100 long games Rybka 4 vs Houdini 1.5

[IWB]

Hallo all,

A few weeks ago I made a test of 100 long games (Game/60min) with Rybka 4 and Houdini 1.03a.

This is the Thread about it: http://www.talkchess.com/forum/viewtopi ... ka+houdini

I repeated the same test with Houdini 1.5. Unfortunately I made a small mistake and set the engine "just" to 256 MB while the original test was done with 1GB. As I recognized this too late I will stick with the result.

Conditions:

GUI: Shredder Classic
CPU: AMD Phenom II 6 core @ 3.275 GHz
Threads: 3(!) (aprox. 4 cores on a Q6600 @ 2.66 GHz)
Hash: 256MB each engine
Opening Positions: FQ 50 Position set
Colours: changing, every engine has to play White and Black, 100 games overall
Time control: 60 minutes per game/each side, played on 5 equal computers.
Ponder: ON (of course, I play real matches!)
Tablebases: 4 piece Nalimov and GTB
Ettiquette: ON (three consecutive 0.00 evals of both engines is a draw, 3 times in a row +/- 5.xx of the loosing engines is a loss!)

Result: + 32,= 51,- 17 in favor for Houdini 1.5

Download all games as PGN: http://www.mediafire.com/?57b4ogtvnjiiusy

Have fun
Ingo

[gaard]

Thanks for the games. Will play through a few tonight. I am particularly interested in seeing how Houdini fares in the king placement department, that is, by initiating and defending against king side attacks.

[George Tsavdaris]

Opening Positions: FQ 50 Position set
Colours: changing, every engine has to play White and Black, 100 games overall
Time control: 60 minutes per game/each side, played on 5 equal computers.
Ponder: ON (of course, I play real matches!)
Tablebases: 4 piece Nalimov and GTB
Ettiquette: ON (three consecutive 0.00 evals of both engines is a draw, 3 times in a row +/- 5.xx of the loosing engines is a loss!)

Result: + 32,= 51,- 17 in favor for Houdini 1.5

Thanks for the matches.

So we have(forgetting about the minor problem, meaningless i guess hash size difference):

Rybka 4 - Houdini 1.03 +23 =60 -17 or 53.0-47.0 or 53.0% for Rybka 4.
And then with the new Houdini 1.5 we have:
Rybka 4 - Houdini 1.5 +17 =51 -32 or 42.5-57.5 or 42.5% for Rybka 4.

What an improvement for Houdini!!

Here is what Bayeselo says:

Code: Select all

Rank    Name                   Elo      +    -    games  score  
 1    Houdini 1.5 3T            64     36    35    100    58%   
 2    Deep Rybka 4 3T           17     25    25    200    48%   
 3    Houdini 1.03a x64 3T       0     34    35    100    47%   

And here is what Elostat says:

Code: Select all

    Program                           Elo    +   -   Games   Score  
 1 Houdini 1.5 3T                 :    74    48  48   100     57.5 % 
 2 Deep Rybka 4 3T                :    21    32  32   200     47.8 % 
 3 Houdini 1.03a x64 3T           :     0    43  43   100     47.0 % 

They don't quite agree.

Opening Positions: FQ 50 Position set

What's this FQ thing?
Obviously i've seen which positions are( ) but what's the abbreviation means?

[Frank Quisinsky]

Hi George,

I wrote in the summer a large documentation about opening positions. Volker Annuss (programmer of Herrman) wrote later a little tool for an idea to this project.

PGN-Random Book 4.1:
http://www.amateurschach.de/master-pgn/ ... tation.htm

I work a long time on it and create here also a test set (only a part of this project) with popular opening systems.

- end position should be interesting
- with a full board of pieces (not many pieces should be switched).

And so on ...

You can find all steps to this project in the bigger download file on my webpages (download area). Also the tool Volker Annuss programmed.

The documentation is in German language.

Best
Frank

With other words:
Ingo wrote FQ positions. Ingo used this 50 opening positions from my summer project!

[beram]

Thx for testing, well done.
So again no diminishing strenght for Houdini 1.5 against DR4 at LTC as I observed too.

grts Bram