Not logged inRybka Chess Community Forum
Up Topic Rybka Support & Discussion / Rybka Discussion / Roman vs. Rybka 3 rematch
1 2 Previous Next  
Poll Score in "Dzindzi" vs. Rybka 3 match, Rybka without f7 pawn? (Closed)
Roman by 4-0 2 2%
Roman by 3.5-0.5 0 0%
Roman by 3-1 1 1%
Roman by 2.5-1.5 11 11%
Tie Match 2-2 18 18%
Rybka by 2.5-1.5 16 16%
Rybka by 3-1 29 30%
Rybka by 3.5-0.5 8 8%
Rybka by 4-0 13 13%
- - By lkaufman (*****) Date 2008-07-20 02:24
     As some of you may recall, after Rybka drew a match 4-4 with GM Roman Dzindzichashvili at odds of each of the eight pawns in turn, plus the White pieces in every game, we agreed to a tiebreaker match of four games at the even larger handicap of the f7 pawn every game, but at a faster time limit (25'+10"). Unfortunately Roman became ill right around the start of the match and we stopped the match after Rybka won three straight games as Roman grew sicker by the hour. It took him about a week to recover.
     Anyway he's feeling okay now, and has returned to chess competition with a victory in the South Carolina Open. We agreed today to redo the aborted playoff match, but in view of the further improvement in Rybka since then to set the time limit at 30'+20", only marginally faster than the limit in the earlier eight game match. He will receive a $200 fee plus $50 per point scored.
     The match is set for Monday July 28 at 11 a.m. Eastern Daylight Time (5 p.m. in Central Europe), with the other games starting at two hour intervals. I expect that by July 28 I will have the final version of Rybka 3 (maybe not in the new GUI, but that won't matter for this match), so this should be the first public display of the new program in action. I'll post information about where the match will be covered when I learn it.
     Aside from the longer time limit, Roman might benefit from reviewing the four games he's played with Rybka at this handicap (three from the playoff, one from the original match) as well as the four games of FM Meyer at pawn and three move odds. Can the improvements in Rybka in the four months since the first match offset the increased handicap and Roman's greater experience with handicap play now? We shall see.
Parent - - By sharpnova (***) Date 2008-07-20 02:50
Will the hardware be the same?

What was the hardware initially?
Parent - - By lkaufman (*****) Date 2008-07-20 03:48
The hardware will be the same (octal at about 3 GHz) as during the aborted tiebreak match. If my memory is right, my upgrade from quad to octal took place in between the initial match with Roman and the playoff match.
Parent - - By dareapa (**) Date 2008-07-29 04:50
HI Larry, I would love to see Rybka play a strong gm like Roman at full bore..If Roman loses no big deal..everyone loses ...If Roman does pull out a win then at least you can develop rybka based on Real chess....No More Handicaps..Real Chess, Real Chess,Real Chess :0..By the way I appreciate your work on Rybka..
Parent - - By lkaufman (*****) Date 2008-07-29 05:16
Roman is indeed a very strong player, particularly in rapid chess, but nevertheless he has failed to win a single game with the f7 handicap in eight tries (true, three were when he was ill, and he should have won at least one today). In order to have any likelikhood of winning a non-handicap game, even with all White pieces, he would probably have to play a hundred serious games, and that is just not going to happen. We develop Rybka primarily from playing against older versions of herself, but I often do learn something of value from these GM matches. I would probably learn a lot less from normal games, because Rybka would never be in any trouble in almost any game unless it related to the opening book, and that's not my department (except for the handicap book). I think each of the three GMs who played Rybka has contributed one or two ideas that are in Rybka 3. Maybe Carlsen or Anand with White could at least make a close enough fight to be interesting or instructive, but that won't happen unless some big corporation gets interested in sponsoring such a match. Perhaps these handicap matches might even help induce some such sponsorship by the implication that no human has a chance without a handicap.
Parent - - By dareapa (**) Date 2008-07-29 06:22
I want to witness the true power of Rybka against a strong human...If rybka thrashes the gm, then i won't feel as bad when it thrashes me :0)....Thanks for your reply..LK
Parent - - By lkaufman (*****) Date 2008-07-29 15:10
Well, you may get to see at least two such games against 2705 rated GM Milov, if sponsorship is found.
Parent - - By Geomusic (*****) Date 2008-08-02 11:17
I would think chessbase would sponsor it.
Parent - By lkaufman (*****) Date 2008-08-02 15:52
We have some sponsorship already, we just need a little more from any source.
Parent - - By FWCC (***) Date 2008-07-20 02:59
I'm afraid that the final version of Rybka will be too much for Roman for this handicap match.Roman did VERY well in the previous match and seemed to do better than previous players against Rybka,he played solidly.I predict a 3-1 Rybka victory.Maybe Roman should try playing Rybka at exchange odds also though I think Meyer is going to play her at exchange odds.
Parent - By lkaufman (*****) Date 2008-07-20 03:52
Roman previously said he preferred the f7 handicap rather than the Exchange (a1 and b8 off) handicap. My computer vs. computer tests rate the two handicaps as about the same. If John Meyer plays it will be at Exchange and White handicap.
Parent - By Uly (Gold) Date 2008-07-20 05:44
I predict 3-1, but I think it's a very optimistic score!
Parent - - By TedSummers (***) Date 2008-07-20 09:49
I predict Roman by 2.5-1.5, Hey I still believe in Mankind, what can I say! :)

