![[us]](/mwf/v12/flags/us.png "United States")
Based on the results so far (Ehlvest-Rybka)I may have been correct in saying that to play even with Rybka,one should play HER at knight odds.A knight odd game will even the playing field maybe only slightly.I am looking forward to such a match perhaps with a GM of around 2700.This should be more interesting.
What do you need a GM for? We have a guy here who says he can beat Rybka at Knight odds. Mike is bragging about it so I say give him a chance.
![[il]](/mwf/v12/flags/il.png "Israel")
I agree that you do not need GM's
It is better to start with weaker players who think that they can do it and only if they fail to give the task for stronger players.
I am not sure if I can do it but I think that with knight odd even I have chances and my fide rating is only 1975.
Uri
It is better to start with weaker players who think that they can do it and only if they fail to give the task for stronger players.
I am not sure if I can do it but I think that with knight odd even I have chances and my fide rating is only 1975.
Uri
I disagree
I think that knight is too much advantage to give to a GM against rybka except blitz.
Uri
I think that knight is too much advantage to give to a GM against rybka except blitz.
Uri
Well it is just a thought here.Rybka is incredibly strong and I think I did hear a couple of people say they can beat HER at knight odds but I am unsure-I don't think so.The only way to find out is to give it a try.Maybe knight odds are too large an advantage to the human but lets find out.Based on HER current results I think Rybka can handle it.Rybka is most likely the strongest playing entity on the planet.Vas has created a monster!Look out for 2.3.2.
OK, I managed to beat Rybka giving knight odds. In the 4th attempt :P They were 5 minute games. [I do believe that I can consistently win such games, my Elo is about 2150, though I am not in good blitz form atm.]
I do have a small suggestion for Vas/Larry. It seems like one can quite easily "dictate" the opening because Rybka will play the same moves. So once you find a way to beat her, you are likely to be able to do it over and over. So it would be great if there were some small standard handicap books, just to make sure that play is varied. Heck, I may even make them myself, now I'm thinking about it.
I played the games with white missing the b-knight. Here's my win (I'm too embarrased to post my losses :P).
[Event "?"]
[Site "?"]
[Date "????.??.??"]
[Round "?"]
[White "Rybka 2.2n2 mp Local"]
[Black "Dagh, Nielsen"]
[Result "0-1"]
[SetUp "1"]
[FEN "rnbqkbnr/pppppppp/8/8/8/8/PPPPPPPP/R1BQKBNR w KQkq - 0 1"]
[PlyCount "92"]
{512MB, PlayChess2006.ctg, REBEKKA} 1. e4 {23} e5 {4} 2. Nf3 {7} Nc6 {9} 3. Bb5
{2} a6 {2} 4. Bxc6 {6} dxc6 {1} 5. O-O {13} f6 {7} 6. d4 {9} exd4 {2} 7. Nxd4 {
8} c5 {5} 8. Ne2 {2} Qxd1 {2} 9. Rxd1 {0} Bd7 {3} 10. b3 {15} O-O-O {5} 11. Bf4
{4} Ne7 {7} 12. Rd2 {1} Ng6 {4} 13. Be3 {5} Bc6 {5} 14. Rxd8+ {1} Kxd8 {1} 15.
Rd1+ {4} Kc8 {2} 16. f3 {1} b6 {3} 17. Ng3 {15} Bd6 {25} 18. Nf5 {2} Bf8 {6}
19. Ng3 {1} Ne7 {2} 20. Nh5 {6} Be8 {9} 21. g4 {5} Bxh5 {4} 22. gxh5 {1} Nc6 {4
} 23. Rd2 {3} Be7 {3} 24. Rg2 {8} Rg8 {2} 25. f4 {3} Kd7 {13} 26. f5 {3} Ke8 {
13} 27. h6 {15} Kf7 {3} 28. hxg7 {9} Rxg7 {1} 29. Rxg7+ {6} Kxg7 {0} 30. c3 {4}
Nd8 {21} 31. h4 {7} h5 {5} 32. Bf4 {3} Bd6 {2} 33. Bxd6 {0} cxd6 {1} 34. Kf2 {
17} Kf7 {4} 35. Ke2 {3} Ke7 {1} 36. Kd2 {2} Kd7 {1} 37. Ke3 {2} Nc6 {1} 38. Kf4
{2} Ne5 {1} 39. Ke3 {2} Kc6 {3} 40. c4 {2} b5 {3} 41. cxb5+ {1} axb5 {1} 42.
Ke2 {2} b4 {3} 43. Kd2 {1} Nf3+ {2} 44. Ke3 {2} Nxh4 {1} 45. Ke2 {9} Ng2 {4}
46. Kf2 {8} Nf4 {7} 0-1
I do have a small suggestion for Vas/Larry. It seems like one can quite easily "dictate" the opening because Rybka will play the same moves. So once you find a way to beat her, you are likely to be able to do it over and over. So it would be great if there were some small standard handicap books, just to make sure that play is varied. Heck, I may even make them myself, now I'm thinking about it.
I played the games with white missing the b-knight. Here's my win (I'm too embarrased to post my losses :P).
[Event "?"]
[Site "?"]
[Date "????.??.??"]
[Round "?"]
[White "Rybka 2.2n2 mp Local"]
[Black "Dagh, Nielsen"]
[Result "0-1"]
[SetUp "1"]
[FEN "rnbqkbnr/pppppppp/8/8/8/8/PPPPPPPP/R1BQKBNR w KQkq - 0 1"]
[PlyCount "92"]
{512MB, PlayChess2006.ctg, REBEKKA} 1. e4 {23} e5 {4} 2. Nf3 {7} Nc6 {9} 3. Bb5
{2} a6 {2} 4. Bxc6 {6} dxc6 {1} 5. O-O {13} f6 {7} 6. d4 {9} exd4 {2} 7. Nxd4 {
8} c5 {5} 8. Ne2 {2} Qxd1 {2} 9. Rxd1 {0} Bd7 {3} 10. b3 {15} O-O-O {5} 11. Bf4
{4} Ne7 {7} 12. Rd2 {1} Ng6 {4} 13. Be3 {5} Bc6 {5} 14. Rxd8+ {1} Kxd8 {1} 15.
Rd1+ {4} Kc8 {2} 16. f3 {1} b6 {3} 17. Ng3 {15} Bd6 {25} 18. Nf5 {2} Bf8 {6}
19. Ng3 {1} Ne7 {2} 20. Nh5 {6} Be8 {9} 21. g4 {5} Bxh5 {4} 22. gxh5 {1} Nc6 {4
} 23. Rd2 {3} Be7 {3} 24. Rg2 {8} Rg8 {2} 25. f4 {3} Kd7 {13} 26. f5 {3} Ke8 {
13} 27. h6 {15} Kf7 {3} 28. hxg7 {9} Rxg7 {1} 29. Rxg7+ {6} Kxg7 {0} 30. c3 {4}
Nd8 {21} 31. h4 {7} h5 {5} 32. Bf4 {3} Bd6 {2} 33. Bxd6 {0} cxd6 {1} 34. Kf2 {
17} Kf7 {4} 35. Ke2 {3} Ke7 {1} 36. Kd2 {2} Kd7 {1} 37. Ke3 {2} Nc6 {1} 38. Kf4
{2} Ne5 {1} 39. Ke3 {2} Kc6 {3} 40. c4 {2} b5 {3} 41. cxb5+ {1} axb5 {1} 42.
Ke2 {2} b4 {3} 43. Kd2 {1} Nf3+ {2} 44. Ke3 {2} Nxh4 {1} 45. Ke2 {9} Ng2 {4}
46. Kf2 {8} Nf4 {7} 0-1
Thanks Alkelele I think they are operating from a handicap book anyway.Knight odds against a VERY strong player at longer time controls may result in the GM winning most but I have to see it to believe it as Rybka is tremendously strong and the GM must remain focused.One slip and you are dead as we saw in G3 of Rybka-Ehlvest.But maybe realistically it might not be feasible to stage such a match with a top player if you are correct but it might be interesting.
OMG, I got another win! I think this was done in true Ehlvest style :-)))
[Event "?"]
[Site "?"]
[Date "????.??.??"]
[Round "?"]
[White "Rybka 2.3h1 mp bench local"]
[Black "Dagh, Nielsen"]
[Result "0-1"]
[SetUp "1"]
[FEN "rnbqkbnr/pppppppp/8/8/8/8/PPPPPPPP/R1BQKBNR w KQkq - 0 1"]
[PlyCount "90"]
{1024MB, DaghNielsen.ctg, REBEKKA} 1. e4 {10} d5 {6} 2. e5 {7} c5 {4} 3. Bb5+ {
4} Bd7 {3} 4. Bxd7+ {2} Qxd7 {1} 5. Nf3 {9} e6 {2} 6. b3 {11} Nc6 {9} 7. Qe2 {6
} a6 {6} 8. O-O {6} b6 {11} 9. Bb2 {9} Nge7 {2} 10. Rac1 {9} h5 {7} 11. d4 {10}
Nf5 {3} 12. Rfd1 {10} a5 {5} 13. c4 {2} cxd4 {7} 14. Nxd4 {8} Ncxd4 {5} 15.
