Not logged inRybka Chess Community Forum
Up Topic Rybka Support & Discussion / Rybka Discussion / Benjamin via Rybka. Game 8. Final result is 6 - 2 for Rybka
- - By Victor Zakharov (*****) [ru] Date 2008-01-07 00:02 Edited 2008-01-07 00:46
Game 8 looks like most interesting in the match (after game 4).

Congratulations to Vas and Larry!

[White "Benjamin"]
[Black "Rybka"]
[WhiteElo "2575"]
[BlackElo "3100"]
[Result "0-1"]
[GameID "215"]
[UniqID "36455"]
[WhiteClock "0:06:56"]
[BlackClock "0:58:21"]
[Stamp "1639"]
[LastMoves "32.Kg5 Rxd6 -11.52"]

1.e4 e5 2.Nf3 d6 3.d4 exd4 4.Nxd4 Nf6 5.Nc3 Be7 6.g3 Nc6 7.Bg2 O-O 8.O-O Bg4 9.f3 Nxd4 10.Qxd4 Be6 11.f4 c5 12.Qd2 Re8 13.b3 c4 14.Kh1 Rc8 15.Bb2 Qa5 16.Qd4 Bf8 17.Rae1 a6 18.Bf3 Bh3 19.Bg2 Qh5 20.Qd2 b5 21.a3 d5 22.exd5 Ng4 23.b4 Bxg2+ 24.Qxg2 Ne3 25.Qf3 Nxf1 26.Qxh5 Rxe1 27.Kg2 Rce8 28.Qf3 Ne3+ 29.Kh3 f5 30.d6 Re6 31.Kh4 Rh6+ 32.Kg5 Rxd6  0-1
Parent - By Felix Kling (Gold) [de] Date 2008-01-07 05:20
Indeed. Maybe I should start playing the Philidor exchange variation again :)
Parent - - By lkaufman (*****) Date 2008-01-07 06:24
Thanks to Sergey Pligin whose games and analysis persuaded me to incorporate the Philidor defense into the book for this match and to play it three times.
Parent - - By Felix Kling (Gold) [de] Date 2008-01-07 11:41
Did Sergey Pligin write a book or where do you have his analysis from? I would be interested in it :)
Parent - - By lkaufman (*****) Date 2008-01-07 15:49
     I believe he is writing a book on the opening, but I received the games/analysis privately with condition of not publishing them or giving them out. Probably many of the games are available in some database of correspondence games.
Parent - By Felix Kling (Gold) [de] Date 2008-01-07 16:16
mmh, no chance to make an exception inside the rybka team? :)
There's a book in German about the philidor exchange variation (by IM Seel, Geheimwaffe Philidor), but some of the analysis there didn't make me happy. I also took part in a seminar by GM Hickl, but he gives some dangerous (dubious?!) lines for black (espacially against g3) which also don't completely satisfy me...
Then I have also "the Phlidor files" where some good ways to play that line for white are shown but a lot of open questions exists how black should play...

So I'm really searching for analysis in that specific line (btw., I enter it via 1.e4 d6 2.d4 Nf6 3.Nc3 e5 4.Nf3 exd and so on)!
Parent - By FWCC (***) [us] Date 2008-01-07 19:36
WOW!! I missed all of the fun this weekend as I was in a tournament.Larry I played in the Baltimore Open and of course you was busy over the weekend and was not there.I see Joel lost 6-2 that is AMAZING! You guys have truly created a MONSTER!!Rybka is virtually unbeatable by human standards.Hats off to Joel for playing her.
Parent - By Hetman (*****) Date 2008-01-07 18:04
I think that 1st thanks could have gone to the Philidor ;-) .
Rgds
Hetman
Parent - - By Hetman (*****) Date 2008-01-07 18:02
The chronological graph of the match has shown that human got tired.
Rgds
Hetman
Parent - - By turbojuice1122 (Gold) [us] Date 2008-01-07 18:20
...or that some adjustments were made in both the openings and in the parameters in the program based on the learning of Larry and the rest of the team of what kinds of strategies could successfully be employed against Rybka.
Parent - - By Hetman (*****) Date 2008-01-07 18:25
It would mean that chances were not equal, human did not have seconds to analyse Rybka adjustments and prepare adaptation of the play strategy.
It would show that not Rybka has won but they seconds ;-).
Parent - - By turbojuice1122 (Gold) [us] Date 2008-01-07 18:32
Well, I recall that in past situations, the challenger, in this case Joel, has been allowed to use Rybka to analyze between the games.  Meanwhile, the changes are probably mainly being applied by Larry.  Thus, it's entirely fair: Rybka has a second (Larry) and Joel has a second (Rybka).  Also, Joel has another second that Rybka doesn't have: himself.  Rybka is (currently) unable to completely change its approach on its own.
Parent - - By Hetman (*****) Date 2008-01-08 18:43
Yes, but it was not  the match Rybka - Human but R + H vs H. :-(
Parent - - By turbojuice1122 (Gold) [us] Date 2008-01-08 19:17
Not at all--see Larry's statements here: http://rybkaforum.net/cgi-bin/rybkaforum/topic_show.pl?dln=32700;pid=38820#pid38820

