Author: Robert Hyatt
Date: 08:46:53 10/04/99
Go up one level in this thread
On October 04, 1999 at 10:43:00, Paulo Soares wrote: >On October 04, 1999 at 03:47:40, Ed Schröder wrote: > >>>Posted by Robert Hyatt on October 03, 1999 at 18:39:17: >> >>>Most books on tactics define 'sacrifice' as giving up material for some sort >>>of compensation (either positional or long term tactical chances). They >>>define 'combination' as a sequence of captures resulting in a gain of >>>material. >>> >>>in this game, my material score is always > 0 in the position you give, >>>meaning >>>that Crafty sees more material coming back to it than it gives up with the >>>original rook capture. That seems to better fit a 'combination'. >>> >>>I will agree that several books talk about 'queen sacrifices' when they are >>>not really sacrifices... as giving up a queen to win the opponent's king gets >>>more material back than it gives up... >>> >>>But I like the term 'combination' here... and usually use the term sacrifice >>>as in 'sacrificing the exchange'... after the rxc3 bxc3 type sac in many >>>Sicilian variations, black is 2 pawns (the exchange) down, yet gets lots >>>of compensation for that material, hopefully... Or in sacrificing a pawn >>>(such as playing a4-a3 to force your opponent to play bxa3 and end up with >>>three isolated pawns that you hope you can eventually win, and which you >>>_know_ can not be used to create a passed pawn... >>> >>>Mainly semantics... But if we call this a sacrifice, then I see one of these >>>every 2 games or so... IE QxR RxQ BxR, because after QxR RxQ I am definitely >>>down 4 pawns, but after the third move I am up a pawn... >>> >>> >>>Bob >> >>Then tell me the difference between a "positional sacrifice", and a "sacrifice". >> >>Anyway here is Jeroen Noomen's view (and analysis of the game in PGN) >> >>Ed >> >>---------------------------------------------------------------------------- >> >> >>Rebel Century - GM R. Sherbakov Monthly GM Challenge >>--------------------------------------------------------- >> >>Before the game we decided to go for 1. e4! No more quiet, positional chess, >>just open positions and play. Ruslan Sherbakov plays the Sicilian Defence, >>the Richter Rauzer variation in particular. Recently I have filled Rebel's >>opening book with lots of ideas in this variation, coming from Peter Wells's >>excellent book 'The complete Richter Rauzer'. We were not disappointed! >> >>1. e4 c5 2. Nf3 Nc6 3. d4 cxd4 4. Nxd4 Nf6 5. Nc3 d6 6. Bg5 e6 7. Qd2 Be7 >>8. O-O-O Nxd4 9. Qxd4 O-O 10. f4 Qa5 11. Bc4 Bd7 12. e5 dxe5 13. Qxe5! >>(An excellent choice. Everybody plays 13 fxe5, which is objectively stronger. >>But that move leads to a slightly better ending for White, and quite dull >>positions. Rebel's choice might be less strong from a theoretical point of view, >>but for a computer the resulting position is much easier to play. Furthermore, >>the queens stay on the board, leaving a lot to play for) >>13. ... Qb6 14. Qe2 Qc7?! (A very strange move. I don't know if this has >>been played before, but in the afore mentioned book by Peter Wells 14 ... >>Rad8 is given, leading to equal play. A game between Karpov and Kamsky - >>Buenos Aires 1994 - continued 15 Ne4 Nd5! and Black got excellent >>compensation for the sacrificed pawn. Besides, 15 f5? would be bad on >>account of 15 .... Qc5! Was Sherbakov afraid of shedding a pawn against a >>computer? Maybe.... But his move is clearly wrong, since now Rebel can >> >>continue....) >>15. f5! (... as 15 ... Qc5 is now impossible: 16 Bxf6 wins a piece) 15. >>... h6? >>(And this is already a big mistake) 16. Rxd7! (Whoops! Clearly not the way >>to handle a computer....) 16. ... Qxd7 17. fxe6 Qc7 18. Bxf6 Bxf6 19. Nd5 Bg5+ >>20. Kb1 Qd6 21. exf7+ Kh8 22. h4 (White has two pawns for the exchange, a >>wonderful bishop on c4, a passed pawn on f7, a giant knight on d5 and also >>the black pieces are not cooperating very well. It is clear who has the >>advantage here!) >>22. ... b5 23. hxg5 bxc4 24. gxh6!! (Fantastic! Rebel sacrifices a knight, >>in order to get at the black king) 24. ... Qxd5? (The only way to keep on >>playing was to try 24 ... g6 25 Qf3) >>25. hxg7+ Kxg7 26. Qg4+ Kf6 27. Rf1+ Ke7 28. Rf5! (This quiet move is the >>big point of the knight sac. The rook joins the attack with decisive effect) >>28. ... Qe6 29. Qh4+ Kd7 30. Qd4+ Kc7 31. Qc5+ Kb7 32. Qb4+! (A wonderful >>queen manoeuvre, after which black is unable to avoid the loss of his queen) >>32. ... Kc7 33. Rc5+ Qc6 (Even worse is Kd7 34 Qb7+) 34. Qxc4 Qxc5 35. >>Qxc5+ (The >>rest is easy. Rebel mops it up without any trouble) 35. ... Kd7 36. c4 Ke6 >>37. Qd5+ Kf6 38. Qb7 Kg7 39. c5 Rad8 40. Qxa7 Rxf7 41. Qa4 Rd2 42. Qg4+ Kf8 >>43. a4 Rff2 44. Qb4 Ke8 45. g4 Rd1+ 46. Ka2 Rff1 47. c6 Ra1+ >>48. Kb3 Rf3+ 49. Kc2 Rf2+ 50. Kd3 and Sherbakov resigned. 1 - 0 >> >>Congrats to Ed and especially Rebel - of course - for this great >>performance. A wonderful game, which I enjoyed very much. And I was pleased >>to see that the preparation for this match worked so well. Until next time! >> >>Jeroen Noomen > >Excellent game, in which the opening was fundamental. I apologize for >insisting in speaking on that, but I was very displeased with the openings >that Rebel was playing, mainly against GM Hoffman, and I found difficult >that Jeroen Noomen got to solve short term this problem. The one that more >leaves me satisfied it is that 13.Qxe5! shows that a theorical inferior move, >can be better for a program in a game against a human. I thought that a GM >could play a theorical inferior move against a program, obtaining advantage >for the type of resulting position, but I never imagined the opposite. >I know that it is still early to arrive to definitive conclusions, >but what left me more satisfied went the form with that Rebel Team reacted >to the adversities. > >Paulo Your point is very interesting... because the most disappointing thing in this (and every other computer vs GM game I see) is that the books have to be tuned _by hand_. Which means that the computers are always going to be one step _behind_ the GM players in opening preparation. That's a bad place to be. I hope that some things going on (IE like the C.A.P. data) might be used to let us automate this opening book move selection problem in a reliable way, without requiring human intervention before every game. So far, that seems difficult to do, based on what everyone has done/is doing to date. Doesn't matter how strong you are tactically if you let the GM dictate the opening...
This page took 0 seconds to execute
Last modified: Thu, 15 Apr 21 08:11:13 -0700
Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.