I found out some very interesting information about both teams. I talked at length with both teams during the past week and I like their enthusiasm, sportsmanship and passion for computer chess. Even though the 2 teams competed very hard against each other, they were very friendly and respected each other. The most interesting fact that I learned is both programmers were born in the United States.
Chess Daily News from Susan Polgar
The test results are already in, Zappa Mexico is muck weaker than Rybka.
http://www.computerchess.org.uk/ccrl/404.live/
Like I said before, these 10 game matches are fun to watch but they really don’t tell you much about engine strength.
sorry Pal, this is not tournament like. This is a two engine match!
Tournament between many engines is different than match between two engines.
CCRL gets their ratings based on a series of matches, not tournaments. The reason why Rybka performs better in those lists is because the hardware is weaker than what displayed in the match. Zappa uses a brute force type search where it won’t get into a great depth, but search many possibilities. Rybka is the opposite, it will cut off it’s search of inferior moves earlier, therefore reaching a greater depth. Since the number of possible moves multiply significantly for each ply, it will take Rybka longer to reach the next ply than it will Zappa, therefore closing the gap. Also, Anthony Cozzie specializes in parallel processing, so it can search a lot faster on monster hardware than Rybka, therefore highlighting what I had said above.
-revengeska
Ok. I agree. Take it this way. If you only have a single cpu or duo computer desktop, I think Rybka is a better program.
But when you got more money, and can afford 8 cpu machine running on a large memory, then zappa can give you a better choice.
All in all, I like the analysis provided by both engines since I got both of them at home.
Also dont forget, the opening book preparation played a great role here, no matter how great your engine is.
And to add to my above comment.
It shows that we, human is still holding an edge over the computer.
All of the opening book preparation is done by human. The computer just do what the programmer tells it to do in a very very fast and quick search way to find the best move.
Go to the link above and see that Rybka with 2 CPUs TRASHES Zappa Mexico with 4! CPUs. Then click on Zappa Mexico and see the TERRIBLE record it has against Rybka. In 50 games played Zappa’s record against Rybka is +5-23=22, that 32%, sorry Pal, but Zappa is nowhere near as strong as Rybka. I wish it was, I wouldn’t mind owning a stronger engine.
Revengeska, your comment contradicts itself. No engine uses a brute force search anyway, but even if it was like you think, Rybka would achieve the next ply of course faster than a b.f.searcher, not slower. Pruning is being done to gain speed.
CCRL uses hardware up to 4 cpu cores.
The point is, 10 games are just meaningless in terms of statistical reliability. This is so simple to understand that I’m beginning to suspect a troll invasion 🙂
Buying Zappa Mexico is an explicit validation of the aphorism: “A fool and his money are soon parted”.
Well,Zappa won the match.
Narrowly ,but won it anyway.
Topalov used to brag about his 60 ELO points advantage over Kramnik prior to their match.
However he lost the match.
So what score would you expect in match Rybka-Zappa played on latest
8CPU hardware if 100 games are played?
I would expect 50-50 ,something like that.
Not by any chance, on latest
8CPU configurations, Rybka trashes Zappa.