Is it possible beat to Rybka in a bullet game ? Yes. It is.

[bob]

Of course Pablo's play was impressive for 2 reasons. He managed to play a 1 minute game so good and he played a game so good creating such a nice closed position.

Have you seen videos on youtube about bullet games? These people move their hands very fast, so playing that fast is a matter of practice. Also, Father is very experienced in reaching these closed positions, so it's a matter of hit or miss (I assume he lost lots of games before playing this one, that's the only one we get to see.)

1-2 games out of 200 is my prediction....

I would bet closer to 1 out of 10, maybe even a bit better than that. Seen these kinds of players on ICC in the past. And some could do it almost every single game. See if you can find some games with captainbluebear on ICC although I am not sure they save bullet games. Might be cptnbluebear as well... This was a few years ago and many computer operators noplayed him "because he could draw at will against the best programs and hardware on ICC" They even put that in their finger notes.

Sorry, Bob. Are you really talking about games played on Bullet TC 1+0?!

Yep. I used to have a directory where I saved these kinds of games. I had several in the 1 0 - 2 0 - 3 0 range that went _way_ past 256 moves. I even had to get Tim Mann to fix xboard so that it would not crash on move 257.

I use playchess fairly regularly and I have watched several games between Super GM's with 3100's and Hiarcs in main playing hall and human with topmost rating winning the game is so so rare.

Note I said "draw". Although many programs back then would lose on time, but not Crafty. But it had plenty of trouble avoiding the locked positions. I worked on this off and on for years and the thing got fairly good at avoiding them. But never "perfect".

I'd guess 1 out of 50 based on Engine's results history.
(Though they win few games on time)

[George Tsavdaris]

Of course Pablo's play was impressive for 2 reasons. He managed to play a 1 minute game so good and he played a game so good creating such a nice closed position.

Have you seen videos on youtube about bullet games? These people move their hands very fast, so playing that fast is a matter of practice. Also, Father is very experienced in reaching these closed positions, so it's a matter of hit or miss (I assume he lost lots of games before playing this one, that's the only one we get to see.)

1-2 games out of 200 is my prediction....

I would bet closer to 1 out of 10, maybe even a bit better than that.

Sorry but this time you are wrong.
I have a large database with playchess games that contains his games from a period of around 1-2 years i think, and the ratio is about 1:47 of non-loses:loses for Pablo(Father).

I will try to search where i have this database because the above is from memory and find the exact statistics and games.....

[gerold]

Hello Pablo. What is your win,lost record with the computer
programs.

Best to you,

Gerold.

P.S. I am talking about the last 200 games you have played
The Programs are much better now so have you still been
able to keep up with their progress.

[Dr.Wael Deeb]

Hello Pablo. What is your win,lost record with the computer
programs.

Best to you,

Gerold.

P.S. I am talking about the last 200 games you have played
The Programs are much better now so have you still been
able to keep up with their progress.

I can answer instead of Pablo:
1-2 games out of 200....

[Father]

Hello Gerold.

In a bad performance 10%, in a godone 100%.
Depends of how many games I have played.

Many times for example I played an unique game against a new machine and draw or win in the first chance. Then if in this moment would end and not plays anymore the score were 50% - 100%; last examples were Fritz 11 and Hiarcs.

Now, for example this last bullet games against JML, were 10%, I tried 10 times and win. But a realistic performance is in range 10% - 30%.

Best,

Pablo

[George Tsavdaris]

Of course Pablo's play was impressive for 2 reasons. He managed to play a 1 minute game so good and he played a game so good creating such a nice closed position.

Have you seen videos on youtube about bullet games? These people move their hands very fast, so playing that fast is a matter of practice. Also, Father is very experienced in reaching these closed positions, so it's a matter of hit or miss (I assume he lost lots of games before playing this one, that's the only one we get to see.)

1-2 games out of 200 is my prediction....

I would bet closer to 1 out of 10, maybe even a bit better than that.

