Not logged inRybka Chess Community Forum
Forum RybkaChess Help Search Register Login
Up Topic Rybka Support & Discussion / Rybka Discussion / Was there ever a match between Rybka 3 or 4 vs top GM?
1 2 Previous Next  
- - By staylor (***) [il] Date 2010-07-11 14:26
I don't know if anyone is following tests with Rybka 4, but i DOES seem quite strong, even if the strength increase over Rybka 3 was small (for 2 years).
How high can chess get? Isn't it taxing on ANY GM to play Rybka 4?
So i was wondering when was the last time there was such a match?
Perhaps vs R4 will be ,more difficult for a human.
I seem to remember that humans managed to do beat Rybka 3 even with odds of a Pawn and move, or similar. Or maybe that is NOT the norm?
Parent - - By Felix Kling (Gold) [de] Date 2010-07-11 14:31
Larry arranged some matches with Rybka 3 were Rybka was giving odds (not the other way around!) http://rybkachess.com/index.php?auswahl=Events

Rybka 4 would crush any grandmaster easily, just like Rybka 3. So there's no point of such a match.
Parent - - By turbojuice1122 (Gold) [us] Date 2010-07-11 14:38
Yes, I think that when Kasparov said that "there's no way [he] would take on that monster!" something like 1 or 1 1/2 years ago, that should have ended all speculation about such matches.
Parent - - By staylor (***) [il] Date 2010-07-11 18:00 Edited 2010-07-11 18:02
Come ONNNN now!
Can't we have a serious world championship match with the strongest anti-computer human GM in the world (preferably the world champion, but if he is not the strongest vs computer, then it should be the one who IS)?

Let the GM prepare a few months, and lt be not a too too long match, and let us see what the human race can still do.

Even if the human get w1 L5 D6 out of 12 games, which is 8-4 to the computer, then this is a great result to the human.

I want to see if the one human win, is from the human superiority which still exists.

I also want to see if any of the draws should have been wins, and indeed, the losses too. I want to see what positions the human excelled in, even if it lost in the end.
Parent - By mikal (**) [us] Date 2010-07-11 18:49

>Even if the human get w1 L5 D6 out of 12 games, which is 8-4 to the computer, then this is a great result to the human.


Even having a draw on half of the games is very unlikely. Anti-computer style is nowadays simply setting up a closed position preferably a pawn fortress style and move your pieces back and forth and hope engine lose on time. But human advantage here is actually easy to avoid by simply setting up openings with early exchanges and open positions.
Parent - - By turbojuice1122 (Gold) [us] Date 2010-07-11 19:29
All of the "anti-computer" games that we have seen are where the GM comes up against a computer with a book tuned for playing against other computers.  Books can easily be tuned to avoid anti-computer openings, and this is exactly what would occur in a match against a GM.  I would predict a sweep in favor of the cluster in a 12-game match against any human.
Parent - - By 8lrr8 (***) Date 2010-07-12 02:34
so u're saying a 2750(+) player wouldnt even get a single draw in 12-games?
Parent - By Athinira (**) [dk] Date 2010-07-12 04:27
Sure they could.

But expecting them to draw more than 2-3 times in such a scenario is unrealistic.
Parent - - By turbojuice1122 (Gold) [us] Date 2010-07-12 05:15
I think that if you had several such matches, you might see a draw or two.  The style of the cluster, though, seems very anti-human in addition to anti-computer.  I think that an anti-human book would put most super-GMs into a losing position before move 20.  I'm sure that others have much more authority to make such statements on this matter, but I have a feeling that they would predict similarly. :twisted:
Parent - - By 8lrr8 (***) Date 2010-07-12 05:38 Edited 2010-07-12 05:44
well, 2 draws is a hugely different result than 12 losses.

my guess is, by adopting a rope-a-dope strategy, a top 5 player should be able to draw half of his white games, and maybe 1 game as black, putting the avg result from +0=3-9 up to +0=4-8.
Parent - - By turbojuice1122 (Gold) [us] Date 2010-07-12 12:04
I'm not talking about 2 draws in a 12-game match.  I'm talking about two draws total in several 12-game matches.  I don't think that the GM would get a single draw in the first 12-game match.
Parent - - By Harvey Williamson (*****) Date 2010-07-12 13:29 Edited 2010-07-12 14:21
Topalov might get a draw or 2 as he has experience of the Cluster.

