monet painting I have for sale? for you, just $10,000 if bought before Oct 14, ignore the "printed by Canon" watermark, that is just a hallucination, it is a real genuine original work by Monet.

[Houdini]

First Rating of Houdini 2.0b

OMG!!!

3381 +-60 !!!

Correct, it is important to stress the very large confidence interval.
As Houdini 2.0, 2.0 Pro and 2.0b are engines with identical strength, the best Elo estimate is obtained by combining the 3 results given above

1 Houdini 2.0b Pro x64 3381
2 Houdini 2.0 x64 3352
3 Houdini 2.0 Pro x64 3345

into a single rating, with the combined number of games.

Robert

[Albert Silver]

Impressive results I have to admit
Dr.D

34 games played.... out of a whopping 60.

[mwyoung]

First Rating of Houdini 2.0b

OMG!!!

3381 +-60 !!!

Correct, it is important to stress the very large confidence interval.
As Houdini 2.0, 2.0 Pro and 2.0b are engines with identical strength, the best Elo estimate is obtained by combining the 3 results given above

1 Houdini 2.0b Pro x64 3381
2 Houdini 2.0 x64 3352
3 Houdini 2.0 Pro x64 3345

into a single rating, with the combined number of games.

Robert

Houdini 2.0b is now being tested at a longer TC in the 2nd group and testing. Houdini 2.0b is still showing better results then 2.0 by a good margin. More games coming.


TC 40moves/15mins

29 of 100 games played

Houdini 2.0b Pro x64 - Critter 1.2 64-bit SSE4 4.5 - 1.5 +3/=3/-0 75.00%
Houdini 2.0b Pro x64 - Deep Rybka 4.1 SSE42 x64 4.0 - 2.0 +3/=2/-1 66.67%
Houdini 2.0b Pro x64 - Stockfish 2.1.1 JA 64bit 4.0 - 2.0 +3/=2/-1 66.67%
Houdini 2.0b Pro x64 - Komodo64 3 4.0 - 2.0 +2/=4/-0 66.67%
Houdini 2.0b Pro x64 - Fire 2.2 xTreme x64 4.0 - 1.0 +3/=2/-0 80.00%


I do know you said the two version are the same strenght, but they are not the same program it seems. As I get different moves and evals and solve times when running the 2 versions side by side in Fritz, no matter the hash lever or CPU cores used.

The ratings gap will come together beween the 2 version if they are the same strength as I log 100's of games.

Robert it is not that I doubt you on your own program, but I always test new versions no matter what the programmer says...most time the programmer is correct, sometime not for what ever reason.

Most times I keep my results to myself, but since no one else was going to test the new version, and some members ask for a test. I made this version test public.