OK i've found the exact numbers:

Pablo(Father) scored 25 points out of 444 games in the database i have for the period of 2005-06 in Playchess.

25/444 means a total= 5.6%

All these in 2005. I guess with the big advance of hardware and the come of Rybka as also the improvement of the other programs this percentage should have decreased in the years after 2006.

BUT note and this is important, that ALL his wins(and they were many out of these 25 games) that contribute considerably in this 5.6%, were wins on time due to the lag factor of playing over the internet in Playchess server.

For example here is the final position of "Father-Pro Deo 1.1" which ended 1-0 with a win for Father on time, but the position is obviously 0-1.
[d]7r/2b5/4p3/3p4/N1pk1Pp1/Pp2p1P1/6K1/4B3 b - - 0 159

Another similar example "Father-Fritz 8" which again was 1-0 although again it's obvious who would win if there wasn't any lag delay(i don't know if it's a Chessbase bug but i guess it is, since i have lost also games on time with Shredder 8 as also other engines on Playchess, while in normal non-Playchess games i have never seen a single game lost one time by Shredder 8) for the moves of black.
[d]5rk1/5r2/8/1p1p3p/p1pPpb2/P1P3Pq/1P3RR1/4BK2 b - - 0 194


Note that his 5.6% is close enough with the score for example of Gata Kamsky or Nakamura on ICC for 2006:
Gata Kamsky, 2723 Elo FIDE on ICC in blitz time controls between 3+1 to 5+5:
as white : +2 =3 -44 (7%)
as black : +1 =2 -48 (4%)

Hikaru Nakamura, 2697 Elo FIDE who hold the record for highest human rating ever at blitz on ICC, in blitz time controls between 3+1 to 5+5:
as white : +1 =19 -80 (10%)
as black : +8 =16 -77 (16%)

[Uri Blass]

Of course Pablo's play was impressive for 2 reasons. He managed to play a 1 minute game so good and he played a game so good creating such a nice closed position.

Have you seen videos on youtube about bullet games? These people move their hands very fast, so playing that fast is a matter of practice. Also, Father is very experienced in reaching these closed positions, so it's a matter of hit or miss (I assume he lost lots of games before playing this one, that's the only one we get to see.)

1-2 games out of 200 is my prediction....

I would bet closer to 1 out of 10, maybe even a bit better than that.

OK i've found the exact numbers:

Pablo(Father) scored 25 points out of 444 games in the database i have for the period of 2005-06 in Playchess.

25/444 means a total= 5.6%

All these in 2005. I guess with the big advance of hardware and the come of Rybka as also the improvement of the other programs this percentage should have decreased in the years after 2006.

BUT note and this is important, that ALL his wins(and they were many out of these 25 games) that contribute considerably in this 5.6%, were wins on time due to the lag factor of playing over the internet in Playchess server.

For example here is the final position of "Father-Pro Deo 1.1" which ended 1-0 with a win for Father on time, but the position is obviously 0-1.
[d]7r/2b5/4p3/3p4/N1pk1Pp1/Pp2p1P1/6K1/4B3 b - - 0 159

Another similar example "Father-Fritz 8" which again was 1-0 although again it's obvious who would win if there wasn't any lag delay(i don't know if it's a Chessbase bug but i guess it is, since i have lost also games on time with Shredder 8 as also other engines on Playchess, while in normal non-Playchess games i have never seen a single game lost one time by Shredder 8) for the moves of black.
[d]5rk1/5r2/8/1p1p3p/p1pPpb2/P1P3Pq/1P3RR1/4BK2 b - - 0 194


Note that his 5.6% is close enough with the score for example of Gata Kamsky or Nakamura on ICC for 2006:
Gata Kamsky, 2723 Elo FIDE on ICC in blitz time controls between 3+1 to 5+5:
as white : +2 =3 -44 (7%)
as black : +1 =2 -48 (4%)

Hikaru Nakamura, 2697 Elo FIDE who hold the record for highest human rating ever at blitz on ICC, in blitz time controls between 3+1 to 5+5:
as white : +1 =19 -80 (10%)
as black : +8 =16 -77 (16%)

I believe that the positions are misleading because I guess that in some games pablo sacrificed material to avoid drawing by the 50 move rule and win on time and in case that the program had more time he could do a draw instead of trying to win by getting lost positions.

