Not logged inRybka Chess Community Forum
Up Topic Rybka Support & Discussion / Rybka Discussion / Rybka vs Zappa in Mexico City
1 2 Previous Next  
- - By Vasik Rajlich (Silver) Date 2007-09-28 01:29
I´d like to take a moment to congratulate the Zappa team for winning their match against us here in Mexico City.

I'd also like to thank the members of the Rybka team for their really hard work. Our active team for this match was:

Opening book: Jeroen Noomen, Dagh Nielsen
Operation: Dagh Nielsen
Engine: Vasik Rajlich, Larry Kaufman
Hardware & Testing: Lukas Cimiotti

A massive amount of work went into this match. In particular, the level of knowledge and effort which goes into the opening book preparation is astounding. I could not be more proud of the Rybka team. This one didn´t go our way, but we will be back.

Vas
Parent - By FWCC (***) Date 2007-09-28 01:38
Congratulations too Vasik to your team for an exciting match.A match result this close proves nothing.Rybka is still king(or shall I say Queen) in my book(as all test results conclude)This was an astonishing match and thank you to both teams for raising the level of computer chess and exploring it's boundaries.I'm sure both programs will leave others behind as they are further developed and as a Terminator friend said"YOU'LL BE BACK!"
Parent - - By pawformation (*) Date 2007-09-28 01:41
Ah, but here is the question: Did you take away anything from the match that you can use to improve Rybka?

Thanks,
Parent - - By lkaufman (*****) Date 2007-09-28 02:06
Yes.
Parent - By revengeska (**) Date 2007-09-28 08:18
Short and sweet:)
Parent - By turbojuice1122 (Gold) Date 2007-09-28 02:20
Yes, congratulations on a great match; the overall level of play by both engines was astounding (in spite of a few gaffes).  I wish you good luck in implementing what you've learned in this match in creating an even stronger Rybka.  It should really help now that Rybka has an opponent who is roughly the same strength and whose name isn't Rybka.
Parent - By Knight (**) Date 2007-09-28 02:41
Congratulations to all. The best thing is that the new rybka will/should never makes the same blunders in this match. I expend the new zappa to do the same.
Parent - - By Mark Young (*) Date 2007-09-28 02:44
I look forward to Rybka 3. Rybka, dispite the match loss is showing it is still the strongest program on the market. I have Zappa Mexico, and Rybka wins consistantly by 70 % on my dual core.  You have made computer chess interesting for me again, keep up the good work.
                              

                                                                                  Rybka Fan,
                                                                                  Mark Young
Parent - By cma6 (****) Date 2007-09-30 23:30
Mark,
  At what rate of play are you getting that 70% result for Rybka over Zappa? And if they were slow games, e.g., 3 minutes/move, what result would you expect?
                                                    CMA
Parent - - By SillyFunction (**) Date 2007-09-28 03:08 Edited 2007-09-28 03:10
We all know that Rybka and Zappa are software, not hardware.
And the Mexico match is all about hardware, not software.
So it is the matter of lucky hardware, not talented software.

And congrat to Zappa team.
You make engine world not too bored
(for not let Rybka lonely champion)
Parent - - By lkaufman (*****) Date 2007-09-28 03:40
We can´t blame the hardware (they were equal) but we can say that the software that was intended for this hardware was not ready until game 7. This was partly due to the failure of the ordered computers to get thru Mexican customs, which I guess cost us a couple of days in getting ready, as there are various problems in relying on a remote internet-accessed computer. Also, although our opening book was first rate and Dagh Nielsen did a great job of adapting it day by day, it seems that Rybka may not play the Spanish opening for either side as well as some other openings (or Zappa plays it relatively better than other openings), and the Spanish was the main battleground of the match.
Parent - - By turbojuice1122 (Gold) Date 2007-09-28 04:32
More precisely, since Rybka had big advantages every time with white in the Spanish (as did Zappa), couldn't we say that these main-line Spanish's have a decent chance of actually being won for white?
Parent - - By Eelco de Groot (***) Date 2007-09-28 15:07 Edited 2007-09-28 15:10
Hello all,

Thanks to both teams for a magnificent match! Congratulations to Zappa for winning this first match! I just hope we can see more of this kind of matches in the future, and I think the World Championship and Rybka - Zappa complemented each other well as a spectator sport. Thank you!

