Author: Robert Hyatt
Date: 13:36:10 09/18/01
Go up one level in this thread
On September 18, 2001 at 14:38:01, Ed Schröder wrote: >On September 18, 2001 at 13:46:58, Dann Corbit wrote: > >>On September 18, 2001 at 13:02:02, Ed Schröder wrote: >> >>>On September 18, 2001 at 12:22:20, Robert Hyatt wrote: >>> >>>>On September 18, 2001 at 10:53:39, Uri Blass wrote: >>>> >>>>>On September 18, 2001 at 10:45:15, José Carlos wrote: >>>>> >>>>>>On September 18, 2001 at 09:56:32, Ed Schröder wrote: >>>>>> >>>>>>>On September 18, 2001 at 09:40:26, Eduard Nemeth wrote: >>>>>>> >>>>>>>>First the game: >>>>>>>> >>>>>>>>[Event "TCh-YUG Herceg Novi YUG"] >>>>>>>>[Site "?"] >>>>>>>>[Date "2001.??.??"] >>>>>>>>[Round "?"] >>>>>>>>[White "Vukic,M"] >>>>>>>>[Black "Marinkovic,I"] >>>>>>>>[WhiteElo "2470"] >>>>>>>>[BlackElo "2440"] >>>>>>>>[ECO "A30"] >>>>>>>>[Result "1-0"] >>>>>>>> >>>>>>>>1. Nf3 Nf6 2. g3 b6 3. Bg2 Bb7 4. O-O g6 5. c4 c5 6. Nc3 >>>>>>>>Bg7 7. d4 cxd4 8. Qxd4 d6 9. Be3 Nbd7 10. Rac1 Rc8 11. b3 >>>>>>>>O-O 12. Qd2 Nc5 13. Rfd1 Nce4 14. Nxe4 Nxe4 15. Qe1 Qd7 >>>>>>>>16. Nd4 Nf6 17. Nf3 Ne4 18. Qb4 Nc3 19. Rd2 Ne4 20. Rd3 Nc5 >>>>>>>>21. Rdd1 Rfd8 22. Ne1 h5 23. Bxc5 Rxc5 24. Nd3 Bxg2 >>>>>>>>25. Kxg2 Qb7+ 26. f3 Rc7 27. Nf2 Rdc8 28. Qd2 b5 29. cxb5 >>>>>>>>Qxb5 30. Kf1 Bc3 31. Qd3 Qa5 32. Rc2 Bf6 33. Rdc1 Rxc2 >>>>>>>>34. Rxc2 Rxc2 35. Qxc2 d5 36. e3 Kg7 37. Nd3 e6 38. Ke2 Qa3 >>>>>>>>39. f4 Kg8 40. h3 Kg7 41. Kf3 Qd6 42. g4 hxg4+ 43. hxg4 e5 >>>>>>>>44. fxe5 Bxe5 45. Nxe5 Qxe5 46. Qc5 Qf6+ 47. Kg3 Qe5+ >>>>>>>>48. Kh3 a6 49. Qd4 Kf6 50. b4 Ke6 51. a4 Qc7 52. g5 Qc2 >>>>>>>>53. Qf6+ Kd7 54. Qxf7+ Kd6 55. Qf6+ Kd7 56. Qd4 Ke6 57. Kg3 >>>>>>>>Qb1 58. Qb6+ Ke7 59. Qb7+ Kf8 60. Qxd5 Qxb4 61. Qd8+ Kg7 >>>>>>>>62. Qd4+ 1-0 >>>>>>>> >>>>>>>>Position after move 61. Kg7: >>>>>>>> >>>>>>>>[D]3Q4/6k1/p5p1/6P1/Pq6/4P1K1/8/8 w - - >>>>>>>> >>>>>>>>After 62. Qf6+ is the win not easy. But after 62. Qd4+ is the game easy to win, >>>>>>>>and the game end! For any programs ist the move Qd4+ not easy to find. >>>>>>>> >>>>>>>>:-( >>>>>>> >>>>>>> >>>>>>>I am sure all Rebel programs find this instantly. A simple chess rule says: >>>>>>>"one pawn up in a pawn ending wins". >>>>>>> >>>>>>>Ed >>>>>> >>>>>> Are you serious Ed? In this position, it's a clear win, but in most pawn >>>>>>endgames, you have to calculate a lot, or otherwise know some rules (which names >>>>>>I don't know how to translate from spanish to english) to figure if it's a win >>>>>>or not. >>>>>> >>>>>> Maybe you were simply joking, and I didn't catch it... :) >>>>> >>>>>In most pawn endgames one pawn advantage is enough to win the game. >>>>>It is not something that always happens but computer can safely evaluate >>>>>+2 for positions with one pawn advnatage if they do not see a positional >>>>>advantage for the side that has less pawns. >>>>> >>>>>Uri >>>> >>>> >>>>I believe that if humans discover this "rule" in Rebel, it will die a slow >>>>and painful death on ICC. I can create _many_ positions where the rule is >>>>wrong. And if I can do it, GMs will do it. Happened to me many times. I >>>>was burned by going into simple king and pawn endings where I was a pawn up, >>>>but my opponent had a distant majority. And he would win. And win. >>> >>> >>>The rule is valid for 90% of all positions and if not search will handle >>>the remaining 10% close to zero. I don't give a +5 score for being a pawn >>>up in a pawn ending but a reasonable score update enough to force a >>>knight/bishop/rook/queen ending to a won pawn ending or to avoid to enter >>>a lost pawn ending position. The rule in Rebel is there for more than 10-12 >>>years and in that time I have seen only one exception. >> >>It certainly seems like a very valuable addition to include before any result >>can be born out by search. In other words, of all positions seem about equal, a >>bonus for an extra pawn seems to be a pretty good idea. If there is some >>problem down the line, it will be born out by search (if the time left is long >>enough) but if you have to guess -- it seems much better to me to guess for the >>better formation. > > >Yes. > >When I implemented this rule 10-12 years ago I did not expect much. However >the contrary was true as it solved many wrong exchanges to a lost pawn ending. > >Ed The simple rule "when ahead in material, exchange pieces but not pawns, and when behind in material, exchange pawns but not pieces" seems to work perfectly in the context you are describing. The problem is that there are _many_ lost positions where one side is a pawn up. And such rules turn into miserable losses rather than stubborn/difficult losses.
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.