Hiarcs played a 2700 GM recently he chose the time control http://hiarcs.net/forums/viewtopic.php?t=3512
Parent - - By Bouddha (****) [ch] Date 2010-07-12 16:07
Time control of 1min for a game are imho not very interesting.
Parent - - By Harvey Williamson (*****) Date 2010-07-12 16:09 Edited 2010-07-12 16:13
I agree but it was what he wanted to play. I guess he wanted to try and block and play Father style. Hiarcs on Playchess in the main room is open to all reasonable challenges. However i do think that any engine in the top 10 on reasonable hardware would get an almost identical rating v Humans as they have little chance of getting a draw. On Playchess I usually set the book on wild also with a .50 contempt. This can help as they do not get repeated openings when trying to get a blocked position to win on time. Although Hiarcs when doing this often comes out of book with a minus.
Parent - - By Werewolf (*****) [gb] Date 2010-07-12 16:21
What's Hiarcs like v Father? Rybka seems to struggle against this still.
Parent - - By Harvey Williamson (*****) Date 2010-07-12 16:27 Edited 2010-07-12 16:29
it all depends. When i know I am about to play him I score 100% it is important to set contempt and the GUI time lag parameter to on. If I am ready for him I score 100%. Most of his wins are due to lag. Setting GUI time lag on will make hiarcs move instantly when it gets to less than 10 seconds on the clock - not even his premove can keep up. I have even beaten him on time.
Parent - - By Ray (****) Date 2010-07-12 20:44

> Setting GUI time lag on will make hiarcs move instantly when it gets to less than 10 seconds on the clock - not even his premove can keep up. I have even beaten him on time


Magnificent, love it !
Parent - By Harvey Williamson (*****) Date 2010-07-12 20:51
:cool:
Parent - - By Lukas Cimiotti (Bronze) [de] Date 2010-07-13 10:22
There are several "anti - Father" setups that work :smile:
A really easy one was using Deep Shredder 8 and enabling "Prefer Open Positions" - this gave 100% success. But I think your Hiarcs setup will also work perfectly. For Rybka it should help lowering the value of own pawns - or setting a relatively high buffer.
Parent - - By Fulcrum2000 (****) [nl] Date 2010-07-13 16:36

> For Rybka it should help lowering the value of own pawns


This doesn't work as Rybka thinks it's far behind in material and will do everything to get a 3-fold repetition. See one of my other posts for the games.
Parent - By Lukas Cimiotti (Bronze) [de] Date 2010-07-13 17:23
So you need a compensation by raising own pieces values :wink:
Parent - - By Harvey Williamson (*****) Date 2010-07-13 14:49
http://hiarcs.net/forums/viewtopic.php?p=47260#47260 The top GM's can draw sometimes.
Parent - By Uly (Gold) [mx] Date 2010-07-14 02:19
Here's a FEN standard copy (the other fails to paste to Winboard)

[Site "Main Playing Hall"]
[Date "2010.07.13"]
[White "Oligarckh (2991)"]
[Black "Comp Hiarcs,HIARCS 13.157 MP (2473)"]
[Result "1/2-1/2"]
[ECO "D12"]
[Opening "Slav"]
[Variation "4.e3 Bf5 5.Nc3 e6 6.Bd3 Bxd3 7.Qxd3 Nbd7"]
[TimeControl "180+1"]
[PlyCount "98"]
[WhiteType "human"]
[BlackType "computer"]