I would just like to comment a little on Turbojuice's and Larry Kaufman's statement here, on the choice of the Spanish, I'm just a lowly clubplayer but I think the situation may not be so complicated:

a) Spanish won for White? That is not very likely but could it not just be it is one of those openings where White has a very longlasting plus, and this translates to good positions for White in the middlegame? The downside then is that there are a lot of forced variations -where yes, maybe Zappa on fast hardware has a little plus-, so risk of opening traps and not very much "play" for the engines. And sometimes these variations are probably too deep to calculate "real time". But a forced win? I don't think so.

b) Since the Spanish and other open games are basically still a lot about attacks on f7, King attacks in general it might already make a lot of difference if in these openings Vas augmented Rybka's King Safety eval a little. I mean, this is very much an open door to state this but it might actually be true? Maybe this was already done in the latest Rybka versions used in this match, I would not know. And finally then you would need a little luck as well, and maybe a bit more depth when playing on 8 cores, depth is always good :)...

Point c) would probably be that the Spanish is a very "strategic" opening, this probably has to do with its long term goals, and computers in general are just not yet as good as humans in this respect. But for this aspect the present match was a very fascinating experiment and experience :)

Okay, here ends "Oom Jan leert zijn neefje schaken" (famous book by our Max Euwe) :)

Regards, Eelco
Parent - By turbojuice1122 (Gold) Date 2007-09-29 01:53
Let me reformulate your second question in point (a) into "...and this translates to good positions for White in the endgame?" since the main line Spanish that I'm talking about involves book lines that go very deep into the middlegame, and then pass the question on to Larry :-) .  I'd certainly bet that the Spanish in general isn't won for white--that's why I singled out the particular main-line variations seen in this match.  I guess the question that really needs asking is, "What is the probability that C97 and C99 are won for white?"

As for changing evals based on the opening, this seems potentially very dangerous in the long-term in that "anti" variations in the opening could make the engine play worse than it otherwise would.
Parent - - By revengeska (**) Date 2007-09-28 08:23
"it seems that Rybka may not play the Spanish opening for either side as well as some other openings"

I've known about that for the longest time!  Definately not an opening I would continue to play with Rybka in your shoes.  I'm surprised, I thought Dagh and Vas would've picked up on that in freestyle:/
Parent - - By lkaufman (*****) Date 2007-09-28 21:23
No one mentioned this until quite late in the match. The problem is that as White, our book was much more complete for 1e4, and especially the Spanish. It´s a dilemma between playing ¨best¨openings and those that suit Rybka, which may be different.
Parent - - By Banned for Life (Gold) Date 2007-09-28 21:38
Erdo did a lot of testing on the CB server to test different openings that Zappa could use in the match against Rybka. Did the Rybka team do any testing like this or was it deemed unnecessary?

Alan
Parent - - By Alkelele (****) Date 2007-09-28 22:20

> Erdo did a lot of testing on the CB server to test different openings that Zappa could use in the match against Rybka.


As far as I know, Erdo didn't use any of the openings that he "tested" on the playchess server during the match. OK, maybe 3.Bb5 in the 2...d6 sicilian, but no games that went along the line of our game in the match. Erdo did say that he had tested the position we got in that game 5, though.

> Did the Rybka team do any testing like this or was it deemed unnecessary?