Bxd4 {4} Nxd4 {3} 16. Rxd4 {1} Bc5 {5} 17. Rd3 {5} d4 {2} 18. Rcd1 {5} g6 {6}
19. Rg3 {7} Qe7 {1} 20. h3 {7} Qb7 {8} 21. a4 {4} Rd8 {9} 22. Rdd3 {6} Kd7 {6}
23. Qe3 {4} Kc8 {3} 24. Qf4 {3} Rd7 {4} 25. Rd1 {27} Kb8 {1} 26. h4 {4} Ka7 {1}
27. Rgd3 {3} Ka6 {1} 28. Rf3 {3} Rhd8 {4} 29. Rfd3 {1} Qc7 {23} 30. Qe4 {2} Ka7
{4} 31. Kh1 {2} Be7 {10} 32. Rxd4 {1} Rxd4 {1} 33. Rxd4 {0} Rxd4 {2} 34. Qxd4 {
1} Qd8 {1} 35. Qxd8 {4} Bxd8 {1} 36. g3 {9} Kb7 {1} 37. Kg2 {2} Kc6 {1} 38. f3
{4} Kc5 {1} 39. Kf1 {0} Kb4 {1} 40. Ke2 {3} Kxb3 {1} 41. Kd3 {2} Be7 {6} 42.
Kd2 {2} Kxc4 {1} 43. Kc2 {0} Bc5 {3} 44. Kd2 {2} Bf2 {1} 45. g4 {7} Bxh4 {2}
0-1
[Event "?"]
[Site "?"]
[Date "????.??.??"]
[Round "?"]
[White "Rybka 2.3h1 mp bench local"]
[Black "Dagh, Nielsen"]
[Result "0-1"]
[SetUp "1"]
[FEN "rnbqkbnr/pppppppp/8/8/8/8/PPPPPPPP/R1BQKBNR w KQkq - 0 1"]
[PlyCount "90"]
{1024MB, DaghNielsen.ctg, REBEKKA} 1. e4 {10} d5 {6} 2. e5 {7} c5 {4} 3. Bb5+ {
4} Bd7 {3} 4. Bxd7+ {2} Qxd7 {1} 5. Nf3 {9} e6 {2} 6. b3 {11} Nc6 {9} 7. Qe2 {6
} a6 {6} 8. O-O {6} b6 {11} 9. Bb2 {9} Nge7 {2} 10. Rac1 {9} h5 {7} 11. d4 {10}
Nf5 {3} 12. Rfd1 {10} a5 {5} 13. c4 {2} cxd4 {7} 14. Nxd4 {8} Ncxd4 {5} 15.
Bxd4 {4} Nxd4 {3} 16. Rxd4 {1} Bc5 {5} 17. Rd3 {5} d4 {2} 18. Rcd1 {5} g6 {6}
19. Rg3 {7} Qe7 {1} 20. h3 {7} Qb7 {8} 21. a4 {4} Rd8 {9} 22. Rdd3 {6} Kd7 {6}
23. Qe3 {4} Kc8 {3} 24. Qf4 {3} Rd7 {4} 25. Rd1 {27} Kb8 {1} 26. h4 {4} Ka7 {1}
27. Rgd3 {3} Ka6 {1} 28. Rf3 {3} Rhd8 {4} 29. Rfd3 {1} Qc7 {23} 30. Qe4 {2} Ka7
{4} 31. Kh1 {2} Be7 {10} 32. Rxd4 {1} Rxd4 {1} 33. Rxd4 {0} Rxd4 {2} 34. Qxd4 {
1} Qd8 {1} 35. Qxd8 {4} Bxd8 {1} 36. g3 {9} Kb7 {1} 37. Kg2 {2} Kc6 {1} 38. f3
{4} Kc5 {1} 39. Kf1 {0} Kb4 {1} 40. Ke2 {3} Kxb3 {1} 41. Kd3 {2} Be7 {6} 42.
Kd2 {2} Kxc4 {1} 43. Kc2 {0} Bc5 {3} 44. Kd2 {2} Bf2 {1} 45. g4 {7} Bxh4 {2}
0-1
YES!!! I can also beat Rybka in the Sicilian!
[Event "?"]
[Site "?"]
[Date "????.??.??"]
[Round "?"]
[White "Rybka 2.3h1 mp bench local"]
[Black "Dagh, Nielsen"]
[Result "0-1"]
[SetUp "1"]
[FEN "rnbqkbnr/pppppppp/8/8/8/8/PPPPPPPP/R1BQKBNR w KQkq - 0 1"]
[PlyCount "108"]
{1024MB, DaghNielsen.ctg, REBEKKA} 1. e4 {12} c5 {2} 2. Nf3 {14} d6 {2} 3. Bc4
{7} Nf6 {4} 4. d3 {13} g6 {2} 5. e5 {6} dxe5 {1} 6. Nxe5 {6} e6 {1} 7. O-O {5}
Bg7 {3} 8. Bg5 {13} h6 {2} 9. Bd2 {11} O-O {2} 10. Qc1 {11} Kh7 {2} 11. Re1 {5}
Nbd7 {4} 12. Nf3 {5} b6 {16} 13. Bf4 {4} Bb7 {6} 14. Bd6 {4} Re8 {2} 15. c3 {4}
Bd5 {6} 16. Bxd5 {7} Nxd5 {5} 17. Qc2 {2} N7f6 {7} 18. Be5 {2} Rc8 {10} 19.
Rad1 {4} Nd7 {3} 20. Bg3 {4} Ne7 {8} 21. d4 {4} Nf5 {5} 22. dxc5 {4} Rxc5 {3}
23. Qa4 {1} Re7 {18} 24. Bh4 {6} Nxh4 {2} 25. Qxh4 {1} Qe8 {17} 26. Re3 {9} Rh5
{22} 27. Qc4 {4} Nc5 {23} 28. Rd2 {4} Rd7 {6} 29. Qe2 {2} Qd8 {3} 30. Rxd7 {1}
Qxd7 {3} 31. c4 {1} Bd4 {22} 32. Nxd4 {2} Qxd4 {1} 33. g3 {2} Qf6 {18} 34. b4 {
3} Nd7 {3} 35. Kg2 {2} Re5 {2} 36. Rxe5 {5} Qxe5 {1} 37. Qxe5 {1} Nxe5 {1} 38.
c5 {0} bxc5 {3} 39. bxc5 {1} Kg7 {2} 40. f4 {1} Nc6 {1} 41. Kf3 {2} Kf6 {1} 42.
Ke4 {1} g5 {5} 43. h3 {3} gxf4 {3} 44. gxf4 {2} Ke7 {2} 45. f5 {2} Kf6 {3} 46.
fxe6 {1} Kxe6 {2} 47. a3 {0} f5+ {2} 48. Ke3 {0} Ke5 {1} 49. h4 {2} f4+ {1} 50.
Kd3 {1} Kf5 {4} 51. Kd2 {3} Ke4 {1} 52. a4 {4} f3 {2} 53. a5 {6} Kf4 {2} 54. h5
{3} Kg3 {1} 0-1
[Event "?"]
[Site "?"]
[Date "????.??.??"]
[Round "?"]
[White "Rybka 2.3h1 mp bench local"]
[Black "Dagh, Nielsen"]
[Result "0-1"]
[SetUp "1"]
[FEN "rnbqkbnr/pppppppp/8/8/8/8/PPPPPPPP/R1BQKBNR w KQkq - 0 1"]
[PlyCount "108"]
{1024MB, DaghNielsen.ctg, REBEKKA} 1. e4 {12} c5 {2} 2. Nf3 {14} d6 {2} 3. Bc4
{7} Nf6 {4} 4. d3 {13} g6 {2} 5. e5 {6} dxe5 {1} 6. Nxe5 {6} e6 {1} 7. O-O {5}
Bg7 {3} 8. Bg5 {13} h6 {2} 9. Bd2 {11} O-O {2} 10. Qc1 {11} Kh7 {2} 11. Re1 {5}
Nbd7 {4} 12. Nf3 {5} b6 {16} 13. Bf4 {4} Bb7 {6} 14. Bd6 {4} Re8 {2} 15. c3 {4}
Bd5 {6} 16. Bxd5 {7} Nxd5 {5} 17. Qc2 {2} N7f6 {7} 18. Be5 {2} Rc8 {10} 19.
Rad1 {4} Nd7 {3} 20. Bg3 {4} Ne7 {8} 21. d4 {4} Nf5 {5} 22. dxc5 {4} Rxc5 {3}
23. Qa4 {1} Re7 {18} 24. Bh4 {6} Nxh4 {2} 25. Qxh4 {1} Qe8 {17} 26. Re3 {9} Rh5
{22} 27. Qc4 {4} Nc5 {23} 28. Rd2 {4} Rd7 {6} 29. Qe2 {2} Qd8 {3} 30. Rxd7 {1}
Qxd7 {3} 31. c4 {1} Bd4 {22} 32. Nxd4 {2} Qxd4 {1} 33. g3 {2} Qf6 {18} 34. b4 {
3} Nd7 {3} 35. Kg2 {2} Re5 {2} 36. Rxe5 {5} Qxe5 {1} 37. Qxe5 {1} Nxe5 {1} 38.
c5 {0} bxc5 {3} 39. bxc5 {1} Kg7 {2} 40. f4 {1} Nc6 {1} 41. Kf3 {2} Kf6 {1} 42.