This really is entirely fair, and I think that Joel's changes based on analysis would have been more significant than Rybka's changes.
Parent - - By Hetman (*****) Date 2008-01-09 19:36
? :-)
Parent - - By turbojuice1122 (Gold) [us] Date 2008-01-09 20:19
i.e. Rybka wasn't really being helped by a human in between the games, though Joel was being helped by Rybka between the rounds.
Parent - - By Hetman (*****) Date 2008-01-10 21:25
Rybka has been helped by human between the games.
Parent - - By turbojuice1122 (Gold) [us] Date 2008-01-10 22:43
Well, that's really the point I'm making--Rybka has not been helped by human between the games--at least, not in this most recent match.  The only thing that Larry did was change the contempt value, and the reason for doing that was simply to find out if doing so wouldn't really give bad results.
Parent - - By Hetman (*****) Date 2008-01-11 15:25
It does not mean that she was helped ? ;-)
Parent - - By turbojuice1122 (Gold) [us] Date 2008-01-11 16:45
Well, "helped" and helped are two different things.  Rybka's results weren't really improved due to this.  Also, Joel uses Rybka to analyze between games.
Parent - - By Hetman (*****) Date 2008-01-11 21:48
Results were not improved ? ;-)  It was 2:2 at the start and 6:2 at the end.
Parent - - By turbojuice1122 (Gold) [us] Date 2008-01-11 21:58
You can believe whatever you like, of course, but (a) this sort of finish could easily be expected of a GM in an intense and tiring match (two games per day against such a monster!), especially since computer chess matches have tended to have this exact pattern, and (b) when Larry indicates that there wasn't improvement, I believe that statement.  Can you point to any places where Rybka played moves that probably wouldn't have been played by the "Rybkas" in previous games?
Parent - By Hetman (*****) Date 2008-01-11 22:12
It is the question to  JB .
Parent - - By Uly (Gold) [mx] Date 2008-01-11 22:41

> You can believe whatever you like, of course


I believe that the results would have been very different if they never changed anything about Rybka, maybe the GM could have won (Though, I'm including opening book change in this.)

If "Rybka's results weren't really improved due to this" then why to do these changes on Rybka in the first place?
Parent - - By turbojuice1122 (Gold) [us] Date 2008-01-11 23:33
Larry said, I think--to make sure that making the changes wouldn't cause Rybka to perform worse.  Obviously if he had set the value at, say, 500, then the performance could easily be jeopardized significantly.  He is trying to see how high he can safely set the value.
Parent - - By Vasik Rajlich (Silver) [hu] Date 2008-01-12 09:54
Actually, it was no big plan. We've been busy on Rybka 3 and didn't take a look at all of this anti-human stuff until a couple of days before the match.

Unlike the previous Benjamin match, we did test before the first game :)

Vas
Parent - - By garyf919 (**) [us] Date 2008-01-12 14:47
Question for Vas:

Would it be possible to give Rybka 3 some chess personalities based on how the contempt variable
is set, but not have the user set the value itself, but rather pick a personality, such as aggressive,
solid, passive, anti-human, default, etc., and then this selection would translate into the GUI
setting the contempt for you to have Rybka play in this fashion? This would allow you and Larry
to "set the best values for the user based on your experience" vs the user picking these values
themselves. You could also still allow the current method of the user setting the values themselves.
I think this could be a nice little feature if there is time to implement it before release.
Parent - By Uly (Gold) [mx] Date 2008-01-13 00:17

> solid


This would be the default setting :)

> passive


A correct implementation of a passive setting would basically lose against anything, but it would be very boring to see it play.

Aside from this, I agree about having some personalities, and I'd want to be able to create my own. Material weighting should also be included, I think.
Parent - By Vasik Rajlich (Silver) [hu] Date 2008-01-14 21:36
We'd like to do this at some point, I agree that it would be nice for users. It's not clear yet that there will be enough time before Rybka 3. 90% of this work would be Larry's.

