Houdini 1.5 at 60' Time Control (100 games)

[Tomcass]

After one week working non stop, here you have the results of this test. 100 games at 60 minutes per game. Houdini 1.5 against DR4 SSE4.2 and the IvanHoe compile (Peterpan) that leads the ranking of my tests:

Computer: i7 975 4 real cores + 4 Hyper Thread. (But HT: Off)
Fritz Benchmark: Speed 26,73 KNS 12.828
GUI: Fritz 12
Book: Masterbook 3.1
Time control: 60 minutes per game.
Ponder: off
No Robbo, Triple or Nalimov Bases

201012Houdini_15_60min 2010

Houdini 1.5 x64(x4) - Deep Rybka 4 SSE42 x64(x4) 31.0 - 19.0 +17/=28/-5 62.00%
Houdini 1.5 x64(x4) - Ivanhoe B50kBx64p(x4)BO 30.5 - 19.5 +15/=31/-4 61.00%

The superiority of Houdini 1.5 is huge at this Time Control. A big THANKS to Robert Houdart for this brilliant X'mas gift!.

Here you have the games:http://www.megaupload.com/?d=EFTIWP3S

Regards from Barcelona.

Tom.

[Frank Quisinsky]

Tom,

make the same test ...

Houdini 1.5 x64 - Rybka 4 x64 Exp. 61.
Settings can be found in Rybka forum or in my download area (my configuration files for the engines I used in SWCR).

More interesting as the default R4 settings!
Believe it or not

Thanks for your test!

Best
Frank

[gerold]

After one week working non stop, here you have the results of this test. 100 games at 60 minutes per game. Houdini 1.5 against DR4 SSE4.2 and the IvanHoe compile (Peterpan) that leads the ranking of my tests:

Computer: i7 975 4 real cores + 4 Hyper Thread. (But HT: Off)
Fritz Benchmark: Speed 26,73 KNS 12.828
GUI: Fritz 12
Book: Masterbook 3.1
Time control: 60 minutes per game.
Ponder: off
No Robbo, Triple or Nalimov Bases

201012Houdini_15_60min 2010

Houdini 1.5 x64(x4) - Deep Rybka 4 SSE42 x64(x4) 31.0 - 19.0 +17/=28/-5 62.00%
Houdini 1.5 x64(x4) - Ivanhoe B50kBx64p(x4)BO 30.5 - 19.5 +15/=31/-4 61.00%

The superiority of Houdini 1.5 is huge at this Time Control. A big THANKS to Robert Houdart for this brilliant X'mas gift!.

Here you have the games:http://www.megaupload.com/?d=EFTIWP3S

Regards from Barcelona.

Tom.

Thanks for your test results Tom.
At 5/3 TC - windows 32 after 2k games my results match your
results.

Best,
Gerold.

[Denzwell]

Thanks for your test results Tom.
At 5/3 TC - windows 32 after 2k games my results match your
results.

Best,
Gerold.

Hi Gerold,

Can you share please your pgn

Thanks

[gerold]

Thanks for your test results Tom.
At 5/3 TC - windows 32 after 2k games my results match your
results.

Best,
Gerold.

Hi Gerold,

Can you share please your pgn

Thanks

Sorry i test for a company it is private.

Best,
Gerold.

[Tomcass]

I will do some pre-test and then this match that you suggest at 60 minutes Time Control. If the strength of this experimental version is what it seems, it will be a very tough match!.

Regards,

Tom.

[Denzwell]

After one week working non stop, here you have the results of this test. 100 games at 60 minutes per game. Houdini 1.5 against DR4 SSE4.2 and the IvanHoe compile (Peterpan) that leads the ranking of my tests:

Computer: i7 975 4 real cores + 4 Hyper Thread. (But HT: Off)
Fritz Benchmark: Speed 26,73 KNS 12.828
GUI: Fritz 12
Book: Masterbook 3.1
Time control: 60 minutes per game.
Ponder: off
No Robbo, Triple or Nalimov Bases

201012Houdini_15_60min 2010

Houdini 1.5 x64(x4) - Deep Rybka 4 SSE42 x64(x4) 31.0 - 19.0 +17/=28/-5 62.00%
Houdini 1.5 x64(x4) - Ivanhoe B50kBx64p(x4)BO 30.5 - 19.5 +15/=31/-4 61.00%

The superiority of Houdini 1.5 is huge at this Time Control. A big THANKS to Robert Houdart for this brilliant X'mas gift!.

Here you have the games:http://www.megaupload.com/?d=EFTIWP3S

Regards from Barcelona.

Tom.

Hi Tom,

Is it possible to have your tournament pgn, not a html version

Thanks

[Tomcass]

Hi Tom,

Is it possible to have your tournament pgn, not a html version

Thanks

Hi, Denzwell.

To be honest, I have no idea about how to produce an output of these games in PGN format, from Fritz 12 GUI. Sorry!.

Tom.

... by the way, the 50 games match, suggested by Frank, between Houdini 1.5 and DR4 Exp 61 at 60 minutes/game, has already started.

[Frank Quisinsky]

Hi Tom,

but await not a wonder result.
We had this topic at the moment in CSS Forum. It seems not easy to understand the results for the most computer chess fans. Most are testing ... Rybka vs. Rybka or Rykpa vs. Houdini only.