Ke4 {1} g5 {5} 43. h3 {3} gxf4 {3} 44. gxf4 {2} Ke7 {2} 45. f5 {2} Kf6 {3} 46.
fxe6 {1} Kxe6 {2} 47. a3 {0} f5+ {2} 48. Ke3 {0} Ke5 {1} 49. h4 {2} f4+ {1} 50.
Kd3 {1} Kf5 {4} 51. Kd2 {3} Ke4 {1} 52. a4 {4} f3 {2} 53. a5 {6} Kf4 {2} 54. h5
{3} Kg3 {1} 0-1
Thanks. I appreciate your game. But we are talking about a match here and on the current hardware that Rybka is running on. The main problem I see at first is Rybka will need a book for the Knight odds game so you can't repeat a known won game after many tries at the same opening. I have seen that too many times on ICC. I have seen GM's play the same opening over and over and over untill they find a line that leads to an advantage big enough to win a game. They are just probing for that one weakness they can find at blitz. I have seen them resign after 15 moves just so they can start over again to try something different. So GM's can actually stack the deck in their favor with a lot of practice against the same program and same Knight odds. I believe this is why some players that are not even GM's can handle many programs at blitz on ICC. They have several pet lines which they know work against certain computers. Just my thoughts on the subject. Mainly I would just like to see an "even" match.
![[hu]](/mwf/v12/flags/hu.png "Hungary")
Actually, it's not a bad idea. Including an internal opening book with a quick editing tool would not be much work.
In the past, I resisted it because it's not possible to view the contents of the book properly in your favorite GUI. In fact, though, there might be some workarounds.
For example, if the engine is in book, she sends as her move-1 PV her highest-percentage move followed by the best book moves after that.
As her move-2 PV, she sends her second-percentage move, followed by the best book moves after that.
The score could be either some scoring % assigned to the moves, or the chance that they are played.
The depth could be the # of games (or something like that).
I'll think about it a bit.
BTW - post your final score when you are done, I'm curious.
Vas
In the past, I resisted it because it's not possible to view the contents of the book properly in your favorite GUI. In fact, though, there might be some workarounds.
For example, if the engine is in book, she sends as her move-1 PV her highest-percentage move followed by the best book moves after that.
As her move-2 PV, she sends her second-percentage move, followed by the best book moves after that.
The score could be either some scoring % assigned to the moves, or the chance that they are played.
The depth could be the # of games (or something like that).
I'll think about it a bit.
BTW - post your final score when you are done, I'm curious.
Vas
My final score was very bad! I did not keep track. I learned that some structures were simply too complicated for me to play. I did manage to reach easily won endings far more times than my score would indicate, but then I would far too often blunder away my extra piece. So I think my 2150 ELO should really be taken with a gran of salt regarding my performance - my blitz game is much, much worse than that these days.
It is my vague feeling that I could train myself up to reaching a + score, by learning which setups give rise to easy play. Essentially, what you want is to win without ever calculating any tactics. Something like, there is one file open, you exhange rooks on that file etc. I think that even if Rybka had a small book to vary play, I would also be able to learn which setup's would provoke simplifications from her.
An example is 1.e4 and now 1...e6 or 1...d5. I would always get a French. If I learned how to develop and block my kingside by Rybka's happy cooperation, I think it would be quite easy to win time after time. Also, after 1...d5 2.e5 c5 3.Nf3 Bg4, I already will have one piece exchanged. An alternative is Bf5, then I know that Rybka almost always prefer Bd3, offering an exchange.
I didn't repeat 1...e5 and a Bxc6 Ruy Lopez. It was simply too easy to win after the queen exchange in that line ;-)
It is my vague feeling that I could train myself up to reaching a + score, by learning which setups give rise to easy play. Essentially, what you want is to win without ever calculating any tactics. Something like, there is one file open, you exhange rooks on that file etc. I think that even if Rybka had a small book to vary play, I would also be able to learn which setup's would provoke simplifications from her.
An example is 1.e4 and now 1...e6 or 1...d5. I would always get a French. If I learned how to develop and block my kingside by Rybka's happy cooperation, I think it would be quite easy to win time after time. Also, after 1...d5 2.e5 c5 3.Nf3 Bg4, I already will have one piece exchanged. An alternative is Bf5, then I know that Rybka almost always prefer Bd3, offering an exchange.
I didn't repeat 1...e5 and a Bxc6 Ruy Lopez. It was simply too easy to win after the queen exchange in that line ;-)
Thanks.
By the way - I tried playing three games yesterday during dinner on my quad Opteron. All three were drawn. It's embarassing how much stuff I overlooked, but usually it wasn't so bad and the game would continue. Then, three moves later, I'd overlook something else. :)
Anyway, I also feel that with a bit of training, I could score pretty well.
Vas
By the way - I tried playing three games yesterday during dinner on my quad Opteron. All three were drawn. It's embarassing how much stuff I overlooked, but usually it wasn't so bad and the game would continue. Then, three moves later, I'd overlook something else. :)
Anyway, I also feel that with a bit of training, I could score pretty well.
Vas
Based on all the games I have played myself, witnessed, or heard about here, I can say the following: The current version (2.3) on a quad can give knight odds (b1 knight) with fairly even results in blitz (5' or 4'2") to players up to around 2400 FIDE, provided that either the player or the computer varies its play in the first couple moves each game (a different good first move can simply be input for Rybka each game). If the time limit is a bit longer, such as 5'+5", 2300 looks about right. For actual tournament chess, 40/2, I think the figure would drop to around 2000. So talk of playing GMs at knight odds now is unreasonable, unless you are talking about bullet chess.
It looks clear to me that for a borderline GM (2500 FIDE) the proper handicap in serious chess for Rybka 2.3 on a quad is the traditional "pawn and Move", meaning the computer plays Black and removes f7. That should be a good battle. It is roughly twice as large a handicap as the one used in the eight game Ehlvest match on average, since the "f" pawn is the worst one to remove according to both theory and my computer-computer tests, and since the first move is significant. Another option that might be a bit too hard for the computer but worth trying is the Exchange - Rybka removes a1 rook, black removes b8 knight.
Aside from the varied pawn handicap of the Ehlvest match, other possibilities for very strong GMs are White plus extra moves (1 or 2 extra), castling odds, or requiring Rybka to play Black and open ...f6 and ...Kf7 (if legal) in every game. There is also draw odds, but that would require special programming to be valid.
It looks clear to me that for a borderline GM (2500 FIDE) the proper handicap in serious chess for Rybka 2.3 on a quad is the traditional "pawn and Move", meaning the computer plays Black and removes f7. That should be a good battle. It is roughly twice as large a handicap as the one used in the eight game Ehlvest match on average, since the "f" pawn is the worst one to remove according to both theory and my computer-computer tests, and since the first move is significant. Another option that might be a bit too hard for the computer but worth trying is the Exchange - Rybka removes a1 rook, black removes b8 knight.
Aside from the varied pawn handicap of the Ehlvest match, other possibilities for very strong GMs are White plus extra moves (1 or 2 extra), castling odds, or requiring Rybka to play Black and open ...f6 and ...Kf7 (if legal) in every game. There is also draw odds, but that would require special programming to be valid.
![[de]](/mwf/v12/flags/de.png "Germany")
I take the challenge! I play 5 0 with knight odds. my rating is around 2100 elo.
my record against rybka 2.2 32 bit on 2.66GHz P4 with alternating knight odds on b1, g1, b8, g8 is now 7-5 in my favor.
I'll beat the hell out of the beast!
my price is 50 €, hehe! 8 game match
congratulations about version 2.3.1 - great improvement and still some potential!
my record against rybka 2.2 32 bit on 2.66GHz P4 with alternating knight odds on b1, g1, b8, g8 is now 7-5 in my favor.
I'll beat the hell out of the beast!
my price is 50 €, hehe! 8 game match
congratulations about version 2.3.1 - great improvement and still some potential!
Do not forget that the challange is with better hardware.
I expect you to lose a blitz match when rybka use some small opening book and does not have knight at b1
Uri
I expect you to lose a blitz match when rybka use some small opening book and does not have knight at b1
Uri
Larry summed it up pretty reasonably.Serious knight odds now at 40/2 against a top player is not feasable at the moment,but a faster time controlled match approaching blitz(maybe NOT blitz because no human could survive 5 minute chess)but maybe G/15 or the like.By the way Larry are you the owner of "The Rybka House"? as Ehlvest calls it,if so how much you charge for a tour?
"(maybe NOT blitz because no human could survive 5 minute chess)"
why not? I played 5 min games
why not? I played 5 min games
LK said with a Quad 64bit machine... I do assume he also means with 3-6 men tablebases on.
On a single processor 32 bit I'd assume it would drop to about 2000 or below at blitz (correct me if I'm wrong LK).
(I'm at 1900-1950 level and I get tons of draws at that setting in blitz).
On a single processor 32 bit I'd assume it would drop to about 2000 or below at blitz (correct me if I'm wrong LK).