Uri

[George Tsavdaris]

Of course Pablo's play was impressive for 2 reasons. He managed to play a 1 minute game so good and he played a game so good creating such a nice closed position.

Have you seen videos on youtube about bullet games? These people move their hands very fast, so playing that fast is a matter of practice. Also, Father is very experienced in reaching these closed positions, so it's a matter of hit or miss (I assume he lost lots of games before playing this one, that's the only one we get to see.)

1-2 games out of 200 is my prediction....

I would bet closer to 1 out of 10, maybe even a bit better than that.

OK i've found the exact numbers:

Pablo(Father) scored 25 points out of 444 games in the database i have for the period of 2005-06 in Playchess.

25/444 means a total= 5.6%

All these in 2005. I guess with the big advance of hardware and the come of Rybka as also the improvement of the other programs this percentage should have decreased in the years after 2006.

BUT note and this is important, that ALL his wins(and they were many out of these 25 games) that contribute considerably in this 5.6%, were wins on time due to the lag factor of playing over the internet in Playchess server.

For example here is the final position of "Father-Pro Deo 1.1" which ended 1-0 with a win for Father on time, but the position is obviously 0-1.
[d]7r/2b5/4p3/3p4/N1pk1Pp1/Pp2p1P1/6K1/4B3 b - - 0 159

Another similar example "Father-Fritz 8" which again was 1-0 although again it's obvious who would win if there wasn't any lag delay(i don't know if it's a Chessbase bug but i guess it is, since i have lost also games on time with Shredder 8 as also other engines on Playchess, while in normal non-Playchess games i have never seen a single game lost one time by Shredder 8) for the moves of black.
[d]5rk1/5r2/8/1p1p3p/p1pPpb2/P1P3Pq/1P3RR1/4BK2 b - - 0 194


Note that his 5.6% is close enough with the score for example of Gata Kamsky or Nakamura on ICC for 2006:
Gata Kamsky, 2723 Elo FIDE on ICC in blitz time controls between 3+1 to 5+5:
as white : +2 =3 -44 (7%)
as black : +1 =2 -48 (4%)

Hikaru Nakamura, 2697 Elo FIDE who hold the record for highest human rating ever at blitz on ICC, in blitz time controls between 3+1 to 5+5:
as white : +1 =19 -80 (10%)
as black : +8 =16 -77 (16%)

I believe that the positions are misleading because I guess that in some games pablo sacrificed material to avoid drawing by the 50 move rule and win on time and in case that the program had more time he could do a draw instead of trying to win by getting lost positions.

Your theory is wrong.
I have watched 15 of the 25 games that he didn't lose and in none of them what you proposed happened.
I will look at the other 10 later.

In the specific 2 positions i gave he did also not sacrificed any material to avoid draw by 50 move rule. Here are the games:

[Event "3m + 0s, rated"]
[Site "?"]
[Date "2005.03.15"]
[Round "?"]
[White "Father"]
[Black "Fonzy, Pro Deo 1.1"]
[Result "1-0"]
[ECO "A03"]
[WhiteElo "2136"]
[BlackElo "2575"]
[PlyCount "317"]
[EventDate "2005.08.11"]

