Not logged inRybka Chess Community Forum
Up Topic Rybka Support & Discussion / Rybka Discussion / Rybka beat the baron
- - By Uri Blass (*****) Date 2007-06-16 11:44 Edited 2007-06-16 12:25
[Event "15th World Computer Chess Championship"]
[Site "Amsterdam, The Netherlands"]
[Date "2007.06.16"]
[Round "7"]
[White "Rybka"]
[Black "The Baron"]
[Result "1-0"]

1. e4 c6 2. d4 d5 3. e5 Bf5 4. Nf3 e6 5. Be2 Nd7 6. O-O Ne7 7. Nbd2 h6 8.
c3 c5 9. dxc5 Nxc5 10. Nd4 Bg6 11. N2b3 Nd7 12. f4 Nf5 13. Bd3 Nxd4 14.
cxd4 Bxd3 15. Qxd3 g6 16. Bd2 Nb6 17. Na5 Qd7 18. Rac1 Be7 19. g4 Rc8 20.
Rxc8+ Qxc8 21. b3 a6 22. Kh1 h5 23. f5 gxf5 24. gxf5 Kd7 25. Rg1 h4 26. h3
Kd8 27. Rc1 Qd7 28. fxe6 fxe6 29. Qg6 Re8 30. Nxb7+ Qxb7 31. Qxe6 Bf8 32.
Qf6+ Be7 33. Qg6 Nc4 34. bxc4 dxc4+ 35. Kg1 Qd5 36. Qb6+ Kc8 37. Qxa6+ Qb7
38. Qe6+ Kb8 39. Qg6 Qc8 40. Rxc4 Qxc4 41. Qxe8+ Qc8 42. Qxe7
1-0

other results

GridChess 1  micro-Max 0 
Shredder 1  Diep 0  
Loop 1  Jonny 0 
Zappa 1  IsiChess 0
The King  0 Deep Sjeng 1

situation now:
1)Rybka  USA  6.0 7 17.5 12.75 
2)Zappa  TUR  5.5 7 21.5 14.50 
3-5)Shredder  DEU  5.0 7 24.5 17.00 
3-5)GridChess  DEU Cluster 5.0 7 19.0 11.75 
3-5)Deep Sjeng  BEL 2 x Intel Xeon 5355 5.0 7 18.5 10.75 
Parent - - By insipid (**) Date 2007-06-16 14:39
It's a real shame Rybka didn't have the proper 6 man TB for the Zappa game.  Had that been a win, Rybka would be at 6.5/7 and Zappa would be in a tie for second at 5.0/7 and the tournament would already be over.  As it is I still like Rybka's chances but sharing first with Zappa also seems like a likely outcome. 
Parent - - By JhorAVi (***) Date 2007-06-16 15:08
If Rybka had 6 man TB during his game with Zappa, Rybka might have LOST!

Because the action spent in installing the TB will result to a different situation..  Rybkas operators might have arrived in the tournament earlier or latter..  It will change everything including the OPENING phase of the game.. the timing of random generator code will change too and Zappa will choose another opening as well :)

It will result to a different MiddleGame and the Endgame you saw will never happen!

Rybka could even have lost had it had TB.

But of course it could have won too :)
Parent - By turbojuice1122 (Gold) Date 2007-06-16 15:38
So basically, the "expectation value" for the score of the Rybka-Zappa game, had the tablebases been completely installed, is closer to 1.0 than to 0.0, but with a large standard deviation...
Parent - - By Nelson Hernandez (Gold) Date 2007-06-16 17:23
I just hate it when people bring up parallel universes and time continuums.  If you don't mind, I will stay in my linear universe.  Rybka would have won with TBs, period!!
Parent - - By Uri Blass (*****) Date 2007-06-16 17:26
I dislike when people give lack of tablebase as an excuse.
Rybka had a simple win that every strong human could win and it was not some complicated endgame.
Rybka without bugs could win without problems and older version of rybka could play better.

The KRP vs KNP tablebases are not important tablebases and not having bugs is more important.