(I'm at 1900-1950 level and I get tons of draws at that setting in blitz).
So if you're 2100 I'd guess that even on dual processor 32 bit version of Rybka you have a chance at winning at this setting at blitz.
(If you're not using all tablebases up to 6, it's even easier on you, but Rybka's full strength is evaluated with them).
a dual processor 64bit with all 3-6 men tablebases already sounds like beyond your capabilities in blitz...
Also note that from a previous post of yours, you seem to mention taking white in some of those games. Talking about knight odds, the meaning is that the machine always plays white (otherwise it's called "knight AND move odds", which is a lot worse).
(If you're not using all tablebases up to 6, it's even easier on you, but Rybka's full strength is evaluated with them).
a dual processor 64bit with all 3-6 men tablebases already sounds like beyond your capabilities in blitz...
Also note that from a previous post of yours, you seem to mention taking white in some of those games. Talking about knight odds, the meaning is that the machine always plays white (otherwise it's called "knight AND move odds", which is a lot worse).
Tablebases make almost no difference statistically, unless the time limit is so short (like small fixed depth) that Rybka cannot do the K and R checkmate.
Actually on my old computer (s.p. 32 bit) a visiting young master (2200+ USCF, a bit lower FIDE) got crushed twice at knight odds by Rybka, and based on other reports here as well I think you need to be well over 2200 FIDE to be evenly matched with Rybka 2.31 at knight odds in true blitz (5-0 or 4-2), assuming varied openings.
Regarding "tons of draws": be sure to set contempt at a decent value when playing knight odds, maybe somewhere in the 50-100 range.
Finally I am beginning to think that the disparity between European national ratings and FIDE ratings may be greater than I thought. A player with a rating around 2000 FIDE tends to be a rather weak player (relatively speaking), because (until the bottom was dropped a couple years ago) he only needed to get lucky a couple times with performances over 2000 to get such a rating; the bad resuls didn't count. So I think that a player with a 2000 German or French or Dutch or Swedish (etc.) rating might be a much stronger player than one with a FIDE 2000 rating. I'd appreciate feedback on this. All my Elo references are to FIDE ratings, not national ones.
Actually on my old computer (s.p. 32 bit) a visiting young master (2200+ USCF, a bit lower FIDE) got crushed twice at knight odds by Rybka, and based on other reports here as well I think you need to be well over 2200 FIDE to be evenly matched with Rybka 2.31 at knight odds in true blitz (5-0 or 4-2), assuming varied openings.
Regarding "tons of draws": be sure to set contempt at a decent value when playing knight odds, maybe somewhere in the 50-100 range.
Finally I am beginning to think that the disparity between European national ratings and FIDE ratings may be greater than I thought. A player with a rating around 2000 FIDE tends to be a rather weak player (relatively speaking), because (until the bottom was dropped a couple years ago) he only needed to get lucky a couple times with performances over 2000 to get such a rating; the bad resuls didn't count. So I think that a player with a 2000 German or French or Dutch or Swedish (etc.) rating might be a much stronger player than one with a FIDE 2000 rating. I'd appreciate feedback on this. All my Elo references are to FIDE ratings, not national ones.
@ Larry Kaufmann: you are quite right about your rating observations. a player with a 2000 german national rating should be considered 2100-2150 Elo. I got an elo rating some years ago around 2050 when my national rating was 1720. now I didn't play many elo tourneys and my national rating is only a bit below my elo, that's also because national rating changes faster than elo.
my last tourney performance was 2360 elo and 2180 DWZ :-)
my last tourney performance was 2360 elo and 2180 DWZ :-)
I've witnessed many more examples of bad behavior without tablebases, though those may disappear if I migrate to a stronger machine.
I already had contempt for draw at 50.
Thought it was enough.
But increasing it all the way up to 100 seems to indeed decrease the number of draws somewhat.
By a lot in anything less than 5-0.
Not by too much on 5-0 though (and above).
I already had contempt for draw at 50.
Thought it was enough.
But increasing it all the way up to 100 seems to indeed decrease the number of draws somewhat.
By a lot in anything less than 5-0.
Not by too much on 5-0 though (and above).
Larry,
FIDE ratings <2100 tend to be unpredictable and inaccurate. Some players with such ratings are quite a bit better than others.
The US national rating system (and also the German national rating system) are quite accurate for lower-rated players. A US 1900 player is very likely to be stronger than a US 1700 player. The same holds true for players with German ratings.
German national ratings are also highly deflated (probably by 100 Elo) relative to FIDE ratings, while US national ratings are somewhat inflated (maybe by 50-75 Elo).
Vas
FIDE ratings <2100 tend to be unpredictable and inaccurate. Some players with such ratings are quite a bit better than others.
The US national rating system (and also the German national rating system) are quite accurate for lower-rated players. A US 1900 player is very likely to be stronger than a US 1700 player. The same holds true for players with German ratings.
German national ratings are also highly deflated (probably by 100 Elo) relative to FIDE ratings, while US national ratings are somewhat inflated (maybe by 50-75 Elo).
Vas
![[fr]](/mwf/v12/flags/fr.png "France")
Hello everyone,
I saw this thread and I played a few games at knight odds against Rybka, three at "long" time controls (40'') and one in blitz (4''). I have no rating so you can consider me as "NN".
I did not have any specific problems in the long games, because a knight is too much a disadvantage; I do not think there is a challenge in a knight-odds match between Rybka and any strong player, even with a specific opening book. However, in a blitz game, things are more "unclear" ... Here is my blitz game (4'') against Rybka, black and without the Nb8 :
1. e4 b6 2. d4 e6 3. Nc3 Bb7 4. Bd3 Nf6 5. Nf3 h6 6. e5 Nd5 7. Nxd5 Bxd5 8. c4 Bb4 9. Bd2 Bxd2+ 10. Qxd2 Bxf3 11. gxf3 f5 12. O-O-O Qh4 13. Qe3 Kf7 14. f4 g5 15. fxg5 hxg5 16. Kb1 Qf4 17. Qxf4 gxf4 18. Rdg1 Rh4 19. h3 Rah8 20. Bf1 f3 21. Rg3 Bxd4 22. Rxf3 Re4 23. Re3 Rxe3 24. fxe3 Rh4 25. Bd3 Ke7 26. Kc2 a6 27. b4 Kf7 28. a4 Ke7 29. Kb3 a5 30. bxa5 bxa5 31. Kc3 c6 32. Kd2 d6 33. exd6+ Kxd6 34. Kc3 Kc5 35. e4 f4 36. e5 f3 37. Bf1 Kb6 38. Kd2 Kc5 39. Ke3 Kb4 40. Kxf3 Kxa4 41. Kg3 Rh7 42. h4 Kb3 43. h5 a4 44. Kf4 a3 45. Bd3 Rf7+ 46. Kg5 a2 47. Kg6 Ra7 48. h6 a1Q 49. Rxa1 Rxa1 50. h7 Ra8 51. Kg7 c5 52. h8Q Rxh8 53. Kxh8 Kc3 54. Bf1 Kd4 55. Kg7 Kxe5 56. Kf7 Kd6 57. Bd3 Ke5 58. Ke7 1-0
After 58... Kd4 59. Kxe6 Kxd3 60. Kxd5 White wins easily. I do not remember if Rybka "resigned" (and if yes, how?) or lost on time, but anyway I had won. So, any thoughts about that ?
I saw this thread and I played a few games at knight odds against Rybka, three at "long" time controls (40'') and one in blitz (4''). I have no rating so you can consider me as "NN".
I did not have any specific problems in the long games, because a knight is too much a disadvantage; I do not think there is a challenge in a knight-odds match between Rybka and any strong player, even with a specific opening book. However, in a blitz game, things are more "unclear" ... Here is my blitz game (4'') against Rybka, black and without the Nb8 :
1. e4 b6 2. d4 e6 3. Nc3 Bb7 4. Bd3 Nf6 5. Nf3 h6 6. e5 Nd5 7. Nxd5 Bxd5 8. c4 Bb4 9. Bd2 Bxd2+ 10. Qxd2 Bxf3 11. gxf3 f5 12. O-O-O Qh4 13. Qe3 Kf7 14. f4 g5 15. fxg5 hxg5 16. Kb1 Qf4 17. Qxf4 gxf4 18. Rdg1 Rh4 19. h3 Rah8 20. Bf1 f3 21. Rg3 Bxd4 22. Rxf3 Re4 23. Re3 Rxe3 24. fxe3 Rh4 25. Bd3 Ke7 26. Kc2 a6 27. b4 Kf7 28. a4 Ke7 29. Kb3 a5 30. bxa5 bxa5 31. Kc3 c6 32. Kd2 d6 33. exd6+ Kxd6 34. Kc3 Kc5 35. e4 f4 36. e5 f3 37. Bf1 Kb6 38. Kd2 Kc5 39. Ke3 Kb4 40. Kxf3 Kxa4 41. Kg3 Rh7 42. h4 Kb3 43. h5 a4 44. Kf4 a3 45. Bd3 Rf7+ 46. Kg5 a2 47. Kg6 Ra7 48. h6 a1Q 49. Rxa1 Rxa1 50. h7 Ra8 51. Kg7 c5 52. h8Q Rxh8 53. Kxh8 Kc3 54. Bf1 Kd4 55. Kg7 Kxe5 56. Kf7 Kd6 57. Bd3 Ke5 58. Ke7 1-0
After 58... Kd4 59. Kxe6 Kxd3 60. Kxd5 White wins easily. I do not remember if Rybka "resigned" (and if yes, how?) or lost on time, but anyway I had won. So, any thoughts about that ?