1. d4 Nf6 2. e3 g6 3. f4 Bg7 4. Nf3 O-O 5. Bd3 d5 6. O-O c5 7. c3 b6 8. Re1 c4
9. Be2 Bf5 10. Nbd2 Qc7 11. Nf1 Ne4 12. N3d2 Nd7 13. Nxe4 Bxe4 14. Bf3 f5 15.
Bxe4 fxe4 16. a3 Rac8 17. g3 e6 18. Bd2 Nb8 19. Qc2 Rf7 20. Red1 Nc6 21. Be1
Na5 22. Ra2 Bf8 23. Rd2 Bd6 24. Rg2 Rcf8 25. Ra1 a6 26. Rb1 Qe7 27. Bf2 Qf6 28.
Re1 Qg7 29. Re2 Qh6 30. Kh1 Qh3 31. Kg1 Kh8 32. Rd2 b5 33. Re2 g5 34. Kh1 h6
35. Be1 g4 36. Kg1 Kh7 37. Rd2 Kg8 38. Rdf2 Kg7 39. Nd2 Kh8 40. Nb1 Bb8 41. Qe2
Kg7 42. Qd2 Bd6 43. Qe2 Kg8 44. Qd2 Nb3 45. Qc2 Kg7 46. Nd2 Na5 47. Nf1 Kg8 48.
Nd2 Bc7 49. Nf1 Kh8 50. Nd2 Qh5 51. Nf1 Bd6 52. Nd2 Nc6 53. Nf1 Qh3 54. Kh1 Na5
55. Kg1 Rf5 56. Nd2 Kg8 57. Nf1 Kg7 58. Nd2 Kh8 59. Nf1 Kh7 60. Nd2 Be7 61. Nf1
R5f7 62. Nd2 Kh8 63. Nf1 Rf5 64. Nd2 Kg7 65. Nf1 Kg8 66. Nd2 R5f7 67. Nf1 Kh8
68. Nd2 Nb7 69. Nf1 Bd6 70. Nd2 Bb8 71. Nf1 Na5 72. Nd2 Bc7 73. Nf1 Kg8 74. Nd2
Nb7 75. Nf1 Nd6 76. Nd2 Nf5 77. Nf1 Kh7 78. Qd2 a5 79. Qc2 a4 80. Qd2 Bb6 81.
Qe2 Kg8 82. Qd2 Bc7 83. Qc2 Bd8 84. Qd2 Be7 85. Qc2 Bd6 86. Qd2 Kh8 87. Qc1 Bb8
88. Qd2 Bc7 89. Qc1 Bd6 90. Qd2 Be7 91. Qc1 Bd8 92. Qd2 Bf6 93. Qc1 Be7 94. Qd2
Kh7 95. Qe2 Bf6 96. Qd2 Bd8 97. Qe2 Bb6 98. Nd2 Bc7 99. Nf1 Bb8 100. Nd2 Kh8
101. Nb1 Bc7 102. Nd2 Bd6 103. Nf1 Kg7 104. Qd2 Kg8 105. Qe2 Ne7 106. Kh1 Nc6
107. Kg1 Kh7 108. Kh1 Ne7 109. Kg1 Kh8 110. Kh1 Nc6 111. Rg1 Na5 112. Rfg2 Kh7
113. Nd2 Kg7 114. Nb1 Kg8 115. Bf2 Kh7 116. Qe1 Nc6 117. Qd1 Ne7 118. Nd2 Nf5
119. Nf1 Bc7 120. Qe2 Bb6 121. Qe1 Kh8 122. Qe2 Kg8 123. Qe1 Bc7 124. Qe2 Kh7
125. Qe1 Bd8 126. Qe2 Be7 127. Qd2 Kh8 128. Qc1 h5 129. Qd2 Bd6 130. Qe2 Bc7
131. Qd2 Bd8 132. Qe2 Bb6 133. Qe1 Kg8 134. Qe2 Kg7 135. Qe1 Kh7 136. Nd2 Bc7
137. Nf1 Bd6 138. Qe2 Bb8 139. Be1 Bc7 140. Nd2 Bd6 141. Nb1 Be7 142. Rf1 Bd8
143. Rgg1 Bc7 144. Qg2 Qxg2+ 145. Kxg2 Nxe3+ 146. Kh1 Nxf1 147. Rxf1 e3 148.
Kg2 Kg6 149. Kg1 h4 150. Kg2 hxg3 151. hxg3 Rh7 152. Rh1 Rxh1 153. Kxh1 Rh8+
154. Kg1 Kf5 155. Kg2 Ke4 156. b3 cxb3 157. c4 bxc4 158. Nc3+ Kxd4 159. Nxa4
1-0


