Not logged inRybka Chess Community Forum
Up Topic Rybka Support & Discussion / Rybka Discussion / Rybka vs. Benjamin - postscript
- - By lkaufman (*****) Date 2007-08-12 21:30
     Here are some additional comments about the match based on a lecture by Joel Benjamin at the Arlington Chess Club the next day, as well as our own discussions and unofficial quick games.
     The d2 handicap is bigger than it looks, because after 1e4 Nc6! (1...e5 as played by Ehlvest is inferior, as White isn't forced to play Nc3 early) 2Nf3 Nf6 as played White cannot play the usual 3e5 because after 3...Ng4 the move 4d4 is impossible. I feared this defense and he played it! So Rybka played 3. Nc3 but the Bb4 pin (after 3..e5) is effective. So 1e4 is not so good here.
     In the c7 handicap game, Joel anticipated the ...Qa5 idea, though not necessarily on move 1. He decided not to try for too much, but just a normal Spanish type position where the missing "c" pawn is significant. This seems to be the "purest" pawn handicap.
    In the b2 game, the tempo White saves on playing Bb2 is not terribly significant; White has no compensation for the pawn. The f4 Bird-style opening was recommended to me by Ehlvest after our game in the first match; it does seem to be better than what we played in that match.
    In the a7 game, I booked the French defense, because the Winawer variation works well, since a3 doesn't attack a bishop on b4. Joel saw this, and so chose the slightly unusual two knight's line. He felt that ...f5 was a terrible move, and if we learned nothing else from this match, perhaps we should try to find ways to avoid similar moves by Rybka.
     Regarding the e2 game, Joel feels that this is a tougher handicap for him than the d2 game, as 1d4 seems to have no drawback and the tempo White saves not needing to play e4 to get castled is useful. He felt in good condition for this game but lost rather badly nonetheless.
     In the traditional f7 handicap game, Joel thought I might have booked the 1...Nh6-f7 defense, just because it is not normally possible and would be especially entertaining for the spectators. It may not be the best defense, but is surely one of the top five. Here he kept a winning advantage but got too short of time and so chose to repeat. He feels that this handicap is a huge one, much more than the others, and that it might be possible for a GM to analyze it out to a fairly easily won position.
     The g2 handicap is less than it looks like according to Joel as the extra tempo for White is more meaningful than with b2 off, and also because White may castle long and attack on the "g" line. So he chose a line with early central action that will force White to castle short, in which case the missing "g" pawn hurts. He felt that 1c4 was the best move, and that was my choice for the book.
     Regarding the h7 handicap, Joel was afraid to open 1e4 because of the Scandinavian (1...d5) which puts Black on track to castle long. Indeed, that is what I had booked. After 1d4 it is very hard for Black to castle long, but Rybka was determined to do so and did indeed find a way. I booked the queen's Indian defense because it at least develops the queen's bishop and so gives Black some hope to castle long.
     General comments: The openings were very interesting, and in general Joel and I independently came to the same conclusions in most cases. Joel did not set out to make draws by repetition, but in game after game he found himself in time pressure when the repetition arose and without seeing a clear plan thought it wise to repeat. In the final game he just missed the a4! idea at the end.
     For any future pawn handicap matches the time limit needs to be slower for the GM to have good chances. After some discussion, we concluded that one hour plus one minute increment (instead of half a minute) would be the best for a two round per day match. Time pressure would be far less serious, but the games would still usually be short enough to keep the total playing time per day to something reasonable. Also this time limit meets the FIDE standards for a GM title tournament, so it would definitely be standard tournament chess. Joel said he would probably not have repeated in several games if the increment had been a full minute.
     Joel also played some extra quick and blitz games with Rybka for fun/practice at various handicaps, partly just to give us a feel for what we might use in future matches as Rybka gets even stronger. At three minute (no increment) chess he had a bad score at knight odds (though these games were played after the final match game so he was quite tired), but he won twice easily when I added just a one second increment. This is consistent with my observation that Rybka can give knight odds to a typical IM in five minute chess (no increment); for a GM three minute looks like the fair time limit. We also played seven games at various handicaps with a time limit of five minutes plus ten seconds increment, which is quick chess but not blitz (games can take half an hour). Joel lost twice at castling plus white odds (turning down many chances to repeat in one game, lost one and drew one at Exchange odds (a1 and b8), and lost by 2-1 at "pawn and two moves" (f7 off, White gets to make two moves before Black replies). He felt that f7 handicap was a larger one than the Exchange.
     Conclusions for future handicap play with GMs: The present pawn handicap is suitable for standard tournament games; pawn and move (f7) is suitable for rapid games like 30' or so, pawn and two moves is suitable for quick chess like 15' or so, and knight odds is suitable for three minute chess. 
     The next match with Joel is planned to be at odds of draw and move; he gets all White pieces (normal chess) but draws count as wins for him.
Parent - - By Felix Kling (Gold) Date 2007-08-12 22:56
"He felt that ...f5 was a terrible move, and if we learned nothing else from this match, perhaps we should try to find ways to avoid similar moves by Rybka."