Even if you have no official rating, I'm sure you have at least an internet rating or you know what rating opponents you can play evenly with. So what is your "unofficial" rating? As Uri mentioned in another thread, even a player rated around 2000 is likely to lose a CORRESPONDENCE (!!) game at knight odds to Rybka! You should know that knight odds means that the odds-giver gets White, so try again that way. Also, what is your hardware? If you just have a 32 bit sp computer, then you "only" need to be around 2300 level to compete with Rybka 2.3.2a at knight odds in blitz. Are you using this version? Also, set contempt for the maximum, and force Rybka to open with 1Nf3 or 1f4 rather than 1e4.
This was my opinion earlier but now
I am not sure he is likely to lose.
It seems that rybka did not evaluate the relevant position correctly and the human still has a clear advantage.
I will post the full game when it is over.
Uri
I am not sure he is likely to lose.
It seems that rybka did not evaluate the relevant position correctly and the human still has a clear advantage.
I will post the full game when it is over.
Uri
Hello,
Sorry, I do not have an Internet rating which could correspond to any Elo rating, so I do not know at all my "strength". Yes, I had forgotten that Rybka should be White... But, speaking of games in which Rybka is white, I have one or two 40'' game in which Rybka is White and begins by 1. Nf3, so if you are interested I can post them. (I am very surprised about a player rated around 2000 likely to lose a correspondence game at knight-odds... in 40'' I had no particular problem, and I am not IM or anything, so... But I believe you.)
Yes, I only have a 32 bit computer. For my hardware: Intel Pentium 1.7 Ghz (400 Mhz FSB, 2 MB L2 cache - I do not know what it means, but if it can help you...). The computer is one-year-old.
OK, I shall try blitz with Rybka as White (and with 1. Nf3 - but actually I never saw Rybka opening with the KP in knight-odds), and if I win any -- which I doubt -- I shall tell you. ;)
Thanks for your answer,
Best regards.
Sorry, I do not have an Internet rating which could correspond to any Elo rating, so I do not know at all my "strength". Yes, I had forgotten that Rybka should be White... But, speaking of games in which Rybka is white, I have one or two 40'' game in which Rybka is White and begins by 1. Nf3, so if you are interested I can post them. (I am very surprised about a player rated around 2000 likely to lose a correspondence game at knight-odds... in 40'' I had no particular problem, and I am not IM or anything, so... But I believe you.)
Yes, I only have a 32 bit computer. For my hardware: Intel Pentium 1.7 Ghz (400 Mhz FSB, 2 MB L2 cache - I do not know what it means, but if it can help you...). The computer is one-year-old.
OK, I shall try blitz with Rybka as White (and with 1. Nf3 - but actually I never saw Rybka opening with the KP in knight-odds), and if I win any -- which I doubt -- I shall tell you. ;)
Thanks for your answer,
Best regards.
Oh, and... One question: what is exactly "contempt"?
"Contempt" is a parameter (click on "engine parameters") you can set at any value from 0 to 100. It represents how strongly the computer will try to avoid a draw (because it has "contempt" for your play compared to its own). In the current commercial version it only does this by trying to avoid repetitions and other obvious draws, but in the next version it will also try to avoid exchanges when this is set. It's not so important to use it for normal play (though you should still set a value of at least 10 so it will not seek a draw just due to playing the Black pieces), but for knight odds play you should set it to the maximum 100. Otherwise, since it starts a knight down, it will play almost singe-mindedly for a draw, and will have trouble winning games unless you blunder. Even with the setting at 100 (one pawn), Rybka will still seek a draw, but at least if you make some small mistakes and your advantage drops below 100 Rybka will try to win.
Frankly I can't believe that a player who can beat Rybka at knight odds could have no idea of his level. How could you have acquired enough ability to do this without playing often with ranked players? I think you are just being coy. I once met a Japanese Go player who suggested a game but denied any knowledge of his strength or level; upon further investigation I finally learned that he was of Grandmaster strength (in chess terms)!
Frankly I can't believe that a player who can beat Rybka at knight odds could have no idea of his level. How could you have acquired enough ability to do this without playing often with ranked players? I think you are just being coy. I once met a Japanese Go player who suggested a game but denied any knowledge of his strength or level; upon further investigation I finally learned that he was of Grandmaster strength (in chess terms)!
OK, I set the value to 100; I played two blitz and was crushed by my own blunders (four-minute chess...), which is always pleasant.
I am being sincere; I cannot give any rating -- and believe me, I would be pleased to tell you, but... I acquired the skill I have through observation: I look how strong programs and strong human players play and I study openings.
I do not know what means "being coy" -- surely you mean something like I'm hiding from you, but frankly no. And I am no GrandMaster! (Perhaps one day...).
I am being sincere; I cannot give any rating -- and believe me, I would be pleased to tell you, but... I acquired the skill I have through observation: I look how strong programs and strong human players play and I study openings.
I do not know what means "being coy" -- surely you mean something like I'm hiding from you, but frankly no. And I am no GrandMaster! (Perhaps one day...).
If you want Rybka to begin by 1.Nf3, I suppose I have to remove always the same knight (Nb1) ?
Yes, it's traditional to remove b1 knight. I know you are no grandmaster, but you must be at least around 2000 level (or better) to win games from Rybka at knight odds, and I never knew anyone to reach that level without playing many games against other people with known ratings (either official or internet), except perhaps in some isolated areas of the world. Are you really saying that you have never played chess with anyone with a known rating?
Yes, I never played such a person.
Did you play a lot against chess programs?
I think that it is possible to get playing strength of 2000 without playing against humans only by experience of reading chess books and playing against chess programs.
I know about one human who did it and in his first tournament against humans got performance above 2000(I do not know it as a fact but I believe him).
Uri
I think that it is possible to get playing strength of 2000 without playing against humans only by experience of reading chess books and playing against chess programs.
I know about one human who did it and in his first tournament against humans got performance above 2000(I do not know it as a fact but I believe him).
Uri
Not a lot, sometimes (draw/losses). As I said, it was mostly observation.
new knight odds threat and a new game played by me :-).
same time control as alkelele 5 0. I used too much time for easy moves - just play through it. my rating is similar to alkeleles and talking about knight odds in corr. chess against 2000 FIDE elo is ridiculus in my opinion.
[Event "Blitz:5'"]
[Site "?"]
[Date "????.??.??"]
[Round "?"]
[White "Rybka 2.3.2a 32-bit"]
[Black "lem"]
[Result "*"]
[SetUp "1"]
[FEN "rnbqkbnr/pppppppp/8/8/8/8/PPPPPPPP/R1BQKBNR w KQkq - 0 1"]
[PlyCount "101"]
[TimeControl "300"]
1. e4 {-1.79/12 10} d5 {21} 2. e5 {-1.73/13 11} e6 {3} 3. Nf3 {-1.55/12 9} c5 {
9} 4. Be2 {-1.56/12 9} Nc6 {13} 5. O-O {-1.79/13 14} Qc7 {5} 6. Re1 {
-1.82/10 10} Bd7 {5} 7. Bd3 {-1.92/10 6} Nge7 {20} 8. b3 {-1.92/12 8} Ng6 {9}
9. Bb2 {-1.90/11 3} Be7 {7} 10. g3 {-1.98/12 6} O-O-O {36} 11. a3 {-1.99/11 7}
f6 {14} 12. exf6 {-2.07/9 3} Bxf6 {6} 13. Bxf6 {-2.15/12 4} gxf6 {1} 14. b4 {
-2.31/12 5} c4 {10} 15. Bf1 {-2.36/12 0} h5 {7} 16. d3 {-2.37/9 6} h4 {22} 17.