[Event "Rated game, 3m + 0s"]
[Site "Engine Room"]
[Date "2005.02.28"]
[Round "?"]
[White "Father"]
[Black "Machiavellian, Deep Fritz 8"]
[Result "1-0"]
[ECO "D00"]
[WhiteElo "1858"]
[BlackElo "2659"]
[PlyCount "387"]
[EventDate "2005.06.10"]

1. d4 Nf6 2. e3 d5 3. c3 Bf5 4. Bd3 Bxd3 5. Qxd3 Nbd7 6. f4 e6 7. Nf3 Bd6 8.
O-O c5 9. g3 Ne4 10. Nbd2 Ndf6 11. Qe2 O-O 12. Nxe4 Nxe4 13. Nd2 f5 14. Nxe4
fxe4 15. Rf2 Rc8 16. Bd2 Qf6 17. Be1 Qg6 18. Rg2 a6 19. a3 Rc6 20. Rd1 b5 21.
Ra1 Be7 22. Kh1 Rb8 23. Kg1 Qf5 24. Kh1 Qf6 25. Kg1 Rf8 26. Kh1 Qf7 27. Kg1 Qe8
28. Kh1 Rc7 29. Kg1 a5 30. Kh1 a4 31. Kg1 Rb7 32. Kh1 Qc6 33. Kg1 Bd8 34. Kh1
Rbf7 35. Kg1 Bb6 36. Kh1 Rf6 37. Kg1 Kh8 38. Kh1 Rh6 39. Kg1 Kg8 40. Kh1 Rff6
41. Kg1 Rfg6 42. Kh1 Rh3 43. Kg1 Ba7 44. Kh1 Rf6 45. Kg1 Qe8 46. Kh1 Rhh6 47.
Kg1 Rf8 48. Kh1 Rf7 49. Kg1 Qc6 50. Kh1 Bb6 51. Kg1 Rg6 52. Kh1 Rgf6 53. Kg1
Ba7 54. Kh1 Bb8 55. Kg1 Rg6 56. Kh1 Ba7 57. Kg1 Rf8 58. Kh1 Bb6 59. Kg1 Rc8 60.
Kh1 Rf6 61. Kg1 Ba7 62. Kh1 Rcf8 63. Kg1 Rh6 64. Kh1 Rb8 65. Kg1 Bb6 66. Kh1
Rb7 67. Kg1 Ba7 68. Kh1 Rc7 69. Kg1 Qe8 70. Kh1 Bb6 71. Kg1 Qc6 72. Kh1 Rf6 73.
Kg1 Ba7 74. Kh1 Rg6 75. Kg1 Rc8 76. Kh1 Rh6 77. Kg1 Kh8 78. Kh1 Rf6 79. Kg1 Rg8
80. Kh1 g5 81. Rd1 Rg7 82. Rd2 g4 83. Rd1 Rf5 84. Ra1 Bb6 85. Kg1 Kg8 86. Kh1
Qe8 87. Kg1 Rg6 88. Kh1 Qc6 89. Kg1 Rh5 90. Kh1 Kg7 91. Kg1 Qd7 92. Kh1 Rf5 93.
Kg1 Rf7 94. Kh1 Kg8 95. Kg1 Kh8 96. Kh1 Qc6 97. Kg1 Rgg7 98. Kh1 Rf6 99. Kg1
Kg8 100. Kh1 Rfg6 101. Kg1 Ba7 102. Kh1 Rf6 103. Kg1 Qd7 104. Kh1 Rf5 105. Kg1
Rg6 106. Kh1 Qe8 107. Kg1 Rh5 108. Kh1 Kh8 109. Kg1 Kg7 110. Kh1 Qc6 111. Kg1
Kg8 112. Kh1 Rf5 113. Kg1 Qb7 114. Kh1 Qb6 115. Kg1 Rh5 116. Kh1 Kg7 117. Kg1
Qa6 118. Kh1 Bb6 119. Kg1 h6 120. Kh1 Kh8 121. Kg1 Ba7 122. Kh1 Rg8 123. Kg1
Qb7 124. Kh1 Qc6 125. Kg1 Bb6 126. Kh1 Qe8 127. Kg1 Qd7 128. Kh1 Rg6 129. Kg1
Kg7 130. Kh1 Qc6 131. Kg1 Kg8 132. Kh1 Ba7 133. Kg1 Kf8 134. Kh1 Bb6 135. Kg1
Kf7 136. Kh1 Rf5 137. Kg1 h5 138. Kh1 Kg7 139. Kg1 Ba7 140. Kh1 Kh7 141. Kg1
Bb6 142. Kh1 Rg7 143. Kg1 Rg8 144. Kh1 Ba7 145. Kg1 Rgf8 146. Kh1 Kg7 147. Kg1
Bb6 148. Kh1 R8f6 149. Kg1 Ba7 150. Kh1 Kh7 151. Kg1 Kg8 152. Kh1 Rf7 153. Kg1
Bb6 154. Kh1 R5f6 155. Kg1 Qd7 156. Kh1 Qe8 157. Kg1 Ra7 158. Kh1 Rb7 159. Kg1
Ba7 160. Kh1 Rbf7 161. Kg1 Re7 162. Kh1 Qd7 163. Kg1 Rf5 164. Kh1 Rg7 165. Kg1
Qc6 166. Kh1 Rff7 167. Kg1 Rf8 168. Kh1 Bb6 169. Kg1 Qd7 170. Kh1 Ba7 171. Kg1
Rb8 172. Kh1 Qf7 173. Kg1 Qf8 174. Kh1 Qf6 175. Kg1 Rf7 176. Kh1 Qf5 177. Kg1
Rbb7 178. Kh1 Qg6 179. Kg1 Rf5 180. Rd1 Kg7 181. Rc1 Rb8 182. Rc2 Rff8 183. Kh1
Qf5 184. Kg1 Rf7 185. Kh1 Kg8 186. Kg1 c4 187. Kf1 Rbf8 188. Rg1 Bb8 189. Qg2
e5 190. h4 gxh3 191. Qxh3 Qxh3+ 192. Rgg2 exf4 193. exf4 Bxf4 194. Rcf2 1-0