Lord Rybka Forgive Me.
Parent - By lkaufman (*****) Date 2008-07-20 21:39
Roman is a very strong rapid player, he learns from experience, he has more time than before, and he starts with a winning position. Your forecast is not unreasonable.
Parent - By semipatz (*) Date 2008-07-22 20:37
Lord Rybka?  Wouldn't that be Lady Rybka the Piscine?
Parent - - By George Tsavdaris (****) Date 2008-07-20 09:52
I guess i should vote for 2-2 but i'm not so sure. What will be the contempt value you will set?
Also since Vas has said that there would be three Rybka 3 versions in the Rybka 3 package: default, humanized and dynamic, which of the 3 you will use for this match?

Is it humanized? And how humanized differs from the default and from dynamic(if you know of course)?
Parent - By lkaufman (*****) Date 2008-07-20 21:42
Probably I'll go with the default version, because the others run somewhat slower and this is rapid chess. I'll set a fairly high contempt value, maybe 50 or 60. As for the other versions, I'll put up something on them pretty soon. Yes, I know how they differ, since I made them!
Parent - - By Felix Kling (Gold) Date 2008-07-20 21:44
What happens if the match ends also in a draw?
Parent - - By lkaufman (*****) Date 2008-07-20 21:58
Maybe I can talk him into an Exchange odds match in that case.
Parent - - By Felix Kling (Gold) Date 2008-07-20 22:05
Parent - - By pokerpawn (***) Date 2008-07-20 23:48
think he will lose 3-1
afraid that on octal rybka 3 is just to strong for any human ....
Parent - - By Felix Kling (Gold) Date 2008-07-20 23:49
I think Dzindzi in a good mood will draw that match again :)
Parent - By (**) Date 2008-07-21 00:22
I believe he will lose this time 3-1 . Its without the f-pawn and not without the g-pawn :-/

Its how he say in his DVD. He was in the first match very very very lucky
Parent - - By Payonki Date 2008-07-21 05:57
I don't see the point of this match.  The match times are too short, the handicap makes a nonsense of the game and the hardware is not like anything that most people are using at home.

What I would like to see is a good human player playing Rybka at longer times without a handicap on the kind of hardware most of us are using day to day.

A match like that would give us a good idea about the strength of the software.  Why not offer a good player an attractive apearance fee and a large bonus if they can beat Rybka in any game.  That would encourage the player to try to win and we would see the real strength of the computer.

After all, Fritz 10 played Kramnik on a laptop.
Parent - - By Uly (Gold) Date 2008-07-21 06:13
I sort of agree, but:

1. - Real chess? That can't be played anymore because the GM would be destroyed and one that wouldn't be is charging ridiculous amounts of money. We're left with handicap matches.

2. - Normal Hardware? They want to prove the strength of the strongest computer on earth and they can't do so if they limit the hardware. It's like limiting the time anyway. Or the software! It's like running Naum 3.1 on the same hardware (my argument doesn't hold because Rybka is too strong, say Rybka 2.3.2a instead of Naum 3.1, but then all the work that has gone for Rybka 3 would now be wasted.)