Rybka 4 Exp. 61 and Rybka 4 Exp. 42 have great results vs. weaker engines. Vs. stronger engines the results are not better as the results from Rybka 4 default.

Look here ...

Code: Select all

3 Rybka 4 x64 Exp. 42       : 2948  900 (+586,=234,- 80), 78.1 %

Shredder 12                   :  26 (+ 16,=  8,-  2), 76.9 %
Zappa Mexico II x64           :  26 (+ 15,= 11,-  0), 78.8 %
Naum 4.2 x64                  :  26 (+ 12,= 11,-  3), 67.3 %
Loop 2007 x64                 :  26 (+ 21,=  3,-  2), 86.5 %
SmarThink 1.20 x64            :  25 (+ 21,=  2,-  2), 88.0 %
Thinker 5.4d Inert x64        :  26 (+ 14,=  8,-  4), 69.2 %
Hiarcs 13.1                   :  26 (+ 21,=  3,-  2), 86.5 %
Komodo 1.2 JA x64             :  26 (+ 12,=  8,-  6), 61.5 %
Fruit 09_07_05 x64            :  26 (+ 17,=  7,-  2), 78.8 %
Onno 1.2.70 x64               :  26 (+ 20,=  5,-  1), 86.5 %
Spark 0.5 x64                 :  25 (+ 15,=  8,-  2), 76.0 %
Hannibal 1.0a x64             :  26 (+ 18,=  8,-  0), 84.6 %
Critter 0.80 x64              :  26 (+ 16,=  9,-  1), 78.8 %
GullChess 1.0a x64            :  26 (+ 14,= 11,-  1), 75.0 %
Jonny 4.00                    :  26 (+ 25,=  1,-  0), 98.1 %
Protector 1.3.6-370 JA x64    :  26 (+ 20,=  5,-  1), 86.5 %
Junior 12.0 x64               :  26 (+ 18,=  4,-  4), 76.9 %
Gaviota 0.80 x64              :  26 (+ 25,=  0,-  1), 96.2 %
Equinox 0.87t x64             :  26 (+ 22,=  3,-  1), 90.4 %
Stockfish 1.9.1 JA x64        :  25 (+ 11,= 10,-  4), 64.0 %
Sjeng c't 2010                :  25 (+ 17,=  8,-  0), 84.0 %
Crafty 23.4 JA x64            :  26 (+ 17,=  8,-  1), 80.8 %
Zarkov 6.44                   :  26 (+ 26,=  0,-  0), 100.0 %
Tornado 4.25 x64              :  26 (+ 24,=  2,-  0), 96.2 %
Umko 1.1 x64                  :  40 (+ 35,=  5,-  0), 93.8 %
Critter 0.90 x64              :  40 (+ 16,= 16,-  8), 60.0 %
GullChess 1.1 x64             :  40 (+ 28,=  9,-  3), 81.2 %
IvanHoe B49jA x64             :  40 (+ 12,= 19,-  9), 53.8 %
Spark 1.0 x64                 :  40 (+ 24,= 14,-  2), 77.5 %
Booot 5.1.0                   :  40 (+ 30,=  8,-  2), 85.0 %
Houdini 1.5 x64               :  40 (+  4,= 20,- 16), 35.0 %

Look what Houdini 1.5 x64 made with Exp. 42.
But Exp. 42 is 20 ELO stronger after 900 games as the default setting.

Look what Stockfish 1.9 have for results vs. Exp. 61 in the still running test. Have a look on the results vs. weaker engines too (important).

This setting is very aggressive with pawns and lost some games very speaktakular too.

But all in all ... if you test this setting vs. the complete group of the TOP-25 you will see that Exp. 61 is a lot stronger as Rybka 4 default.

Best
Frank

[beram]

Hi Tom,

but await not a wonder result.
We had this topic at the moment in CSS Forum. It seems not easy to understand the results for the most computer chess fans. Most are testing ... Rybka vs. Rybka or Rykpa vs. Houdini only.

Rybka 4 Exp. 61 and Rybka 4 Exp. 42 have great results vs. weaker engines. Vs. stronger engines the results are not better as the results from Rybka 4 default.

....

Look what Houdini 1.5 x64 made with Exp. 42.
But Exp. 42 is 20 ELO stronger after 900 games as the default setting.

Look what Stockfish 1.9 have for results vs. Exp. 61 in the still running test. Have a look on the results vs. weaker engines too (important).

This setting is very aggressive with pawns and lost some games very speaktakular too.

But all in all ... if you test this setting vs. the complete group of the TOP-25 you will see that Exp. 61 is a lot stronger as Rybka 4 default.

Best
Frank

hi Frank,

I did a quick test 50 games H15-R4exp61 at TC 3m2s and 6m4s at my AMDx6 1090T (4cores for each) R4exp61 performs well against H15 at blitz, 3m2s (48%), but at 6m4s R4exp61 already performs same as R4 default (39% R4exp61; 39%R4(default)
So in line with observation for relatively worse performance against strong programs.
I agree that the aggressiveness of the exp61 setting (at longer TC) works against it in matches against equal opponents(1) and the longer the TCthe more outspoken this becomes (2)

In other words this spectacular R4xp61 setting is nice in matches against weak opponents, but to put it mildly, less useful for matches against strong opponents, game analysis and correspondence chess.
So in the end what does it mean:"..a lot stronger than R4 default"

Kind regards, Bram