Rc1 {-2.61/11 0} hxg3 {9} 18. hxg3 {-3.01/11 6} Nce5 {23} 19. Nxe5 {-2.90/12 12
} Nxe5 {2} 20. d4 {-3.09/12 1} Nc6 {12} 21. Bg2 {-2.92/12 5} Rdg8 {4} 22. Qd2 {
-2.93/13 4} Ne7 {2} 23. c3 {-3.01/11 5} Nf5 {2} 24. a4 {-3.08/11 3} Nxg3 {2}
25. fxg3 {-2.83/9 0} Qxg3 {1} 26. Re3 {-5.39/10 0} Qh2+ {2} 27. Kf1 {-5.48/13 2
} Rg4 {5} 28. Qf2 {-5.63/13 2} f5 {5} 29. Rc2 {-3.51/11 3} Rhg8 {4} 30. Rh3 {
-2.33/10 1} Qd6 {4} 31. Bf3 {-2.48/12 1} R4g5 {5} 32. a5 {-2.42/12 1} Qf4 {5}
33. Bg2 {-2.18/12 3} Qxf2+ {4} 34. Rxf2 {-2.01/11 0} Rg3 {3} 35. Rxg3 {
-2.11/13 1} Rxg3 {0} 36. Bf3 {-2.01/14 1} Kc7 {2} 37. Ke2 {-2.36/14 4} Kd6 {1}
38. Ke3 {-2.22/15 4} Rg1 {1} 39. Bg2 {-2.39/14 3} Rc1 {1} 40. Kd2 {-2.82/14 0}
Ra1 {1} 41. Bf3 {-3.08/15 1} Ra2+ {1} 42. Ke1 {-3.36/16 1} Rxf2 {0} 43. Kxf2 {
-3.36/16 0} e5 {1} 44. dxe5+ {-3.74/15 4} Kxe5 {0} 45. Ke2 {-3.90/17 2} d4 {1}
46. Bxb7 {-3.68/16 3} d3+ {1} 47. Ke3 {-1.89/18 4} f4+ {1} 48. Kd2 {-1.88/19 1}
Ba4 {1} 49. Ba6 {-1.88/19 2} Kf6 {2} 50. Bxc4 {0.85/13 1} Bc2 {2} 51. b5 {
2.31/16 2} *
lost on time (and position finally)
same time control as alkelele 5 0. I used too much time for easy moves - just play through it. my rating is similar to alkeleles and talking about knight odds in corr. chess against 2000 FIDE elo is ridiculus in my opinion.
[Event "Blitz:5'"]
[Site "?"]
[Date "????.??.??"]
[Round "?"]
[White "Rybka 2.3.2a 32-bit"]
[Black "lem"]
[Result "*"]
[SetUp "1"]
[FEN "rnbqkbnr/pppppppp/8/8/8/8/PPPPPPPP/R1BQKBNR w KQkq - 0 1"]
[PlyCount "101"]
[TimeControl "300"]
1. e4 {-1.79/12 10} d5 {21} 2. e5 {-1.73/13 11} e6 {3} 3. Nf3 {-1.55/12 9} c5 {
9} 4. Be2 {-1.56/12 9} Nc6 {13} 5. O-O {-1.79/13 14} Qc7 {5} 6. Re1 {
-1.82/10 10} Bd7 {5} 7. Bd3 {-1.92/10 6} Nge7 {20} 8. b3 {-1.92/12 8} Ng6 {9}
9. Bb2 {-1.90/11 3} Be7 {7} 10. g3 {-1.98/12 6} O-O-O {36} 11. a3 {-1.99/11 7}
f6 {14} 12. exf6 {-2.07/9 3} Bxf6 {6} 13. Bxf6 {-2.15/12 4} gxf6 {1} 14. b4 {
-2.31/12 5} c4 {10} 15. Bf1 {-2.36/12 0} h5 {7} 16. d3 {-2.37/9 6} h4 {22} 17.
Rc1 {-2.61/11 0} hxg3 {9} 18. hxg3 {-3.01/11 6} Nce5 {23} 19. Nxe5 {-2.90/12 12
} Nxe5 {2} 20. d4 {-3.09/12 1} Nc6 {12} 21. Bg2 {-2.92/12 5} Rdg8 {4} 22. Qd2 {
-2.93/13 4} Ne7 {2} 23. c3 {-3.01/11 5} Nf5 {2} 24. a4 {-3.08/11 3} Nxg3 {2}
25. fxg3 {-2.83/9 0} Qxg3 {1} 26. Re3 {-5.39/10 0} Qh2+ {2} 27. Kf1 {-5.48/13 2
} Rg4 {5} 28. Qf2 {-5.63/13 2} f5 {5} 29. Rc2 {-3.51/11 3} Rhg8 {4} 30. Rh3 {
-2.33/10 1} Qd6 {4} 31. Bf3 {-2.48/12 1} R4g5 {5} 32. a5 {-2.42/12 1} Qf4 {5}
33. Bg2 {-2.18/12 3} Qxf2+ {4} 34. Rxf2 {-2.01/11 0} Rg3 {3} 35. Rxg3 {
-2.11/13 1} Rxg3 {0} 36. Bf3 {-2.01/14 1} Kc7 {2} 37. Ke2 {-2.36/14 4} Kd6 {1}
38. Ke3 {-2.22/15 4} Rg1 {1} 39. Bg2 {-2.39/14 3} Rc1 {1} 40. Kd2 {-2.82/14 0}
Ra1 {1} 41. Bf3 {-3.08/15 1} Ra2+ {1} 42. Ke1 {-3.36/16 1} Rxf2 {0} 43. Kxf2 {
-3.36/16 0} e5 {1} 44. dxe5+ {-3.74/15 4} Kxe5 {0} 45. Ke2 {-3.90/17 2} d4 {1}
46. Bxb7 {-3.68/16 3} d3+ {1} 47. Ke3 {-1.89/18 4} f4+ {1} 48. Kd2 {-1.88/19 1}
Ba4 {1} 49. Ba6 {-1.88/19 2} Kf6 {2} 50. Bxc4 {0.85/13 1} Bc2 {2} 51. b5 {
2.31/16 2} *
lost on time (and position finally)
First of all you (and anyone else playing knight odds) should not let Rybka open 1e4, as it is just a poor move due to 1...d5. I do well in knight odds blitz with Rybka (latest, quad 64 bit) if she opens 1e4, but I normally make her open 1Nf3 or 1f4 (perhaps the two best moves at the odds) and then it's very hard for me. I agree with you that knight odds in correspondence chess vs. 2000 SHOULD be ridiculous, but Uri showed that Rybka could totally outplay such a player enough to reach an unclear position, even if it turns out that the human finally wins the game. By the way it's very strange that you report Rybka lost on time after 51 moves; this is not supposed to happen. Can you account for this? Is there something about your interface or setup that could explain this?
hah! you didn't replay it :-). I lost the game on time. I just wanted to show the first game at 5 0, that I played, although I lost it. 30 more secs and I would have won :-)
I tested now some games with 1.f4 and 1.Nf3 and these moves are obviously better and I had indeed some difficulties. I won 2, but lost 4 or 5, mostly due to horrible blunders, when I already reached a favourable ending. then when it's only a pawn plus in a rook ending you are virtually lost at this time control..
here is one win. when I had only 1.3 secs left rybka luckily gave up.
[Event "Blitz:5'"]
[Site "?"]
[Date "????.??.??"]
[Round "?"]
[White "Rybka 2.3.2a 32-bit"]
[Black "lem"]
[Result "0-1"]
[SetUp "1"]
[FEN "rnbqkbnr/pppppppp/8/8/8/8/PPPPPPPP/R1BQKBNR w KQkq - 0 1"]
[PlyCount "168"]
[TimeControl "300"]
1. f4 d5 {0} 2. b3 {-2.02/12 14} Nf6 {2} 3. Nf3 {-2.12/12 4} g6 {2} 4. e3 {
-2.09/12 2} Bg7 {1} 5. Bb2 {-2.17/13 22} O-O {2} 6. Be2 {-2.21/12 12} b6 {4} 7.
O-O {-1.99/11 11} Bb7 {4} 8. c4 {-1.97/10 7} Nbd7 {9} 9. cxd5 {-1.88/10 12}
Bxd5 {2} 10. d3 {-1.97/11 5} c5 {3} 11. e4 {-1.89/11 3} Bb7 {2} 12. Qd2 {
-1.82/10 6} Ng4 {14} 13. e5 {-1.76/10 3} e6 {6} 14. Ng5 {-1.80/12 8} Nh6 {3}
15. Bf3 {-1.98/12 38} Bxf3 {3} 16. Rxf3 {-1.99/12 9} Nb8 {4} 17. b4 {-1.97/10 5
} cxb4 {4} 18. Qxb4 {-2.02/9 1} Nc6 {1} 19. Qe4 {-2.27/11 4} Qd7 {15} 20. Rc1 {
-2.50/9 6} Rac8 {19} 21. Rc4 {-2.68/11 0} Na5 {7} 22. Rd4 {-2.76/10 4} Qc6 {6}
23. Qe1 {-2.81/10 4} Qc2 {11} 24. Rf2 {-2.59/10 3} Qc5 {7} 25. Ne4 {-2.30/10 3}
Qc6 {6} 26. Nd6 {-2.07/10 3} Rcd8 {5} 27. Ba3 {-2.13/10 3} Nf5 {10} 28. Nxf5 {
-2.29/10 1} exf5 {4} 29. Bd6 {-2.68/12 1} Rfe8 {18} 30. Qd1 {-3.65/15 1} Rc8 {7
} 31. Ba3 {-2.86/10 3} Bf8 {6} 32. Bxf8 {-2.99/11 3} Rxf8 {3} 33. Rd6 {
-3.07/12 3} Qc1 {6} 34. Qxc1 {-3.17/13 0} Rxc1+ {1} 35. Rf1 {-3.46/13 2} Rxf1+
{2} 36. Kxf1 {-3.28/15 4} Rc8 {10} 37. Rd7 {-3.49/15 0} Rc1+ {5} 38. Kf2 {
-3.60/16 5} Rc2+ {1} 39. Kg3 {-3.61/16 13} Rxa2 {4} 40. Rxa7 {-3.47/13 2} Rd2 {
1} 41. Rd7 {-3.61/11 1} Nc4 {11} 42. Kf3 {-3.39/11 2} Nb2 {3} 43. Rd8+ {
-2.91/11 2} Kg7 {1} 44. d4 {-3.09/13 1} Nc4 {16} 45. d5 {-3.31/12 4} b5 {7} 46.