Vas
Parent - - By Uly (Gold) [mx] Date 2008-01-09 22:17
This was posted over at CCC by Dann Corbit:

"This is really interesting, as it shows some sort of problem with either the search or the eval of Rybka.
Given the position after moving 23. b4 and black has responded Bxg2+, Rybka still does not see the trouble:

2r1rbk1/5ppp/p7/1p1P3q/1Pp2Pn1/P1N3P1/1BPQ2bP/4RR1K w - - bm Qxg2;



    Searching move: Qd2xg2
    Best move (Rybkav2.3.2a.w32): Qd2xg2
    identical moves! Found in: 00:00
      5  00:00           494  505.856  -0.41  Qd2xg2 Ng4e3
      6  00:00         1.050  63.247  -0.45  Qd2xg2 Ng4e3 Qg2f3
      7  00:00         2.622  83.904  -0.59  Qd2xg2 Ng4e3 Qg2f3 Qh5xf3+ Rf1xf3 Ne3xc2
      8  00:00         5.665  73.429  -0.66  Qd2xg2 Ng4e3 Qg2f3 Qh5xf3+ Rf1xf3 Ne3xc2 Re1xe8 Rc8xe8 Rf3f2
      9  00:00         8.256  76.163  -0.57  Qd2xg2 Ng4e3 Qg2f3 Qh5xf3+ Rf1xf3 Ne3xc2 Re1xe8 Rc8xe8 Rf3f2 Nc2d4
     10  00:00        12.071  71.449  -0.51  Qd2xg2 Ng4e3 Qg2f3 Qh5xf3+ Rf1xf3 Ne3xc2 Re1xe8 Rc8xe8 Rf3f2 Nc2d4 Rf2d2
     11  00:01        19.632  61.103  -0.42  Qd2xg2 Ng4e3 Qg2f3 Qh5xf3+ Rf1xf3 Ne3xc2 Re1xe8 Rc8xe8 Rf3f2 Nc2d4 Rf2d2 Nd4f5
     12  00:01        32.624  59.232  -0.42  Qd2xg2 Ng4e3 Qg2f3 Qh5xf3+ Rf1xf3 Ne3xc2 Re1xe8 Rc8xe8 Rf3f2 Nc2d4 Rf2d2 Nd4f5 Kh1g1
     13  00:01        57.046  58.356  -0.43  Qd2xg2 Ng4e3 Qg2f3 Qh5xf3+ Rf1xf3 Ne3xc2 Re1xe8 Rc8xe8 Rf3f2 Nc2d4 Rf2d2 Nd4f5 Kh1g1 Bf8d6
     14  00:02       112.365  57.502  -0.37  Qd2xg2 Ng4e3 Qg2f3 Qh5xf3+ Rf1xf3 Ne3xc2 Re1xe8 Rc8xe8 Rf3f2 Nc2e1 Rf2e2 Re8e7 Kh1g1 Ne1d3
     15  00:04       187.649  56.135  -0.37  Qd2xg2 Ng4e3 Qg2f3 Qh5xf3+ Rf1xf3 Ne3xc2 Re1xe8 Rc8xe8 Rf3f2 Nc2e1 Rf2e2 Re8e7 Kh1g1 Ne1d3
     16  00:06       316.024  54.783  -0.33  Qd2xg2 Ng4e3 Qg2f3 Qh5xf3+ Rf1xf3 Ne3xc2 Re1xe8 Rc8xe8 Rf3f2 Nc2e1 Rf2e2 Re8e7 Kh1g1 Ne1d3
     17  00:11       560.709  53.876  -0.29  Qd2xg2 Ng4e3 Qg2f3 Qh5xf3+ Rf1xf3 Ne3xc2 Re1xe8 Rc8xe8 Rf3f2 Nc2e1 Rf2e2 Re8e7 Kh1g1 Ne1d3
     18  00:24     1.262.156  53.883  -0.28  Qd2xg2 Ng4e3 Qg2f3 Qh5xf3+ Rf1xf3 Ne3xc2 Re1xe8 Rc8xe8 Rf3f2 Nc2e1 Rf2e2 Re8e7 Kh1g1 Ne1d3
     19  00:38     2.001.084  54.663  -0.33  Qd2xg2 Ng4e3 Qg2f3 Qh5xf3+ Rf1xf3 Ne3xc2 Re1xe8 Rc8xe8 Rf3f2 Nc2e1 Rf2e2 Re8e7 Kh1g1 f7f5
     20  01:13     3.948.433  55.659  -0.37  Qd2xg2 Ng4e3 Qg2f3 Qh5xf3+ Rf1xf3 Ne3xc2 Re1xe8 Rc8xe8 Rf3f2 Nc2e1 Rf2e2 Re8e7 Kh1g1 f7f5
     21  02:13     7.859.223  60.637  -0.54  Qd2xg2 Ng4e3 Qg2f3 Qh5xf3+ Rf1xf3 Ne3xc2 Re1xe8 Rc8xe8 Rf3f2 Nc2e1 Rf2e2 Re8xe2 Nc3xe2 g7g6
     22  03:52    14.692.426  64.923  -0.59  Qd2xg2 Ng4e3 Qg2f3 Qh5xf3+ Rf1xf3 Ne3xc2 Re1xe8 Rc8xe8 Rf3f2 Nc2e1 Rf2e2 Re8xe2 Nc3xe2 Ne1c2
   1/8/2008 7:42:45 PM, Time for this analysis: 00:06:25, Rated time: 19:46