> Why not offer a good player an attractive apearance fee

Perhaps this will change if chessbase sponsors, but right now it's even hard to find enough money to play Nakamura.
Parent - By Geomusic (*****) Date 2008-07-21 10:37
Some GMs are better opponents than others. Roman is one of the best versus rybka.  GO ROMAN!
Parent - - By (**) Date 2008-07-21 11:17
Chessbase sponsors ??? When did Chessbase sponsors a match in the past ? In Kramnik vs. Fritz, Chessbase was NOT the sponsor !
Parent - By Uly (Gold) Date 2008-07-21 11:50
I stand corrected :(
Parent - - By Felix Kling (Gold) Date 2008-07-21 12:11
yes, I think it was the Bundeskunsthalle in Bonn, which is known for wasting money of our government, and some companies.
Parent - By Felix Kling (Gold) Date 2008-07-21 12:13
from "rank zero" (a chess blog in germany):

In einem Bericht vom 15. Mai diesen Jahres stellte der Bundesrechnungshof

"zahlreiche Verstöße gegen die Grundsätze einer ordnungsgemäßen und wirtschaftlichen Geschäftsführung fest. [..] Die Vergabe von Freikarten sei in großem Maße rechtswidrig. [..] Schriftliche Aufzeichnungen […] wurden im September 2006 im Sekretariat der Geschäftsleitung auf Weisung des kaufmännischen Geschäftsführers vernichtet."

you can find more here:
Parent - - By pokerpawn (***) Date 2008-07-21 23:38
is naka playing Rybka ?
Parent - - By Uly (Gold) Date 2008-07-22 01:14

> is naka playing Rybka ?

Yes, it has been arranged. They just need to gather enough money.
Parent - - By pokerpawn (***) Date 2008-07-22 01:17
nice , why didn't i read anything about it here ?
this is without odds ?
Parent - By Uly (Gold) Date 2008-07-22 01:58

> nice , why didn't i read anything about it here ?;hl=Nakamura

> this is without odds ?

No, odds are explained in that link. However, while searching for that bit of info I also read that the match with Nakamura is doubtful (because of money issues.)
Parent - - By lkaufman (*****) Date 2008-07-21 19:02
Roman is an extremely good rapid player but no longer has the stamina for 4 hour+ games. He is very comfortable with the time limit chosen. The hardware doesn't make that much difference, I think Rybka 3 on a good laptop might be stronger than Rybka 2.32a on my octal. And as for offering top players a large bonus for a win, it's like offering them a lottery ticket. Even Anand or Carlsen understands that his chances of an actual victory in a game are somewhere around 2%. Without a handicap, strong players will only consider the guarantee; they know that's all they will get.
Parent - - By semipatz (*) Date 2008-07-22 21:46
Do Anand and Carlsen actually "understand" this, or do you just think they do?

Also, there could be a sizeable payout for a draw...that would be a real incentive.
Parent - - By lkaufman (*****) Date 2008-07-22 22:21
I don't know what they really think, and they don't make a point of saying that they have almost no chance against the best computer program (although I think Aronian said so), but one of the characteristics of the best players is that they are usually rational and intelligent; certainly this applies to Anand and Carlsen. They have enough experience with computers to know the truth. Of course if you make it a draw odds match by giving them a big enough payout per draw then they would play, as their drawing chances with the White pieces might be close to even. Care to fund the match?
Parent - By semipatz (*) Date 2008-07-23 16:47
Sorry, I'm broke.  I'm having enough trouble figuring out how to pay for Deep Rybka!    :)
Parent - - By crescens (*) Date 2008-07-21 20:30
I'd like to see Rybka at full strength on good hardware against a team of GMs.
Parent - - By lkaufman (*****) Date 2008-07-21 21:45
It's not clear that GMs consulting will play much better than the strongest of them by himself, and in any case this would be too expensive to arrange.
Parent - - By Kreuzfahrtschiff (***) Date 2008-07-21 23:23
rybka will win 3,5:0,5

p.s.: i played with roman over 200 (maybe 500) blitzgames in the beginning 90s, he was very angry when he lost some of them  :) (about 10%)
Parent - - By Felix Kling (Gold) Date 2008-07-21 23:34
I heard someone won a 5 minutes blitz game against Dzindzi like this:
After the first move he stood up, went to the fridge, slowly took a beer, came back, had 2 minutes left on the clock and then crushed Dzindzi :)

Don't know if this is just a rumour, but beating a very strong player like Dzindzi that way would really be cool :) I heard that rumour in my chess club.
Parent - - By theoak (**) Date 2008-07-21 23:56
Are you sure he didn't have way too many beers to remember what exactly has happened? :) Well, at least this tactic didn't help Alekchine :)
Parent - By Uri Blass (*****) Date 2008-07-23 22:13
Alekchine did not use 60% of the game time to drink the beer so his results are not a reliable test for the question if the tactic is effective.