Only two possible answers? Btw, since Erdo made all those tests publicly available, why should we waste our CPU time performing them? ;-) Also, how to test reliably when we are going to play with different versions in the match? We did use Zappa in our prep, and using Zappa during opening analysis as well as reviewing the games from the playchess server allowed us to find out about important strengths and weaknesses of Zappa relative to Rybka.
Parent - - By Banned for Life (Gold) Date 2007-09-28 23:20
I suspect that Team Zappa has a much better appreciation of the types of positions that Zappa plays well against Rybka than vice versa.
Parent - - By Alkelele (****) Date 2007-09-28 23:24
When you win, you're a genious, when you lose, you're an idiot, sure :-) Anyway, I firmly believe you are wrong, and I also believe that we were quite effective in taking advantage of our knowledge. Still, when it is not reflected in the result, everybody is of course welcome to consider Jeroen and me idiots. I still wonder if we watched the same games, though :-)
Parent - - By Banned for Life (Gold) Date 2007-09-28 23:33
Yes, you seem very sure of yourself. I suspect that traditional testing at 60-20 would show that the latest version of Rybka is 50-100 Elo above the latest version of Zappa, but if a rematch were held tomorrow, I'd handicap it as dead even.
Parent - By Alkelele (****) Date 2007-09-28 23:56
Incredible. Words fail me.
Parent - - By Lee Ma Hong (**) Date 2007-09-29 01:57
If a rematch were held tomorrow, I believe Rybka would win hands down this time
Parent - - By turbojuice1122 (Gold) Date 2007-09-29 02:08
If a rematch were held tomorrow, the result would be roughly even--perhaps 5.5-4.5 in Rybka's favor, perhaps same as now, perhaps drawn.  Neither side has significantly improved their engines since the match.
Parent - - By Lee Ma Hong (**) Date 2007-09-29 02:19
Larry has pointed out elsewhere that the latest version was used in the last 4 games of the match, and the result was in Rybka's favor

"Neither side has significantly improved their engines since the match."

Oh? I didn't know you were in Vas' inner circle ...
Parent - - By turbojuice1122 (Gold) Date 2007-09-29 02:33
The match just ended, so it would be very, very unreasonable to assume, as you might have done, that either engine has significantly improved in strength.  As for the result being in Rybka's favor in the last four games of the match, I knew that someone would bring this up; perhaps I should have addressed it preemptively.

In the last several games, the Rybka team successfully employed a strategy of playing openings that are fairly well suited toward Rybka's style of play.  The Zappa team needs some time to adjust for this, just as the Rybka team needed time to adjust for the fact that the Zappa team was successfully sending the opening into channels in which Zappa plays well.  Second, even in match conditions in which each game is prepared for extensively ahead of time, you cannot place any emphasis on the result of only four games.
Parent - - By Banned for Life (Gold) Date 2007-09-29 02:53
I think your assessment here is more on target than up above, but I don't think Zappa was ever as good an engine as Rybka, either 2.3.2a or the follow ons. There was too much ground to make up, and even the optimistic end of Anthony's estimate of strength improvement wouldn't have been enough to catch even 2.3.2a, let alone the newer versions.

I agree with your assessment that at the start of the match, the Zappa team seemed better prepared to play to their strengths and Rybka's weaknesses. This wasn't always enough. Erdo did spend quite a bit of time in the engine room, and while some might believe he was just there to while away the hours, I suspect he was trying to figure out what positions Zappa found most favorable in games against Rybka.

Regards,
Alan
Parent - By cma6 (****) Date 2007-09-30 23:35
Is there any consensus on which aspects of  Zappa Mexico provided the edge over Rybka: superior endgame analysis; better scaling to multi-core; etc?
                If better scaling is the main issue, then Vas' current/future work on scaling to quad-core, which will be come common in chess, may be the most productive use of his time?!
Parent - By Lukas Cimiotti (Bronze) Date 2007-09-29 07:50
There was no improovement since the match so far, that's right. But the engine was improoved during the match. Games 7, 8 and 9 were played with a version that was significantly better than the versions, that played in games 1-6. The best version only played in game 10.
I'm currently running a test of that version 3min. + 1sec. against Zappa on 2 8 core computers, using a neutral book. Zappa's score after 10 games is +0 =5 -5.

