MEGA MATCH VERY LONG TIME CONTROL: Houdini 1.5 vs Rybka 4

[aLfApRiMo]

100 games played (alternating colors)


Start Date Match: 2010.12.16
End Date Match: 2011.01.20


TIME CONTROL:
2 hours each for the first 40 moves +
2 hours each for the next 40 moves +
2 hours each for finish the game


HASH: 256 MByte
Book: OFF
Ponder: OFF
Tablebase: OFF
Threads: 2 for both engines

Both engines are restarted by GUI after each game ended.

Hardware: Intel(R) Xeon(R) CPU E5240 @ 3.00 GHz Dual with 1.024 MB memory

OS: Microsoft Windows XP Professional Service Pack 3 (Build 2600) 32 bit

GUI: Arena 2.0.1 Build 2399

PGN-File: Houdini 1.5 vs Deep Rybka 4 VLTC.pgn

Country: Italy

Website: www.giocareascacchi.it

Nick name: aLfApRiMo

E-Mail Address: catello.erbi@tin.it



*FINAL RESULT
-------------------------------------------------------------------------------------------------------
*Houdini 1.5_w32 - Deep Rybka 4 w32:
*63,5 - 36.5
*40+/47=/13-
*(=0==11==1=1=011=11101=1=111=0=1=0==1==1===0=====00111=1=1=1=1=1=0=1=1=10=011=11=1=1=1=1=1=====1011=0)
*63.5% => ELO +96
-------------------------------------------------------------------------------------------------------



My personal opinion:
------------------------------------------------------------------------
In very long time control, Houdini 1.5 is much stronger than Rybka 4



Next VERY LONG TIME CONTROL MATCH will be:
------------------------------------------
Houdini 1.5 versus Robbolito 009

with 1 thread for both (Robb isn't multithreading),
64 bit edition for both (on another 64 bit machine).

The other options are identical to the present match.

For Robert Houdart:
I sent you at 'houdini@cruxis.be' the PGN file of this tournament.

[Kurt Utzinger]

I cant find the PGN file on your homepage. Please give a direct download link. Many thanks.
Kurt

[aLfApRiMo]

Direct download link for Houdini 1.5 vs Deep Rybka 4 VLTC.pgn:
http://www.zshare.net/download/85540570009ecd3c/

[Frank Quisinsky]

Hi Catello,

you need 33 days for such an match.

Easy:
With this time controls move average from normaly 86 = 8 hours per game x 100 = 800 hours : 24 = 33 days.

You must played without ponder and with resign mode = on. Think so! Houdini 1.5 is now 35 days available

At the first moment I am thinking not possible on one machine only, but its possible

Perhaps you have interest to play the same match again but vs. Rybka 4 Exp. 42. I am sure you will get an other result. Houdini 1.5 will win too but I think Rybka will make much points more because Rybka is much more aggressive with this setting.

And ...
With x64 the different will be smaler!

I visiting different Houdini 1.5 games with such long time controls in SWCR Champions-League 2011. Often to see that Houdini find with more time better moves.

But I am not sure that we can say ... Houdini is with more times stronger. I wrote for minutes a long message to Houdini in my German actual page. Reason are the 7.5 : 0.5 result vs. Deep Fritz 12 in SWCR Champions-League.

It's an interesting topic!
I mean to try to find out why programs like Houdini 1.5 x64 are so strong. What make Houdini better as the others. Very interesting question!

Have thanks for your work.
I will donwload the games and will have a look too.

Best
Frank

[Frank Quisinsky]

Hi Catello,

its an interesting download file.
The statisic program will not work under my Windows XP Professional x64 edition, unfortunately. Not search for the reason so far.

I have a look shortly in your files. You are playing without opening books. You have a lot of same first moves.

You are using Arena GUI.
Download from my webpage the bigger opening file "swcr-fq-openings-v4.1.zip" We developed in Arena (in the time I am working on this project) the PGN Random option.

You can used version 3.5 with 5.120 start positions with the Arena PGN-Random options. If you don't like many opening moves you can say "used the first 4 moves from a PGN file as PGN - Random".

This produced more different openings as to test such things without book. Only Arena have this options, very nice and ideal for testing.

Houdini 1.5 won 13 games up to move 50.
Rybka 4 won 0 games up to move 50.

Also nice to see, because I have in SWCR same statistics. King safty in middlegame isn't very good in Rybka 4. Houdini is much more aggressiv. But I am sure you will never get such a short game statistic if Rybka 4 Exp. 42 will play such a match vs. Houdini 1.5.

Without endgame database, thats good and interesting.
I will look tomorrow in detail in the games, very very interesting!

More interesting if you used an opening book.
Try out Arena PGN Random option perhaps with my PGN Random database. Version 3.5 is right ... you will have more fun if you see more different openings and the file is checked of mistakes, can you see if you download the bigger file. Furthermore checked in 35.000 SWCR games.

Thanks again!

Best
Frank

PS:
2 Cores without ponder isn't more interesting as 1 core and ponder = on.
Ponder = around 35-40 ELO. With two cores and factor 1.7-1.8 you will get around + 35-40 ELO too. But with ponder = on is much more interesting to visit the games. Time management is often better with ponder = on.