I will add the video of the press conference, the exclusive interviews with team Zappa and Rybka. There are a lot of fascinating stuff. It will also be archived on www.ChessDiscussion.com soon.
Chess Daily News from Susan Polgar
Zappa – Rybka Press Conference
I thought Rybka was the best. So disappointed now, will never use Rybka again.
Same here. Turns out it is a bad engine.
This is bad news for future Rybka sales. But if Rybka is still tops over Zappa Mexico on the engine rating sites then this win for Zappa will mean little.
Where can I find the games from the Rybka v Zappa match?
“There ARE a lot of …”: no need to complain about the quality of the USCF website.
I am assuming the first two anon posts here are from very young children.
I have played 40 or 50 games against Zappa Mexico with Rybka in the last few days, and my winning percentage is about 60%.
I use a watercooled quad core machine to play engine games, and have yet to have serious problems against Zappa.
Zappa’s advantage in Mexico came from the fact that it scaled well over an 8 core machine. On a Quad core Rybka has a slight edge, on anything less, Rybka is significantly better.
To the average user of a chess engine, Rybka is still the strongest. As can be seen on all the engine test sites.
Phil,
How do you know which program scales better on multicore machines ? Do you have any evidence supporting your claims ? Is it possible that Zappa had a much better opening book then Rybka so that Zappa brought Rybka to positions where Zappa plays well and Rybka does not ?
You don’t seem to understand the difference between an engine and a chess playing system which consists of the hardware, opening book, engine and the endgame tablebases.
I think my understanding is reasonable, I built my own computer for engine chess, I have played over 20,000 games at playchess.com, and my computers rating there is about 2800 with a highest of over 2900.
Of course opening books are a hugely significant factor, my point was that the people who posted the first two replies to this entry were suggesting that Rybka was somehow a poor engine because of the Mexico result. This is patently not the case.
The scaling issue has been well covered in many places on the net, evidence of Zappa making better use of higher core counts is not hard to find. That is not a criticism of Zappa, I am simply saying that unless you have an 8 core server to play on, Rybka is still the stronger engine. Also that the advantage of Zappa over Rybka reduces with the number of cores it plays on.
If you want evidence, have a look at this site
http://www.husvankempen.de/nunn/40_40%20Rating%20List/40_40%20BestVersion/rangliste.html
There you will see that Zappa Mexico running on two cores has an ELO of 2934. Rybka 2.3.2a on two cores has an ELO of 3055.
These are the commercially available versions of the programs, so although both Rybka and Zappa in Mexico had modifications, the results shown at CEGT are more relevant to someone considering which engine to buy.
Here Phil, take a look:
http://www.computerchess.org.uk/ccrl/404.live/
Thank you, that site illustrates my point well.
Rybka on 2 cores is still slightly ahead of Zappa on 4 cores.
For anyone deciding which engine to buy, thats a useful fact to know.
I agree with Phil, having tried to explain the same in recent comments. Nevertheless, at the same time I think we can say that we have seen the TWO currently strongest engines, at least on 2 cpu cores or more. Not to be missed is the upcoming release of Shredder 11 which is already reported to have been very much improved. I am mentioning that because I think some users, i.e. correspondence players, will not rely on just one presumably strongest engine but will want to have a “racing stable” with a couple of strong but different engines for their analyses.
(On single cpu, some others are candidates for rank #2 too, including the freewares Fruit and Toga.)
Phil,
You said “If you want evidence, have a look at this site
http://www.husvankempen.de/nunn/40_40%20Rating%20List/40_40%20BestVersion/rangliste.html
There you will see that Zappa Mexico running on two cores has an ELO of 2934. Rybka 2.3.2a on two cores has an ELO of 3055.”. Unfortunately your reasoning is not scientific. If the programs playing at Mexico would be exactly the same as (opening book, tablebases etc) programs used for making the rating list you have mentioned then your claim could make sense (Rybka running on two processors is stronger running on 8 cores weaker, therefore it does not scale as good). But that is not the case. Opening books were different at Mexico, so you don’t have any evidence.
If someone was to buy Zappa Mexico, for the purposes of playing in an engine room, and he ran it on a computer with 4 or less processors, he would find that Rybka would be stronger in most cases.
I have seen quite a few people do exactly this in the playchess engine room, and yes, their rating has dropped.
Thats my point. As supported by the results shown at the sites posted above.
Clearly the opening book preparation is a big factor here.
The computer which ever engine you use, just calculate the moves based on the value point assign to its pieces on the board.
I assume when they tuned the engine, the programmer actually gives a different sampling value to the program, so it will be different than the engine that you bought out of the shell.
All of this factor, plus the zappa probably have the best way to run on multi machines, makes zappa won the tournament this time.
But however for people who owns Rybka, dont worry. There is nothing wrong with your engine.
I just hope that one day, either zappa or rybka team would like to sell us their so called “opening preparation book”. Or maybe somebody can tell us how to tune the value system in an engine, and make it configurable somehow.
So we can play around with it, and try to maximize it.
Ok. Enough said. Hope this will help those who are sad because they felt their engine are suck.
Forget it, a tuned match book is
useless for a normal user! Or is the entire purpose you buy a chess progam for, to play computer games against 8 core Zappa or against 8 core Rybka??
What do you do with your chess program? Think about where it would help you to have a book which has some “uncomfortable” variations and probabilities against a specific engine. What for?
I thought chess players are thinking human beings. The comments here are so full of bullshit that I don’t know anymore at whom this blog is directed.
It’s obvious you either don’t understand or don’t like engine chess, thats fine, it’s certainly not for everyone.
I bought my program and built my computer to play against other computers, that may seem weird to you, but it’s my choice as to how I spend my spare time.
There is really no point or no need to be offensive about it.