Tony Thomas wrote:Looks like Erdo probably did a wonderful job with the books. It is possible that Anthony used some tricks to get perfect scaling on 8 cores. After all with double the speed (considering that rybka doesnt have perfect scaling) Zappa is only a few elo points weaker than Rybka.
Everything I said above should be considered as the blabberings of an insane Zappa fan , person . So the score is Zappa 2.5- Rybka 1.5. Wow, not an expected result. I am about to go order my Zappa.
Nothing wrong with ordering Zappa Mexico. I already have it. But if you think the version you are buying is the same one being used in this match, have some vaseline when you bend over
I do not know if it is the same but zappa Mexico could also beat rybka2.3.2a in the CCRL
Of course Zappa used 4 processors when rybka used only 1 processor but usually one processor of rybka2.3.2a 64 bits beat 4 processors of other programs.
Uri Blass wrote:
I doubt if the problem was positive contempt factor.
It seems that the problem was not having negative contempt factor
that was needed.
Rybka evaluated the postion as more than +2 for herself during the game
(I do not mean to the position that was posed but to earlier positions).
In this case it is probably better to evaluate draw as +1 because if you are winning the game you will never go down below +1 and if you go down below +1 but your evaluation is still positive then it means that you probably do not understand the position and are too optimistic so it is better to prefer draw.
Uri
Hi Uri,
Do you know which version or Rybka they were using?
Thanks in advance.
I read that they were using 2.3.2a in the first game
version with improved evaluation in the second game and the 4th game
and
version with improved evaluation and improved search in the third game
They were afraid from a bug after losing the third game so they did not use the version with improved search in the 4th game.
Total blow by Rybka in this game. weird. maybe they should just play with the retail version, avoiding all those obsecure new versions that play weird moves(bugs).
Oh and BTW .. please no more Ruy Lopez.. its growing ugly.
To be honest, I can't understand how they have permission to switch the version during the match.
They should start the match with 1 version and play it through the whole match (like Zappa seems to be doing).
Total blow by Rybka in this game. weird. maybe they should just play with the retail version, avoiding all those obsecure new versions that play weird moves(bugs).
Oh and BTW .. please no more Ruy Lopez.. its growing ugly.
To be honest, I can't understand how they have permission to switch the version during the match.
They should start the match with 1 version and play it through the whole match (like Zappa seems to be doing).
In an ideal world I assume you would be correct.
However in ideal world the machines would arrive on time and there wouldn't be need to play with remote machines.
Or having all the machines in WCCC being equal.. aint gonna happen.
I think its fair. Those changes are so little, probably little eval tunings I bet, done very quickly, often to iron out bugs or such. not huge code rewrites that would really give you different engine.
I see no reasons not doing it. better versions there gives us better retail versions later. those weaks are welcomed from my POV.(ofcourse, my original point was - its better to play with a 'safe' version, and not a new untested one, to avoid sloppy moves and possible bugs).
Total blow by Rybka in this game. weird. maybe they should just play with the retail version, avoiding all those obsecure new versions that play weird moves(bugs).
Oh and BTW .. please no more Ruy Lopez.. its growing ugly.
To be honest, I can't understand how they have permission to switch the version during the match.
They should start the match with 1 version and play it through the whole match (like Zappa seems to be doing).
Yes Martin, that bugs me also. While the match is fun to watch, it's hard for me to take this as a serious match under these conditions. I guess the Rybka team does not feel this is a serious match either. Alot has went wrong right from the start. Starting with the ordered Computers. At least we are getting a good show. If the Zappa team wins. They can say that they beat the best the Rybka team had to offer.
"Good decisions come from experience, and experience comes from bad decisions."
__________________________________________________________________
Ted Summers
AdminX wrote:
Yes Martin, that bugs me also. While the match is fun to watch, it's hard for me to take this as a serious match under these conditions. I guess the Rybka team does not feel this is a serious match either. Alot has went wrong right from the start. Starting with the ordered Computers. At least we are getting a good show. If the Zappa team wins. They can say that they beat the best the Rybka team had to offer.
This match has turned out quite differently than I had originally expected.
That I'm a Zappa fan isn't exactly a secret, it is good for the computer chess that the #2 on the lists is strong enough to make Team Rybka sweat.
I don't think Zappa will win the match, but at least the excitement is there.
I wish Anthony would devote himself 100% to his engine like Vasik has, then maybe Zappa could even claim the #1 spot one day.
Total blow by Rybka in this game. weird. maybe they should just play with the retail version, avoiding all those obsecure new versions that play weird moves(bugs).
Oh and BTW .. please no more Ruy Lopez.. its growing ugly.
To be honest, I can't understand how they have permission to switch the version during the match.
They should start the match with 1 version and play it through the whole match (like Zappa seems to be doing).
When humans play chess in a tournament or a match they are allowed to learn and improve their level based on experience of previous games or based on analysis of games.
It is not only improving book but also better chess understanding.
I see no reason why do you think not to allow programs to do the same.
Uri Blass wrote:
When humans play chess in a tournament or a match they are allowed to learn and improve their level based on experience of previous games or based on analysis of games.
Because it is eventually a human who decide to put differently tuned or otherwise different version to play in the next game. Your analogy would be valid if a program itself would learn from a game and tuned some weights in the eval for example.
In my opinion a program should stay the same thorough the match.
Uri Blass wrote:
I doubt if the problem was positive contempt factor.
It seems that the problem was not having negative contempt factor
that was needed.
Rybka evaluated the postion as more than +2 for herself during the game
(I do not mean to the position that was posed but to earlier positions).
In this case it is probably better to evaluate draw as +1 because if you are winning the game you will never go down below +1 and if you go down below +1 but your evaluation is still positive then it means that you probably do not understand the position and are too optimistic so it is better to prefer draw.
Uri
How do you know this? I only see games posted/published with _Zappas
evaluation only_? Zappa evaluated the position as that good for Rybka,
despite being a dead draw.
So far I have not seen any game file showing Rybkas evaluation...
or do you mean your own Rybka version at home evaluates the positions
in question that way? In this case you cannot be sure, as you do not have
the version playing in Mexico.
. So the score is Zappa 2.5- Rybka 1.5. Wow, not an expected result. I am about to go order my Zappa.