I'm interested in the oppinion of other strong players, Vas said that it is not obvious to him that f5 should be bad and I think the move looks indeed ugly first, but could be the best try to get some play in this position, since moves like Nf6 offer black no chances for counterplay.

In Mainz we could ask some top GM, I would like to know what Morozevich (I hope he plays in Mainz) or Anand think :)
Parent - - By lkaufman (*****) Date 2007-08-12 23:12
Please do so, if you are there. My own opinion was that ....f5 is a dubious move, but not as bad as Joel considered it to be.
Parent - By Vasik Rajlich (Silver) Date 2007-08-13 08:21
When we have Krasenkow here for the freestyle tournaments, sometimes I get the impression that he's playing a different game. After a lot of pretty normal moves, he'll comment that "this creates a weakness". Or that in some positions, something is "weak".

It's hard to say how much these concepts are valid and how much they are a matter of style. Probably, they are reasonably valid.

Parent - - By turbojuice1122 (Gold) Date 2007-08-13 03:21
The next match with Joel is planned to be at odds of draw and move; he gets all White pieces (normal chess) but draws count as wins for him.

Have there been discussed many other details of this next match?  You mention in the article having a time control of 60'+60"--what will be Rybka's time control?  Will this be a time-odds match?  Will Rybka have a restricted opening book?  Or is draw-odds the primary handicap?  Also, if draws count as a win, will he get "bonus points" in the unlikely event that he actually wins a game?  Have you two decided roughly when the match will take place and how many games it will be?  I know this is a lot of sudden questions about "the next match" just after this one has completed, but since you brought it up, I figured I'd ask :-)
Parent - - By lkaufman (*****) Date 2007-08-13 03:38
Draw odds and all White pieces are the two handicaps -- time should be equal (probably 60'+60"), with no restrictions on the opening book. In my opinion, draw odds is a larger handicap than the various advantages (other than all White pieces) given to Ehlvest in his second match, so I don't think other advantages are appropriate. I thought about the idea of bonus points for a win, but this might encourage Joel to take chances which would spoil the draw-odds character of the match. As to when it will be played, we are talking about the second half of September, but this is far from decided yet.
Parent - - By JhorAVi (***) Date 2007-08-13 07:34
Draw odds is interesting but I believe it can't be simply handled by contempt adjustment. Rybka code should be modified instructing that draw is lost for her.
Parent - By lkaufman (*****) Date 2007-08-13 13:29
Yes, we definitely have some work to do before playing such a match.
Parent - - By turbojuice1122 (Gold) Date 2007-08-13 12:37
I assume that in the version that will play in September with draw-odds, the techniques used by Father to lock up the position shouldn't work?
Parent - By lkaufman (*****) Date 2007-08-13 13:28
I hope not, but this has not yet been addressed.
Parent - - By tasdourian (**) Date 2007-08-29 03:56
Is this match still "on"?  And do you actually anticipate Rybka winning this match (that is, winning more than it draws)?  I am wondering if you and Vas in some sense are drawn to the idea of ever larger challenges, even to the point where you think a match loss is more likely than a win, or if you in fact are trying to stick to challenges that you have estimated Rybka "should" win.
Parent - By lkaufman (*****) Date 2007-08-29 05:14
We are hoping to hold this match along with the Zap match in Mexico City, but we do not yet have final approval on this. If not, then we will probably hold it a few weeks later. Yes, we are drawn to ever larger challenges (though I wouldn't say that this one is a "larger" challenge than the previous pawn handicap) and will probably keep increasing the odds until we do draw or lose a match. But we don't increase the odds too quickly, so we never actually "expect" to lose, we just want to find out what's fair. And it's a moving target, because both Rybka and the hardware keep getting better. I doubt that we would have won the last match with Benjamin if we played on the hardware and software we had in March for the first Ehlvest match.
Parent - - By ernest (****) Date 2007-08-13 17:31
Hi Larry,

Did I miss something? Wasn't it also planned to play one FRC exhibition game?
Parent - - By lkaufman (*****) Date 2007-08-13 21:56
Sorry, I don't know what you refer to, unless perhaps you mean that Rybka will play in the FRC event at Mainz in the next few days, but this is only against other engines, not against human.
Parent - By ernest (****) Date 2007-08-14 17:53
Thanks, Larry.
I must have been daydreaming :-)...

(I seem to recall I read something about such a 9th game, just as a test of the Rybka FRC version...?)... never mind!
Up Topic Rybka Support & Discussion / Rybka Discussion / Rybka vs. Benjamin - postscript

Powered by mwForum 2.27.4 © 1999-2012 Markus Wichitill