Regards,
Lukas
Parent - - By Banned for Life (Gold) Date 2007-09-29 03:22
There's no question in my mind that Rybka is a better engine than Zappa and has been all along. Rybka also has, in my opinion, the best bookmaker in Jeroen. This combination is clearly sufficient for tournament play. But matches are different. They require more preparation. I think the Zappa team took a more constructive approach to this and did a lot of testing against Rybka to come up with openings and settings, while the Rybka team, blinded by overconfidence, didn't bother to carefully study Zappa's play.

And just between you and me, I agree with you that Rybka would almost certainly win if there was a rematch tomorrow, but only because the lessons from the original match were too painful to completely ignore.

Regards,
Alan
Parent - By dareapa (**) Date 2007-09-29 05:35
I would have to agree with you, Alan ,on this..
Parent - - By Alkelele (****) Date 2007-09-29 05:47

> But matches are different. They require more preparation. I think the Zappa team took a more constructive approach to this and did a lot of testing against Rybka to come up with openings and settings, while the Rybka team, blinded by overconfidence, didn't bother to carefully study Zappa's play.


I believe you are wrong. I also believe I have slightly better conditions for knowing about this than you. I just don't understand how you came to your conclusions, for example, about us being overconfident. I personally always held that the match was extremely difficult. We also spent oceans of time analysing strengths of Zappa and weaknesses of Rybka. We discarded dozens of opening systems and lines due to these discoveries, or repaired them or tuned them so we would avoid unpleasant stuff. We chose specific lines when we would detect a specific weakness in Zappa.

And we succeeded:

In the 1st game, we booked slightly wrongly with bxa4 (instead of Bxd5), but our general knowledge of this Zaitsev system was that Zappa underestimated black's queenside pawns. And the game confirmed this, when Rybka turned a slightly inferior position into a slightly superior position.

In game 2 we got a huge advantage out of the opening in a Ruy Lopez, and won.

In game 3, we were not booked deeply enough in a rare line, and Rybka chose a bad move. So, go ahead and blame us, but the reason we chose to play 1...e5 again already was EXACTLY because we had discovered specific Zappa-Rybka problems in our intended sicilian, so this doesn't fit your ideas about our blind overconfidence either.

In game 4, we got a winning position straight out of book. Yeah, go ahead and blame us for not predicting that Rybka would first miss the win, then throw away the draw.

In game 5, you may be able to give Erdo credit for choosing a line that Rybka handed badly. So shoot us. Yeah, naturally, we were dumb to even think that 1.e4 c5 2.Nf3 d6 may be a viable option. Btw, Nf6 instead of Nc6 was booked also exactly because of highly sensible analysis of Zappa-Rybka issues in a mainline after Nc6.

Game 6, I have already explained about it in the forum.

Game 7, we had to win as black, so I worked all night to allow us to play 2...Nc6 instead of 2...d6. We feared the Maroczy bind, but Erdo didn't choose that, and we got a fine position for Rybka to play for a win in.

Game 8, we outbooked Zappa in a Ruy Lopez again, choosing slightly offbeat 6.d3, with transpositions into various anti-marshalls, achieving a clear advantage out of the opening. Rybka chose bad plan Nf5 (instead of d4 or Rf1). Do you want us to discard 1.e4 or 3.Bb5 because Rybka may or may not in one specific position choose the right plan?

Game 9, I had to prepare during night for this along with game 8. I prepared Caro pretty much from scratch, and Rybka was allowed to play for a win again.

Game 10, Zappa played unintended move 3...c5 (instead of 3...d5) and relied on random stats from then on. Our handpicked moves gave Rybka a huge advantage, but she failed to convert. Yeah, shoot us again.

Bottom line is, we lost, we had freakish bad luck in two games, but, sure, when the stronger engine lost, who else can you blame than the bookmakers :-) And yes, surely 10 games and my few postings here intended to help provide decent coverage are enough to draw wild conclusions about our attitude, methods, quality etc. etc. You, as a scientist, should know this better than anyone.