1. d4 d5 {B/0 0} 2. c4 c6 {B/0 0} 3. Nc3 Nf6 {B/0 0} 4. Nf3
e6 {B/0 0} 5. cxd5 exd5 {B/0 0} 6. Bg5 Bd6 {B/0 0} 7. e3 h6
{B/0 0} 8. Bh4 Bf5 {-0.01/0 0} 9. Bd3 Bxd3 {-0.01/0 0}
10. Qxd3 Nbd7 {B/0 0} 11. O-O O-O {B/0 0} 12. Rab1 Re8
{0.05/16 23} 13. b4 (13. Rfe1) 13... a6 {0.04/15 6} 14. a4
(14. Rfe1) 14... Nb6 {0.00/16 12} 15. Nd2 (15. Rfe1)
15... Qe7 {0.03/15 6} 16. b5 axb5 {0.09/15 1} 17. axb5 Nbd7
{0.10/16 5} 18. bxc6 (18. Rfc1) 18... bxc6 {0.10/14 1}
19. Bg3 (19. Rfc1) 19... Bxg3 {-0.02/15 5} 20. hxg3 c5
{0.00/16 4} 21. dxc5 Nxc5 {0.01/17 0} 22. Qd4 Nce4
{-0.06/17 0} 23. Nf3 (23. Rbc1) 23... Rac8 {-0.09/14 4}
24. Rbc1 (24. Ne2) 24... Qa3 {-0.60/17 5} 25. Ne2 Rxc1
{-0.21/16 0} 26. Nxc1 Ra8 {-0.12/17 0} 27. Ne2 (27. Ne5)
27... Qa2 {-0.28/16 5} 28. Nf4 (28. Nc3) 28... Rb8
{-0.06/16 10} 29. Nd3 (29. Qa1) 29... Rb1 {-0.03/16 7}
30. Rxb1 Qxb1+ {0.00/19 0} 31. Nfe1 h5 {0.00/18 2} 32. Qb4
Qxb4 {0.00/18 2} 33. Nxb4 Nc3 {0.03/20 4} 34. f3 (34. Kf1)
34... Ne8 {-0.15/18 4} 35. Kf2 (35. Ned3) 35... Nd6
{-0.25/20 4} 36. Nec2 Nc4 {Oligarckh offers a draw -0.25/19
0} 37. Nd3 (37. Na6) 37... g5 {-0.21/19 5} 38. Ke1
(38. Nc1) 38... Kg7 {-0.17/19 3} 39. Nf2 (39. Nc1) 39... f5
{-0.19/17 4} 40. Nd1 Na2 {-0.10/18 4} 41. Nf2 (41. Nd4)
41... Kf6 {-0.12/18 4} 42. Nd1 Nd6 {-0.03/19 5} 43. Nf2
(43. Nd4) 43... Nc3 {-0.13/17 5} 44. Nd1 (44. Kd2)
44... Ncb5 {-0.15/17 2} 45. Ke2 Nb7 {-0.07/18 9} 46. Nd4
Nxd4+ {-0.06/20 0} 47. exd4 Ke6 {-0.06/19 2} 48. Nf2 Nd8
{-0.06/21 0} 49. Nd3 Kd6 {Comp Hiarcs,HIARCS 13.157 MP
offers a draw -0.06/21 0} 1/2-1/2
Parent - - By 8lrr8 (***) Date 2010-07-12 17:59 Edited 2010-07-12 18:04
wow, u really dont have much hope for humans anymore, eh? :wink:

maybe i'm overestimating white's advantage esp. if the human aims for the draw from the get-go?  while anand, carlsen, et al, may not draw at will as white, it's really hard for me to imagine they cant draw ~50% of the time (as white) if they put that as their objective.

imo, the hydra v. adams and rybka v. benjamin match results arent particularly good data pts to hypothesize this conjecture.
Parent - - By turbojuice1122 (Gold) [us] Date 2010-07-12 19:18
No, the Hydra match isn't a good data point at all--the Rybka cluster is about 400 elo points stronger than the Hydra system. :twisted:
Parent - - By 8lrr8 (***) Date 2010-07-13 05:11
if hydra was 2900, u're saying the cluster is 3300.  i'm not convinced the cluster is quite that strong (at 40/120 w/ no restriction on opening books).

otoh, at the highest levels, i get the feeling the opening book is at least as important if not more important than the actual engine/hardware.  so if hydra's book sucks ass, all bets are off.
Parent - - By Werewolf (*****) [gb] Date 2010-07-13 08:44
Hydra = 3100 according to Chrilly.