[gerold]

Hello Gerold.

In a bad performance 10%, in a godone 100%.
Depends of how many games I have played.

Many times for example I played an unique game against a new machine and draw or win in the first chance. Then if in this moment would end and not plays anymore the score were 50% - 100%; last examples were Fritz 11 and Hiarcs.

Now, for example this last bullet games against JML, were 10%, I tried 10 times and win. But a realistic performance is in range 10% - 30%.

Best,

Pablo

Thanks for your reply.
Is it harder to get a win on the new programs. Or is it about the
same as a couple years ago.

Best to you,

Gerold.

P.S. I know you have played some programs that is suppose to
play better vs. the stonewall. I have watch you play vs. some of
them are they all that easy to beat.

[Ovyron]

Note that his 5.6% is close enough with the score for example of Gata Kamsky or Nakamura on ICC for 2006:
Gata Kamsky, 2723 Elo FIDE on ICC in blitz time controls between 3+1 to 5+5:
as white : +2 =3 -44 (7%)
as black : +1 =2 -48 (4%)

Hikaru Nakamura, 2697 Elo FIDE who hold the record for highest human rating ever at blitz on ICC, in blitz time controls between 3+1 to 5+5:
as white : +1 =19 -80 (10%)
as black : +8 =16 -77 (16%)

Ah, but these GMs' wins are real:

http://www.chessgames.com/perl/chessgame?gid=1497429