Uri
Parent - - By Nelson Hernandez (Gold) Date 2007-06-16 18:13
According to my database (consisting of engine games only, millions of games) KRPKNP is the 18th-most frequently occurring combination, occurring in 0.53% of all games that reduce to six pieces.  But more to the point, this particular combination produced 30.7% draws, nearly duplicate of a random sample of all games.  Meaning: this is one of the trickiest and most important endgames to have, because finding the right moves could be challenging for a computer without EGTBs (compare to KQPPKP, which occurs more often but is drawn 2.8% of the time).  In fact, I would rank KRPKNP as a must-have tablebase, certainly one you would have within the first 100GB of 6-man downloads in spite of its large (11.54 GB) size.  Honestly, all this is irrefutable to me and to anyone who has done a similar statistical compilation.
Parent - - By Harvey Williamson (*****) Date 2007-06-16 18:27
This ending is not on the CB 9 DVD turbo 3 set. Anyone know if its on the similar endgame product sold by Convekta?

I would bet  that none of the competitors at WCCC had this tb?

I use the CB turbo 3 set. All 5  man some 6 - I really doubt any use much more than this 40 gb set?

Best Wishes,

Harvey
Parent - - By davidwhite (***) Date 2007-06-16 21:38
Harvey,
            the ending is also NOT on Convekta's latest 6 DVD Nalimov Tablebase set(50 Gbs total)

            One entire DVD(8.5 Gbs) is used to deal with KNPKBP alone !
             In fact,the above tb is one of only THREE K,piece,pawn/K,piece,pawn positions dealt with by Convekta in the entire set.
             The other two are KNPKNP and KBPKBP.

             I'm,also,willing to bet that none of the competitors at WCCC had the necessary tb.

             Best Regards,
             David
Parent - By insipid (**) Date 2007-06-17 01:23
Well they all should, shouldn't they?  50 GB is not a huge number by any stretch when there are terrabyte hard drives.  It's the world computer chess championship .. if you're going to spend effort to get the best hardware why wouldn't you have the best software as well? 
Parent - - By Uri Blass (*****) Date 2007-06-16 18:29
I think that krp vs knp should be usually easy to win(assuming that the stronger side do not lose material)
krp vs Kbp seems to be harder than krp vs knp.

I guess that in most of the draws in your database the draw is not result of a blunder but of a situation when the stronger side lose a pawn or lose a rook for a pawn or a rook for a knight.

Uri
Parent - By Vempele (Silver) Date 2007-06-16 18:40

> I think that krp vs knp should be usually easy to win(assuming that the stronger side do not lose material)


There are fortresses. For example, if Rybka-Zappa had been shifted one rank up, many (all?) of the positions in the game would have been drawn. Giving a penalty when the pawns are close (on adjacent files/on the same file, facing each other as close as possible) and recognizing the most common fortresses would solve some problems, I suppose.

But it'd still be something that would only affect less than 0.5% of all games. The TB is certainly worth it if you have that much extra disk space sitting around, but coding it in the eval probably isn't.
Parent - - By h1a8 (***) Date 2007-06-17 02:44
Why don't you set up the exact krpknp position where rybka drew zappa and then play that position against rybka and then post the game.
Because I'm having a hard time figuring out how to win this type of ending (without tablebase info).
Parent - - By Uri Blass (*****) Date 2007-06-17 05:15
I am going to try to do it later but my impression is that the plan of white to win a pawn (kg1 kh2 kg3) is a simple plan and earlier it seems even simpler to win the f pawn in order to have a passed pawn and not to win the d pawn.

Uri
Parent - By Uri Blass (*****) Date 2007-06-23 20:03
I have to admit that it is not so simple to win and when it seems that rybka played with no plan when the tablebase moves encouraged pushing the king toward the pawn it seems that finding the right plan to do it correctly is not so easy.

Uri
Parent - - By Alkelele (****) Date 2007-06-16 18:30
Yes, I completely agree. I would rank this endgame up there with endings like krppkr or krpkrp, or right after. It's not only that the right moves are difficult to find -- it's also that evaluation is impossible to get right for an engine on its own. Being able to tell from afar if such an ending is 1-0 or ½-½ is very valuable.
Parent - By Uri Blass (*****) Date 2007-06-16 18:38
I disagree
I believe that in big majority of the cases KRP vs KNP is a win if you can keep the material for few moves.
Things are less clear with bishop but knight against rook is weaker than a bishop.

Uri
Parent - By Vasik Rajlich (Silver) Date 2007-06-18 07:19
Thanks for the stats. I find it quite amazing that 30% of KRPKNP endings are drawn. Perhaps many involve the pawn from the stronger side quickly being lost, or the pawn of the side with the knight immediately requiring a rook sac.

Vas
Up Topic Rybka Support & Discussion / Rybka Discussion / Rybka beat the baron

Powered by mwForum 2.27.4 © 1999-2012 Markus Wichitill