d6 {-3.69/13 0} g5 {7} 47. fxg5 {-3.75/16 1} Nxe5+ {2} 48. Kf4 {-3.68/13 1} Nc6
{6} 49. Rd7 {-3.61/11 1} Rd5 {3} 50. g3 {-3.59/11 2} b4 {3} 51. Rb7 {-3.63/13 2
} Rxd6 {3} 52. Kxf5 {-3.81/14 0} b3 {3} 53. Rxb3 {-2.84/11 2} Nd4+ {1} 54. Ke5
{-3.37/14 1} Nxb3 {1} 55. Kxd6 {-3.75/15 0} Nd4 {2} 56. h4 {-3.19/13 1} Ne2 {1}
57. Ke5 {-3.84/15 2} Nxg3 {1} 58. Kf4 {-3.93/15 0} Nh5+ {1} 59. Kf5 {-3.93/15 1
} Ng3+ {1} 60. Kf4 {0.00/24 0} Ne2+ {0} 61. Ke3 {-3.96/14 1} Nc3 {0} 62. Kf4 {
-4.55/14 0} Nb5 {0} 63. Kg4 {-3.99/14 1} Nc7 {0} 64. Kf5 {-4.11/13 1} Ne6 {0}
65. Kg4 {-5.24/14 1} Kg6 {0} 66. h5+ {-5.29/12 1} Kg7 {0} 67. Kh4 {-5.29/12 1}
Kf8 {0} 68. Kg4 {-5.13/13 4} Ke7 {0} 69. Kg3 {-5.23/11 1} Kd6 {0} 70. Kg4 {
-5.29/11 1} Ke5 {0} 71. Kg3 {-5.50/10 1} Kf5 {0} 72. Kh4 {-5.87/11 1} Nxg5 {0}
73. Kg3 {-5.36/8 0} Ne6 {1} 74. Kh4 {-5.87/10 1} Nf4 {1} 75. h6 {-6.06/11 1}
Kg6 {1} 76. Kg3 {-5.75/9 1} Kxh6 {0} 77. Kh4 {-6.35/10 1} Kg6 {0} 78. Kg3 {
-6.19/8 0} f5 {0} 79. Kf2 {-6.58/9 1} Kf6 {0} 80. Kg3 {-6.30/7 0} h5 {0} 81.
Kf3 {-6.82/9 1} Kg5 {0} 82. Ke3 {-12.79/10 1} h4 {0} 83. Kf3 {-12.79/10 0} h3 {
0} 84. Kg3 {-#16/8 1} Kh5 {1} 0-1
I tested now some games with 1.f4 and 1.Nf3 and these moves are obviously better and I had indeed some difficulties. I won 2, but lost 4 or 5, mostly due to horrible blunders, when I already reached a favourable ending. then when it's only a pawn plus in a rook ending you are virtually lost at this time control..
here is one win. when I had only 1.3 secs left rybka luckily gave up.
[Event "Blitz:5'"]
[Site "?"]
[Date "????.??.??"]
[Round "?"]
[White "Rybka 2.3.2a 32-bit"]
[Black "lem"]
[Result "0-1"]
[SetUp "1"]
[FEN "rnbqkbnr/pppppppp/8/8/8/8/PPPPPPPP/R1BQKBNR w KQkq - 0 1"]
[PlyCount "168"]
[TimeControl "300"]
1. f4 d5 {0} 2. b3 {-2.02/12 14} Nf6 {2} 3. Nf3 {-2.12/12 4} g6 {2} 4. e3 {
-2.09/12 2} Bg7 {1} 5. Bb2 {-2.17/13 22} O-O {2} 6. Be2 {-2.21/12 12} b6 {4} 7.
O-O {-1.99/11 11} Bb7 {4} 8. c4 {-1.97/10 7} Nbd7 {9} 9. cxd5 {-1.88/10 12}
Bxd5 {2} 10. d3 {-1.97/11 5} c5 {3} 11. e4 {-1.89/11 3} Bb7 {2} 12. Qd2 {
-1.82/10 6} Ng4 {14} 13. e5 {-1.76/10 3} e6 {6} 14. Ng5 {-1.80/12 8} Nh6 {3}
15. Bf3 {-1.98/12 38} Bxf3 {3} 16. Rxf3 {-1.99/12 9} Nb8 {4} 17. b4 {-1.97/10 5
} cxb4 {4} 18. Qxb4 {-2.02/9 1} Nc6 {1} 19. Qe4 {-2.27/11 4} Qd7 {15} 20. Rc1 {
-2.50/9 6} Rac8 {19} 21. Rc4 {-2.68/11 0} Na5 {7} 22. Rd4 {-2.76/10 4} Qc6 {6}
23. Qe1 {-2.81/10 4} Qc2 {11} 24. Rf2 {-2.59/10 3} Qc5 {7} 25. Ne4 {-2.30/10 3}
Qc6 {6} 26. Nd6 {-2.07/10 3} Rcd8 {5} 27. Ba3 {-2.13/10 3} Nf5 {10} 28. Nxf5 {
-2.29/10 1} exf5 {4} 29. Bd6 {-2.68/12 1} Rfe8 {18} 30. Qd1 {-3.65/15 1} Rc8 {7
} 31. Ba3 {-2.86/10 3} Bf8 {6} 32. Bxf8 {-2.99/11 3} Rxf8 {3} 33. Rd6 {
-3.07/12 3} Qc1 {6} 34. Qxc1 {-3.17/13 0} Rxc1+ {1} 35. Rf1 {-3.46/13 2} Rxf1+
{2} 36. Kxf1 {-3.28/15 4} Rc8 {10} 37. Rd7 {-3.49/15 0} Rc1+ {5} 38. Kf2 {
-3.60/16 5} Rc2+ {1} 39. Kg3 {-3.61/16 13} Rxa2 {4} 40. Rxa7 {-3.47/13 2} Rd2 {
1} 41. Rd7 {-3.61/11 1} Nc4 {11} 42. Kf3 {-3.39/11 2} Nb2 {3} 43. Rd8+ {
-2.91/11 2} Kg7 {1} 44. d4 {-3.09/13 1} Nc4 {16} 45. d5 {-3.31/12 4} b5 {7} 46.
d6 {-3.69/13 0} g5 {7} 47. fxg5 {-3.75/16 1} Nxe5+ {2} 48. Kf4 {-3.68/13 1} Nc6
{6} 49. Rd7 {-3.61/11 1} Rd5 {3} 50. g3 {-3.59/11 2} b4 {3} 51. Rb7 {-3.63/13 2
} Rxd6 {3} 52. Kxf5 {-3.81/14 0} b3 {3} 53. Rxb3 {-2.84/11 2} Nd4+ {1} 54. Ke5
{-3.37/14 1} Nxb3 {1} 55. Kxd6 {-3.75/15 0} Nd4 {2} 56. h4 {-3.19/13 1} Ne2 {1}
57. Ke5 {-3.84/15 2} Nxg3 {1} 58. Kf4 {-3.93/15 0} Nh5+ {1} 59. Kf5 {-3.93/15 1
} Ng3+ {1} 60. Kf4 {0.00/24 0} Ne2+ {0} 61. Ke3 {-3.96/14 1} Nc3 {0} 62. Kf4 {
-4.55/14 0} Nb5 {0} 63. Kg4 {-3.99/14 1} Nc7 {0} 64. Kf5 {-4.11/13 1} Ne6 {0}
65. Kg4 {-5.24/14 1} Kg6 {0} 66. h5+ {-5.29/12 1} Kg7 {0} 67. Kh4 {-5.29/12 1}
Kf8 {0} 68. Kg4 {-5.13/13 4} Ke7 {0} 69. Kg3 {-5.23/11 1} Kd6 {0} 70. Kg4 {
-5.29/11 1} Ke5 {0} 71. Kg3 {-5.50/10 1} Kf5 {0} 72. Kh4 {-5.87/11 1} Nxg5 {0}
73. Kg3 {-5.36/8 0} Ne6 {1} 74. Kh4 {-5.87/10 1} Nf4 {1} 75. h6 {-6.06/11 1}
Kg6 {1} 76. Kg3 {-5.75/9 1} Kxh6 {0} 77. Kh4 {-6.35/10 1} Kg6 {0} 78. Kg3 {
-6.19/8 0} f5 {0} 79. Kf2 {-6.58/9 1} Kf6 {0} 80. Kg3 {-6.30/7 0} h5 {0} 81.
Kf3 {-6.82/9 1} Kg5 {0} 82. Ke3 {-12.79/10 1} h4 {0} 83. Kf3 {-12.79/10 0} h3 {
0} 84. Kg3 {-#16/8 1} Kh5 {1} 0-1
Yes, that is often my experience; I reach a won ending, maybe up a pawn or a knight for two pawns or such, but with less than a minute left I can only hope to get a draw by repetition or insufficient material. You should probably try playing with a small increment so you can win easily won endings. At 5'+5" I would estimate that you need to be around 2300 to make an even score at knight odds against Rybka on a Quad (opening Nf3 or f4 with max. contempt and no-resign on) and perhaps 2150 on 32bit sp hardware.
now I tried pawn f7 odds at 5' 5'' which is of course much harder, but it's still fun and more like normal chess. I guess I can get some occasional draws but no wins.