Even more amazing, look at this (we are at 27. about to move Kg2):

2r2bk1/5ppp/p7/1p1P3Q/1Pp2P2/P1N3P1/1BP4P/4rn1K w - - bm Kg2;


If you look at the pv, Rybka does not see the devastating reply f5! that is coming at move 29:"

   
    Searching move: Kh1-g2
    Best move (Rybkav2.3.2a.w32): Kh1-g2
    identical moves! Found in: 00:00
      5  00:00         1.591  95.834  -0.43  Kh1g2 Nf1e3+ Kg2f2
      6  00:00         2.807  89.824  -0.49  Kh1g2 Nf1e3+ Kg2f2 Ne3xc2
      7  00:00         7.620  70.935  -0.19  Kh1g2 Nf1e3+ Kg2h3 Rc8d8 Nc3e4
      8  00:00        10.182  66.409  -0.24  Kh1g2 Nf1e3+ Kg2h3 Rc8d8 Nc3e4 h7h6
      9  00:01        14.624  59.661  -0.25  Kh1g2 Nf1e3+ Kg2h3 Rc8d8 Nc3e4 h7h6 Bb2c3
     10  00:01        33.442  60.717  -0.12  Kh1g2 Nf1e3+ Kg2h3 Rc8e8 Qh5f3 Re1f1 Qf3e2 Rf1g1
     11  00:01        66.326  63.832  -0.37  Kh1g2 Nf1e3+ Kg2h3 Rc8e8 Qh5f3 Re8d8 Nc3e2 Ne3xd5 Bb2d4
     12  00:02       107.154  63.794  -0.41  Kh1g2 Nf1e3+ Kg2h3 Rc8e8 Qh5f3 Re8d8 Nc3e2 Ne3xd5 Bb2d4 Bf8e7
     13  00:03       173.644  60.831  -0.21  Kh1g2 Nf1e3+ Kg2h3 Rc8e8 Qh5f3 Re8d8 Nc3e4 Ne3xd5 Bb2d4 Bf8e7 Ne4c3
     14  00:10       625.785  63.971  -0.58  Kh1g2 Nf1e3+ Kg2h3 Rc8e8 Qh5g5 h7h6 Qg5h5 Ne3xc2 Qh5f5 Nc2d4 Qf5d7 Re1e7
     15  00:14       815.009  62.834  -0.58  Kh1g2 Nf1e3+ Kg2h3 Rc8e8 Qh5g5 h7h6 Qg5h5 Bf8d6 Nc3e2 Ne3xc2 Ne2d4 Nc2xd4 Bb2xd4 Re1e4
     16  00:21     1.352.454  65.505  -0.58  Kh1g2 Nf1e3+ Kg2h3 Rc8e8 Qh5g5 h7h6 Qg5h5 Bf8d6 Nc3e2 Ne3xc2 Ne2d4 Nc2xd4 Bb2xd4 Re1e4
     17  01:28     5.438.147  63.391  -0.73  Kh1g2 Rc8e8 Kg2h3 Re8d8 Qh5g5 Bf8e7 Qg5h5 Rd8d6 Kh3g2 Nf1e3+ Kg2f2 Ne3xc2 Qh5f3 Rd6h6
     18  01:44     6.468.552  64.133  -0.67  Kh1g2 Rc8e8 Kg2h3 Re8d8 Qh5g5 Bf8e7 Qg5h5 Nf1e3 Nc3e2 Ne3xd5 Qh5f3 Be7f6 Bb2xf6 Nd5xf6
     19  02:28     9.616.749  66.811  -0.75  Kh1g2 Rc8e8 Kg2h3 Re8d8 Qh5g5 Bf8e7 Qg5h5 Nf1e3 Nc3e2 Ne3xd5 Qh5f3 Be7f6 Bb2xf6 Nd5xf6
     20  03:41    14.863.965  68.934  -0.78  Kh1g2 Rc8e8 Kg2h3 Re8d8 Qh5g5 Rd8d6 Kh3g2 Nf1e3+ Kg2f2 Ne3xc2 f4f5 Re1e8 Qg5d2 Nc2e3
   1/8/2008 8:06:07 PM, Time for this analysis: 00:06:25, Rated time: 00:00
Parent - By Victor Zakharov (*****) [ru] Date 2008-01-10 08:08
Sorry, what trouble Rybka should see?

You put Rybka analysis but but don't say what is wrong with it.
Up Topic Rybka Support & Discussion / Rybka Discussion / Benjamin via Rybka. Game 8. Final result is 6 - 2 for Rybka

Powered by mwForum 2.27.4 © 1999-2012 Markus Wichitill