Cluster about +300 elo over Rybka 4 UCI on a fast quad - this issue came up a lot at the last tournament.
Parent - By 8lrr8 (***) Date 2010-07-13 16:10
i realize 2900 is most likely too low a figure for hydra, but 3100 is likely too high, imo.

as for the cluster being +300 over an o/c quad, it really depends on the testing conditions.
Parent - - By Labyrinth (****) [us] Date 2010-07-13 09:21
I think I agree with your general sentiment.

I feel like there is so little data on serious human vs. computer matches that I think we must be skeptical of the predictions.

Of course the computer is the favorite but I think that people tend to exaggerate the difference.
Parent - By 8lrr8 (***) Date 2010-07-13 16:23
right, the data is sparse, and a bit tainted.

the payout for the hydra v. adams match was such that he was heavily influenced to play for a win w/ both colors.  if i recall, he would've received $25g for one, but only $5g for each draw.

as for the benjamin match, if we learned anything from it, it may simply being a <2650 player simply isnt strong enough draw w/ white more than a quarter of the time.  but it doesnt necessarily mean a 2750+ player cant do it half the time.

and it could also be said both mickey and joel were likely seriously underestimating the strength of their silicon opponent (their play being affected as a result).
Parent - By Harvey Williamson (*****) Date 2010-07-11 19:34
for a match like this you are talking 1 million dollars then you can open negotiations. Sadly those days are gone.
Parent - By Labyrinth (****) [us] Date 2010-07-11 20:21
I'd love to see a an exhibition match with Carlsen vs. the cluster. Long time controls of course.

I have a feeling that Carlsen would do better than expected.

Too bad that even if he was interested, he would have to demand a hefty fee. I don't think there is enough interest for someone to sponsor the match.
Parent - By SpiderG (***) [us] Date 2010-07-12 02:11
I totally agree staylor... I would like to watch a Super GM-Super Comp match... The problem is finding a top GM who would do it imo! :smile:
Parent - By Quapsel (****) [de] Date 2010-07-12 10:55 Edited 2010-07-12 10:58

> Can't we have a serious world championship match with the strongest anti-computer human GM in the world


I'm not interested to see a human fighting against a very large book and a very large tablebase, with some engines-moves in the middle.
Lets find a strong anti-computer human GM, and take Rybka4 (or whatelse seems to be the strongest engine), give her a very small (or really no) book and no tablebase.
And lets see how that GM and that engine will finnish then.
Perhaps the GM has no chance even then. But I do want to see this.

Other idea:
Give R4 only 5 seconds exactly for each move on a normal Quad (perhaps i7),
and let the GM play 40 moves in 2 hours.
What engine-strength could result then?

Quap
Parent - By apollox (***) [gb] Date 2010-07-12 15:50
I think you would be disappointed. I watched a few GM/IM v machine matches in the Rybka era, usually with odds of knight, rook or pawn and move. The matches tend to be pretty tepid affairs and it is usually the human trying to devise a draw strategy or as some like to say 'anti-computer tactics' to thwart the aggression of the engine.

> Let the GM prepare a few months, and lt be not a too too long match, and let us see what the human race can still do.