I find it interesting, that you estimate the difference in hardware as much as 150 elo when comparing a quad and a 32 bit sp. you have more experience with this, but my guess would be much lower in man-machine games, because the speedup results in even better tactics and only slightly better positional play, but tactics against a human doesn't matter so much, since he is worse there anyway. therefore in comp-comp games this 150 elo difference might be right, as better tactics pay off there also. in these handicap games the human tries to avoid tactics anyway and more or less he can afford to do this.
I find it interesting, that you estimate the difference in hardware as much as 150 elo when comparing a quad and a 32 bit sp. you have more experience with this, but my guess would be much lower in man-machine games, because the speedup results in even better tactics and only slightly better positional play, but tactics against a human doesn't matter so much, since he is worse there anyway. therefore in comp-comp games this 150 elo difference might be right, as better tactics pay off there also. in these handicap games the human tries to avoid tactics anyway and more or less he can afford to do this.
Probably in comp vs comp play the hardware difference might be more like 200 Elo points in blitz play; I agree that with humans the differences from speedup are less. But don't underestimate the improvement to positional play from a big speedup; it's hard to prove or detect, as it just means slightly better moves every two or three turns, but it really adds up. Perhaps my 150 estimate is a bit too high for knight odds play. By the way, grandmaster Joel Benjamin came out with only one draw in three tries at pawn (f7) and two moves (White gets an extra move at the start) at 5'+10" against my quad, so if you can score more than a very occasional draw (with contempt set to 50) you can be very proud. I did have a book for this, but it is not worth as much as the free move I'm sure.
Yesterday in blitz I had a Knight + Bishop ending with a pawn up against Rybka's rook... The problem was that his only pawn was only two squares away from queening, and my bishop could not control the promotion square without getting crushed by the rook (that damned d5-pawn...! my bishop could only go from a8 to c6, but when the Rook went to c8 -- with pawn at a6 --, it was time to resign).
I tried to win against Rybka in 3+0 and after ~15 games I managed to do it :)
It's really harder than I thought, many time she finds a way to get enough counterplay to create problems and with that short time control I overlook some tactics in the middlegame and suddenly it's only equal which means that I lose :)
the game I won:
[Event "?"]
[Site "?"]
[Date "????.??.??"]
[Round "?"]
[White "Rybka 2.3.2z3 mp 32-bit bench"]
[Black "preferred, Change"]
[Result "0-1"]
[SetUp "1"]
[FEN "rnbqkbnr/pppppppp/8/8/8/5N2/PPPPPPPP/R1BQKB1R b KQkq - 0 1"]
[PlyCount "99"]
{16MB, 1.ctg, AthlonXP 1800} 1... d6 {0} 2. g3 {7} e5 {1} 3. Bg2 {10} f5 {1} 4. O-O
{4} Nf6 {1} 5. d4 {4} e4 {2} 6. Nh4 {0} Be7 {2} 7. Bh3 {3} g6 {2} 8. Ng2 {11}
O-O {1} 9. Bh6 {3} Rf7 {3} 10. Nf4 {6} c6 {7} 11. Qd2 {4} Na6 {1} 12. c4 {7}
Nc7 {2} 13. f3 {4} d5 {2} 14. c5 {3} Ne6 {2} 15. Bg2 {3} Nxf4 {2} 16. Bxf4 {1}
Nh5 {2} 17. Be5 {3} Bg5 {6} 18. Qc3 {3} Bf6 {8} 19. Rac1 {3} Bxe5 {1} 20. dxe5
{0} Re7 {2} 21. Rcd1 {3} Qe8 {5} 22. f4 {1} Ng7 {2} 23. Qa5 {4} Ne6 {3} 24. e3
{1} Qd8 {16} 25. Qc3 {0} Rc7 {5} 26. Rf2 {2} Qe7 {6} 27. Rc2 {5} b6 {3} 28.
cxb6 {1} axb6 {1} 29. Bf1 {1} Ba6 {4} 30. Bxa6 {1} Rxa6 {2} 31. b4 {0} Ra8 {5}
32. Qb3 {1} Rd8 {7} 33. a4 {2} Kg7 {7} 34. Rcd2 {0} Rcd7 {5} 35. b5 {2} cxb5 {6
} 36. Qxb5 {0} d4 {3} 37. Qxb6 {0} dxe3 {4} 38. Rxd7 {1} Rxd7 {1} 39. Rd6 {0}
Rxd6 {3} 40. exd6 {0} Qd7 {2} 41. a5 {10} Qa4 {5} 42. Qa7+ {1} Kh6 {5} 43. Qxe3
{0} Qxa5 {3} 44. g4 {2} Qc5 {3} 45. Qxc5 {3} Nxc5 {1} 46. h3 {0} Kg7 {1} 47.
gxf5 {0} gxf5 {1} 48. Kf2 {1} Kf6 {0} 49. Ke3 {2} Ke6 {1} 50. d7 {0} Nxd7 {1}
0-1
She gave up here, but I would win this easily, I had 24 seconds left.
Btw., what do you think, are such games against Rybka a good training? I guess it is, since Rybka plays more "human like" than other engine I know (the only "strange" move was 7.Bh3) and in the games I've seen some good ideas by Rybka...
It's really harder than I thought, many time she finds a way to get enough counterplay to create problems and with that short time control I overlook some tactics in the middlegame and suddenly it's only equal which means that I lose :)
the game I won:
[Event "?"]
[Site "?"]
[Date "????.??.??"]
[Round "?"]
[White "Rybka 2.3.2z3 mp 32-bit bench"]
[Black "preferred, Change"]
[Result "0-1"]
[SetUp "1"]
[FEN "rnbqkbnr/pppppppp/8/8/8/5N2/PPPPPPPP/R1BQKB1R b KQkq - 0 1"]
[PlyCount "99"]
{16MB, 1.ctg, AthlonXP 1800} 1... d6 {0} 2. g3 {7} e5 {1} 3. Bg2 {10} f5 {1} 4. O-O
{4} Nf6 {1} 5. d4 {4} e4 {2} 6. Nh4 {0} Be7 {2} 7. Bh3 {3} g6 {2} 8. Ng2 {11}
O-O {1} 9. Bh6 {3} Rf7 {3} 10. Nf4 {6} c6 {7} 11. Qd2 {4} Na6 {1} 12. c4 {7}
Nc7 {2} 13. f3 {4} d5 {2} 14. c5 {3} Ne6 {2} 15. Bg2 {3} Nxf4 {2} 16. Bxf4 {1}
Nh5 {2} 17. Be5 {3} Bg5 {6} 18. Qc3 {3} Bf6 {8} 19. Rac1 {3} Bxe5 {1} 20. dxe5
{0} Re7 {2} 21. Rcd1 {3} Qe8 {5} 22. f4 {1} Ng7 {2} 23. Qa5 {4} Ne6 {3} 24. e3
{1} Qd8 {16} 25. Qc3 {0} Rc7 {5} 26. Rf2 {2} Qe7 {6} 27. Rc2 {5} b6 {3} 28.
cxb6 {1} axb6 {1} 29. Bf1 {1} Ba6 {4} 30. Bxa6 {1} Rxa6 {2} 31. b4 {0} Ra8 {5}
32. Qb3 {1} Rd8 {7} 33. a4 {2} Kg7 {7} 34. Rcd2 {0} Rcd7 {5} 35. b5 {2} cxb5 {6
} 36. Qxb5 {0} d4 {3} 37. Qxb6 {0} dxe3 {4} 38. Rxd7 {1} Rxd7 {1} 39. Rd6 {0}
Rxd6 {3} 40. exd6 {0} Qd7 {2} 41. a5 {10} Qa4 {5} 42. Qa7+ {1} Kh6 {5} 43. Qxe3
{0} Qxa5 {3} 44. g4 {2} Qc5 {3} 45. Qxc5 {3} Nxc5 {1} 46. h3 {0} Kg7 {1} 47.
gxf5 {0} gxf5 {1} 48. Kf2 {1} Kf6 {0} 49. Ke3 {2} Ke6 {1} 50. d7 {0} Nxd7 {1}
0-1
She gave up here, but I would win this easily, I had 24 seconds left.
Btw., what do you think, are such games against Rybka a good training? I guess it is, since Rybka plays more "human like" than other engine I know (the only "strange" move was 7.Bh3) and in the games I've seen some good ideas by Rybka...
I think that Ktulu 8 has a very human-like style. Also like humans, it is very strong in the opening (possibly more so than any other engine), weaker in the middle-game (thus why it isn't so high on the rating lists), and very strong in the endgame compared with other engines. It also doesn't typically do "crazy" moves like Junior.
Yes, I think this is good training. You should be proud to win a game at 3-0, even GM Benjamin couldn't do it until he added one second increment (though he was playing on a 64 bit quad).
Powered by mwForum 2.27.4 © 1999-2012 Markus Wichitill