I have read a lot of stuff in this forum the last few days that basically amounts to abuse. I am just surprised that a guy like you could not restrain yourself. Anyway, if you would like to substantiate your impressions, maybe we can carry on this discussion in an interesting way.
Parent - - By Roland Rösler (****) Date 2007-09-29 06:26
Conclusion: 1.) Rybka is the better engine and 2.) had the better book and 3.) freakish bad luck! That´s all!? Maybe you are rigth in 2.), but then we have to look at 1.), because I don´t believe so much in 3.).
Parent - - By Alkelele (****) Date 2007-09-29 06:48
Yes, we had very bad luck in two games, but that doesn't mean that the Zappa victory was not well-deserved! Anthony built a strong engine, and Erdo navigated the book battle according to the demands of the situation. In other words, they did what was necessary, and that's all you can ask from a winning team :-)

I personally view a match like this as a sporting event. Stats from testing groups can be useful as guidelines for overall strategy, but once the engines and books match up, it's all about who is stronger on the day. It's sport, not science. I don't think it is right to expect every 10-game match to confirm the "scientific" truth.
Parent - - By Uri Blass (*****) Date 2007-09-29 09:39
If we talk about luck I can say that you had good luck in game 7 when zappa missed a win so I do not think that the rybka team
can complain about luck.

It seems to me that the zappa team did a better job in book preperations because I do not believe that zappa is stronger engine than rybka.
The question is not objective evaluation of the position out of book but if rybka knows to play the book position.

If zappa get inferior positions with black when zappa knows to draw them against rybka then it means that the book preperation of zappa worked because the book pushed rybka to positions when she does not know what to do.

Uri
Parent - - By Venator (Silver) Date 2007-09-29 12:45
It seems to me that the zappa team did a better job in book preperations

Really!? Lost position quickly after leaving book in game 2, lost position in game 4 right after book ended, bad position in game 10 right out of book...
No advantage as white in games 5, 7 and 9. Really, start to be a bit more objective Uri! Just a little hint.

Have a nice weekend,

Jeroen
Parent - - By Uri Blass (*****) Date 2007-09-29 14:15
Talking about being more objective then the positions are inferior positions and not lost positions.

Remember that zappa missed a draw in the endgame of game 2.
also in game 4 there was no position that I can consider as win but only better position(Rybka missed Qe1 but this was not a simple move to find and many moves after the book was finished).

No advantage as white in game 5 7 9 is not exactly correct because rybka won game 5 and missed a win in game 7 so it seems that zappa knew to play part of these positions better than rybka.

Getting equal position with white for zappa when zappa knows to play it better than rybka is also a success for the zappa team.

Note that
I do not say that you did a bad job in the opening but only that I guess that the zappa team did a better job
and I do not buy the bad luck story for explaining the result because I could see that zappa also had a bad luck in missing a win in game 7(and here probably the question if zappa could find hxg7 is really a question of luck because parallel search is not deterministic and analysis after the game showed that zappa could find it).

Uri
Parent - - By Venator (Silver) Date 2007-09-29 15:42
Zappa won the match because it made fewer mistakes. If we would reverse the openings and replay the match,
I fear it could be something like 7-3 for Zappa.

Objectively black was lost three moves after leaving book in game 2, immediately after leaving book in game 4
and a couple of moves after leaving book in game 10. No matter how you want to call it, it is the way it is. That
Zappa didn't lose these games, was due to it's tough resistance and Rybka making mistakes. But as I told you,
what happens after book is a whole game: middlegame and endgame. Programmers nor book authors can do
anything anymore. In the middlegame and endgame the engine strength, luck, blundering and so forth kick in.
In these 10 games Zappa was overall better on these points. Therefore it deserves the win.
Parent - - By Uri Blass (*****) Date 2007-09-29 16:11
You may be correct in the claim that zappa could do better with opposite colors
but if it is the case it proves that zappa understands the positions better than rybka
and maybe it means that you did a mistake in choosing the opening and you chose openings that rybka does not know to play well.