I am afraid the writing was on the wall in 1997. The Deep Blue match signified that the chess engine was coming of age and the human GM was no longer going to dominate the board. Even getting a draw against Rybka on even strength is a herculean task. Rybka does not tire or feel the stress of a situation - it is a huge advantage over flesh and blood.
Parent - - By 8lrr8 (***) Date 2010-07-12 02:33
"...Kasparov said that "there's no way [he] would take on that monster!"

where was this from, a chessbase article?
Parent - - By turbojuice1122 (Gold) [us] Date 2010-07-12 05:16
I've actually forgotten--it might have been somewhere different (my feeling is that it was).  I don't feel like searching. :-)
Parent - - By 8lrr8 (***) Date 2010-07-12 05:32
was it from this thread?: http://rybkaforum.net/cgi-bin/rybkaforum/topic_show.pl?pid=135186;hl=kasparov

that's the only thing i can remember that had any reference to kasparov making a comment on rybka's strength.  even so, i'm not sure kasparov was "shaking in his boots."
Parent - - By turbojuice1122 (Gold) [us] Date 2010-07-12 12:02
No, that wasn't it.  He definitely used the term "that monster".  Strangely, I cannot find this with a Google search, but I am certain of that quote.
Parent - - By Athinira (**) [dk] Date 2010-07-12 12:28
Video interview perhaps?
Parent - - By turbojuice1122 (Gold) [us] Date 2010-07-12 17:30
Yes, I realized afterward that it must have been in a video, though precisely which one is fading.  I feel, though, like it was something that I read instead of watched, but then it should be able to be Googled.
Parent - By Sesse (****) [fr] Date 2010-07-12 21:18
Perhaps you read the subtitles.

/* Steinar */
Parent - - By lkaufman (*****) Date 2010-07-12 16:31
I personally asked Kasparov about a year and a half ago if he had any interest to play a match with Rybka. He put up his hands in a gesture as if to say "no way".
Parent - - By 8lrr8 (***) Date 2010-07-12 18:01
do u think he meant "no way" as in "i'm not even close to my prime and i dont want to be publicly embarrassed," or "no way" as in "even in my prime, i'd get beaten like a rented mule"?
Parent - By lkaufman (*****) Date 2010-07-12 18:13
The second. While Kasparov may not be fully prepared opening-wise for a world championship match, I don't think he would admit to being much off his prime as far as basic chess skills are concerned. For playing a computer, he would presumably use very different openings anyway from his usual ones (as he did against Deep Blue).
Parent - By Banned for Life (Gold) Date 2010-07-13 22:58
Perhaps he was just showing you how big a pile of money it would take to get him to consider the match...
Parent - - By Sesse (****) [fr] Date 2010-07-12 09:48
I'm sure there's been plenty of matches of R3 versus GMs in the comfort of their own homes, the results of which are probably the reason you won't see such a match in public ;-)

/* Steinar */
Parent - By staylor (***) [il] Date 2010-07-12 10:33
:wink::wink:
Parent - By Leto (***) [us] Date 2010-07-13 16:28
I think I read a chessbase article in which the current world champion Anand states he uses Rybka 3 for practice and when he was asked how he does against it he admitted it dominates him.
Parent - - By lkaufman (*****) Date 2010-07-12 16:39
In my view, the interesting matches we could play vs. top GMs include:
1. Draw and move handicap (we did this with Joel Benjamin, winning 6-2).
2. Castling odds (computer forfeits castling rights). Never been tried, but seems about right to me and perhaps more interesting than material odds.
3. Exchange handicap with no queenside castling. Rybka 3 lost by 1.5 to 2.5 (3 draws and a loss) to GM Milov (around 2700 then) at Exchange odds, but he could castle long while Rybka without the queen's rook could not. This rule change might make it fair for a 2700 level opponent.
4. c7 pawn handicap. We had lots of pawn odds games with both colors, but this pawn is the cleanest handicap in my opinion. The traditional f7 is too much for a strong GM, as is g7, the edge pawns too little, and the center pawns change the game too much. We only had a couple games with c7 odds.
Up Topic Rybka Support & Discussion / Rybka Discussion / Was there ever a match between Rybka 3 or 4 vs top GM?
1 2 Previous Next  

Powered by mwForum 2.27.4 © 1999-2012 Markus Wichitill