Page 5 of 11

Re: Rybka 4 vs Houdini 1.02 - Long TC Match - LIVE BROADCAST

Posted: Thu Jul 01, 2010 4:29 am
by Martin Thoresen
Hehe Tom, I have little control over this match while I am sleeping. :D

Re: Rybka 4 vs Houdini 1.02 - Long TC Match - LIVE BROADCAST

Posted: Thu Jul 01, 2010 1:59 pm
by S.Taylor
My internet was down for 6 days. Now i see that Rybka has a big lead.
If it's going like this till now, i don't expect Rybka will suddenly have problems.
To crush Rybka 4, it seems we must wait for a different program, or version.

Re: Rybka 4 vs Houdini 1.02 - Long TC Match - LIVE BROADCAST

Posted: Thu Jul 01, 2010 2:47 pm
by gerold
S.Taylor wrote:My internet was down for 6 days. Now i see that Rybka has a big lead.
If it's going like this till now, i don't expect Rybka will suddenly have problems.
To crush Rybka 4, it seems we must wait for a different program, or version.
This match is closer than a person may think. Look at all the
draw games.
Rybka use TB also . I don't think Houdini use TB.

Best,
Gerold.

Re: Rybka 4 vs Houdini 1.02 - Long TC Match - LIVE BROADCAST

Posted: Thu Jul 01, 2010 5:26 pm
by Albert Silver
gerold wrote:
S.Taylor wrote:My internet was down for 6 days. Now i see that Rybka has a big lead.
If it's going like this till now, i don't expect Rybka will suddenly have problems.
To crush Rybka 4, it seems we must wait for a different program, or version.
This match is closer than a person may think. Look at all the
draw games.
Rybka use TB also . I don't think Houdini use TB.

Best,
Gerold.
I guess crushing depends on what one's expectations are. So far, after 28 games, Rybka is performing at +50 Elo.

the next match

Posted: Thu Jul 01, 2010 5:28 pm
by Martin Thoresen
For the next match I was thinking Houdini vs Stockfish.

BUT

A few choices I have which I can't decide on:

1: Keeping the same format
2: Using 2CPU and Ponder ON, with the same time controls
3: Using 2CPU and Ponder ON, with double the time controls

What do you think?

Also, I will double the hash size to 2048 MB.

Re: the next match

Posted: Thu Jul 01, 2010 5:49 pm
by Martin Thoresen
I got a few requests but forgot to add this:

Games 1-24 in a proper PGN format. Download here.

Re: the next match

Posted: Thu Jul 01, 2010 6:02 pm
by Albert Silver
Martin Thoresen wrote:For the next match I was thinking Houdini vs Stockfish.

BUT

A few choices I have which I can't decide on:

1: Keeping the same format
2: Using 2CPU and Ponder ON, with the same time controls
3: Using 2CPU and Ponder ON, with double the time controls

What do you think?

Also, I will double the hash size to 2048 MB.
If Stockfish is capable of gaining more from an i7 due to its use of threads, as opposed to cores, I'd like to see them with ponder off as here, to get a feel as to how much it really gains from the hyperthreading.

As to hash, give them as much as is practically possible with such a system. There is no question that both Rybka and Houdini are filling the 1GB within 30 seconds to a minute at most.

Re: the next match

Posted: Thu Jul 01, 2010 7:46 pm
by Martin Thoresen
Albert Silver wrote: If Stockfish is capable of gaining more from an i7 due to its use of threads, as opposed to cores, I'd like to see them with ponder off as here, to get a feel as to how much it really gains from the hyperthreading.
Albert, I am not sure I understand you here. Hyperthreading is not enabled in my system and will not be. Feel free to elaborate a bit on your point. :D

Confused Regards,
Martin

Re: Rybka 4 vs Houdini 1.02 - Long TC Match - LIVE BROADCAST

Posted: Thu Jul 01, 2010 7:48 pm
by S.Taylor
Albert Silver wrote:
gerold wrote:
S.Taylor wrote:My internet was down for 6 days. Now i see that Rybka has a big lead.
If it's going like this till now, i don't expect Rybka will suddenly have problems.
To crush Rybka 4, it seems we must wait for a different program, or version.
This match is closer than a person may think. Look at all the
draw games.
Rybka use TB also . I don't think Houdini use TB.

Best,
Gerold.
I guess crushing depends on what one's expectations are. So far, after 28 games, Rybka is performing at +50 Elo.
You mean, between the two. But if they both played weaker programs, I wonder if Houdini would prove more dangerous vs them.

(crushing? first comes merely to win more than losing vs Rybka 4. Crushing would mean more than that, and this is obviously not available at the momment)

Re: the next match

Posted: Thu Jul 01, 2010 7:52 pm
by S.Taylor
Martin Thoresen wrote:For the next match I was thinking Houdini vs Stockfish.
Obviously, you see what you see, and that makes you curious how they would do vs each other.
Ultimately, it would be interesting to see how soon something proves superior to Rybka.