I wonder if you tried to play games and calculated statistics of rybka against zappa in different openings in order to help you to decide about the opening choice.

An example:
If you find that statistically rybka does better against zappa with 1.e4 c6 and not with 1.e4 e5 then it may be better to prefer 1.e4 c6

Uri
Parent - - By turbojuice1122 (Gold) Date 2007-09-29 17:07
Such statistics would be unreliable at this stage--Zappa Mexico is not Zap Zanzibar, and I think statistics obtained from playing against Zap Zanzibar wouldn't work very well.  Zappa Mexico was placed on the consumer market just before the match, but to obtain reliable openings statistics, quite a large number of games at reasonably long time controls would need to be played (since Zappa Mexico stinks at blitz and rapid).
Parent - - By Banned for Life (Gold) Date 2007-09-29 18:24
If you compared the move ordering of Zap Zans, Zappa Mexico, Rybka 2.3.2a, and Rybka 2.3.x in a large number of positions, you would see that the ordering was highly correlated for Zap and Zappa. Same for Rybka 2.3.2a and 2.3.2x. You would also find that move ordering correlation was much lower between the Zs and the Rs.

You would want the final versions of the engine (running on the match hardware at match time controls) to come up with accurate statistics on how the different openings faired, but for predicting opponent moves, Zap Zans would be very beneficial.

Regards,
Alan
Parent - - By Venator (Silver) Date 2007-09-29 19:36
Zappa Mexico was only available shortly before the match, so that left little time for such conclusions. Of course we used
Zap!Zanzibar to base our book conclusions upon.

BTW, you don't know if the commercial Zap is the same Zap as playing in the Mexico match. And to have good statistics,
you need hundreds, maybe thousands of games in various lines, preferably with the same hardware and the same playing
tempo as in Mexico. This cannot be done in a short period of time.

Our opponent had a much easier time here: he only needed to collect 2 million Rybka-Rybka Playchess games to look for
weaknesses in the Rybka repertoire.

Regards, Jeroen
Parent - By Bouddha (****) Date 2007-09-29 21:00
Hi !

I do not understand you last comment as it seems that the Rybka used was not 2.3.2a.
So to me it seems that in this case Rybka team was in same conditions than Zappa team.

regards
Parent - By Banned for Life (Gold) Date 2007-09-29 21:03
Our opponent had a much easier time here: he only needed to collect 2 million Rybka-Rybka Playchess games to look for
weaknesses in the Rybka repertoire.


Yes! This is exactly right. It was much easier for Team Zappa to collect games, find weaknesses, and do initial testing because they had access to the Rybka Room and there is no corresponding Zappa Room.

BTW, you don't know if the commercial Zap is the same Zap as playing in the Mexico match.

First, an aside: If you're an engine developer,  unless you are already number 1, keeping your best stuff for tournaments is a double-edged sword. It will certainly help you in tournaments, but you will get mauled in the rating lists and on the servers. I can't see this as a good trade-off, but its certainly a possibility.

Second, improved versions of engines tend to exhibit move orderings that correlate very strongly with previous versions. As an example of this, people can follow game 10, which purportedly used the latest and greatest Rybka, with 2.3.2a without major surprises in move choices. I suspect that the changes to Zappa are of the same magnitude or less. The bottom line here is that engine updates may have a significant effect on the outcome of match, but the top move will change in only a small number of cases.

And to have good statistics,
you need hundreds, maybe thousands of games in various lines, preferably with the same hardware and the same playing
tempo as in Mexico. This cannot be done in a short period of time.


This is true both for book making and engine development purposes. Larry has stated that he validates changes in the evaluation function using large numbers of quick games. This isn't ideal, but its the best he can do. You must have your own methods of dealing with the almost infinite number of possible opening lines in a finite period of time.

Regards,
Alan
Parent - By NATIONAL12 (Gold) Date 2007-09-29 21:56
cozzie says it is the same prog as the commercial version.
Parent - - By Banned for Life (Gold) Date 2007-09-29 18:40
Suppose you had played Rybka versus Zappa 20 times with the opening that transpired in game 4. Clearly, this is won for white, but suppose Rybka kept making the same mistakes and losing the same way. Would you leave this "winning" opening in your book?