I believe that the tactics is not effective but  not because of Alekchine.
Parent - By (**) Date 2008-07-22 17:58
Dzindzi was a long time ago in Germany/Frankfurt (in the late 80's). He won a lot of money by playing chess in coffeehouses. Most of the time blitzgames. But one day, he played against a strong playing 2300 rated player and lost 90% of all games against him. He lost so much money. I dont know the name of this player, but he played everytime unusual openings and real coffeehouse-chess (for example,- push h-pawn with white and then sac rook on the h-file to open king-position). Roman lost and lost and lost. Years later, he made a video about "how you can use blitzgames to become better in chess". In this video he says, that its not the right way, to play uncorrect moves in the opening. Not in long games and not in blitz :-) If you want to annoy Roman, just play 1.h4 :-D
Parent - - By lkaufman (*****) Date 2008-07-22 01:12
I played him hundreds of games myself back around 1980 when we were both at peak strength; we found that it was about fair if he gave me 5 minutes to 1.5 minutes time odds.
Parent - - By Felix Kling (Gold) Date 2008-07-22 16:05
"1980 when we were both at peak strength"

Maybe your peak strength will be in the future ;)
Parent - By lkaufman (*****) Date 2008-07-22 17:58
I may actually play as well or better now as in 1980, but my relative standing in the world is vastly lower now. According to the Jeff Sonas historical ratings, I just narrowly missed being in the world top 100 then, and this is with my best result of my life missing from his data! Now the number of players who outrate me is a 4 digit number. It's just that the standard of play is dramatically higher now in general, due to computers, books, and internet.
Parent - By Geomusic (*****) Date 2008-07-23 09:37
lots of brilliant players get upset when patzers beat them.
Parent - - By semipatz (*) Date 2008-07-22 23:24
"It's not clear that GMs consulting will play much better than the strongest of them by himself, and in any case this would be too expensive to arrange."

It could pay for itself.  You have 3 GM's in a room, discussing what to play, and they are being videoed and webstreamed live.  Put that on Playchess with some proper can watch the game itself for free, but you have to pay some Ducats to hear GM's in the throes of deciding what moves to make, explaining their reasoning out loud.  You could bring in some serious dough with that, plus increase sponsor interest because of the amount of publicity that would attach.  Loads of chess players would kill to be able to see a GM's brain's gears turning, and you'd effectively have three.

I'd make another suggestion as well, whether against one player or three:  allow the human to play ahead on a separate board or computer screen.  This would be particularly fun if used by a team, but either way, it gives the human or humans a better chance.  Even though all strong players can play blindfold, they play a lot better when not blindfolded.  This moves in the other direction: from what I hear, tests show a sizable ratings boost to a player if he/she can use a second board to aid visualization.

This is one way to make a human more competitive with Rybka that is arguably NOT a handicap.  There is a big philosophical debate about what terms of play are equal between human and computer.  Some would said that letting the human play ahead gives an unfair advantage to the human, but others would say that the current system gives an unfair advantage to the computer.  After all, the computer's internal representation of the current position is in the same format as its representation of future positions.  This is only true for the human player if you let him move the pieces on another board.

Add in letting the human(s) have access to the same opening book and tablebases as Rybka, while obviously disallowing any engine access, and you have yourself the most favorable possible terms for the human that are even arguably NOT a handicap.  You might think it IS a handicap...let the philosophy debate begin.  At least no one can say it’s biased in the computer’s favor.  It's also a lot more palatable aesthetically, IMHO, than giving the human a pawn up or whatnot.  It's just regular chess from a theoretical perspective.  Only the practical conditions of the game are changed, and realistically the practical conditions can never be the same in human-computer chess as they are in human-human.  Current rules sweep this under the rug.
Up Topic Rybka Support & Discussion / Rybka Discussion / Roman vs. Rybka 3 rematch
1 2 Previous Next  

Powered by mwForum 2.27.4 © 1999-2012 Markus Wichitill