There are two ways of thinking about this:

1) The position is won. The engine developers should change the engine to win it! or
2) The position is won, but in the short run the engine can't win it and in the long run we are all dead. Put something else in until the engine is improved.

In my opinion, the book has to follow the engine so you would have to take option 2. Others, (probably including Vas and Larry) might say, stick with 1 and we'll fix the engine. This is more of a philosophical point maybe, but its certainly something the underdog needs to do, in my opinion, to have good chances in this type of match (and I'm pretty sure that Team Zappa did this). I think the favorite should do this too, but one could argue otherwise (and that's why I asked the initial question).

Regards,
Alan
Parent - - By Venator (Silver) Date 2007-09-29 19:04
When booking this line, I never thought it would pop up in a very important match. Erdo had never played the Zaitsev before, so when I booked the whole stuff with Qd2! and saw Rybka liking it a lot that was enough for me. So I was truly very surprised to see this nice piece of preparation being played in game 4! But it is of course impossible to follow your suggestion, as you never know when the opponent gets out of book. Suppose I knew where Zappa would be out of book, test that position and the score would be something like 16-4. That indicates a succesrate of 80%, but it would be completely useless, as Erdo showed in game 6 that black has a forced draw.

Of course I know what you mean and for me it is simple: if a line is good for white but Rybka doesn't understand it, misevaluates it or has a bad score with it, then I will not play that line. So your option 2 is the best one.

Regards, Jeroen

PS  The whole line with Qd2! in the Zaitsev has actually been played before, I recall that GM Kotronias is the inventor of that move. Ironically I saw the Kotronias game on ChessVibes when I was preparing the Zaitsev and still not knowing what to play against the c4-line. Sometimes you need that kind of luck :-)
Parent - - By Kapaun (****) Date 2007-09-29 20:59
Actually I do believe that, while there are some wins due to good book preparing, the majority of wins was decided late in the game by mistakes of one or both engines. Which means that book makers certainly can improve their work (who can't?) but the real problem this time appears to be the software. So, mainly it's up to Vas and Larry, right?
Parent - - By Venator (Silver) Date 2007-09-30 11:31
The only clear-cut book issue was game 6, in which the game ended in a very correct draw right after the opening. Although Rybka was much better after the opening moves in game 2, 4 and 10, you still need to win the game. A lot can happen, as this match clearly indicated. From opening until the final position there are many moves to be played. I think both programs had their missed chances, Zappa could also have drawn game 2 and do much better in game 8, which it lost. So it is not a matter of who's issue it is, as these are IMO normal things in computer matches. In many games there are ups and downs. We are used to the fact that if Rybka gets the better position, the chance is high she will win the game. But this was a match of 10 games against the number 2 of the world, which is different from a tournament like WCCC or Leiden. Besides, Erdo knows what he is doing and did a good job, while in other tournaments opponents are less well prepared to play Rybka. So the software makes mistakes, the book authors could do better, but this applies for both sides. In this respect it is not much different from a human match: those games have ups and downs as well, blunders, great opening play, mistakes in preparation and it is not sure the higher rated player always wins. This is just all about chess :-)
Parent - By Kapaun (****) Date 2007-09-30 12:52
I can certainly agree with that and would add only one little comment: Rybka definitely has her weak spots which can be used (by someone who knows what he is doing, like Erdo) to knock her out of the water. I am not up to decide whether these weak spots are more serious and dangerous now than they were before or not. But anyway the match made it clear enough that it's time to do something substantial about it. After all it's not like some of these weak spots (King safety for instance) haven't been known since quite a while now... ;-)
Up Topic Rybka Support & Discussion / Rybka Discussion / Rybka vs Zappa in Mexico City
1 2 Previous Next  

Powered by mwForum 2.27.4 © 1999-2012 Markus Wichitill