Is it possible to beat Rybka 3 in Bullet? Yes it is.

[Father]

Rybka 3 32-bit: 13.0 ply; 17kN/s Intel(R) Celeron(R) M CPU 410 @ 1.46GHz 1463MHz, (1 CPUs) 19:42

[Event "Rated game, 1m + 0s"]
[Site "Engine Room"]
[Date "2008.08.18"]
[Round "?"]
[White "Father"]
[Black "Ridvan79"]
[Result "1-0"]
[ECO "D00"]
[WhiteElo "2003"]
[BlackElo "2194"]
[Annotator "Restrepo,Pablo Ignacio"]
[PlyCount "37"]
[EventDate "2008.08.18"]
[BlackTeam "Rybka 3 32-bit: 13.0 ply; 17kN/s Intel(R) Cel"]
[TimeControl "60"]

1. d4 {0} Nf6 {B/0 0} 2. e3 {0} d5 {B/0 0} 3. c3 {0} Nc6 {-0.33/8 3} 4. f4 {
1-f3) 0} Bf5 {-0.51/8 6} 5. Nf3 {0} e6 {B/0 0} 6. Bb5 {1} Bd6 {-0.47/9 1} 7.
O-O {1} O-O {-0.48/10 0} 8. Re1 {(Sf3-e5) 0} a6 {-0.65/8 1} 9. Bd3 {(Lb5-e2) 1}
Ne7 {-0.66/7 1} 10. Qe2 {(Sf3-e5) 1} c5 {-0.74/8 6} 11. g3 {(Sb1-d2) 2} Bxd3 {-
0.82/7 7} 12. Qxd3 {1} h6 {-0.76/9 0} 13. Nbd2 {1} Qb6 {-0.67/8 1} 14. Nf1 {
-e4) 1} Nf5 {-0.97/6 6} 15. Re2 {(Sf3-e5) 1} Qc7 {-1.18/5 6} 16. Kg2 {(Sf3-e5)
1} b5 {-1.44/5 6} 17. Bd2 {(Kg2-g1) 1} Rad8 {-1.39/4 5} 18. a3 {(Kg2-g1) 0} Be7
{-1.48/4 6} 19. Be1 {(Kg2-g1) Zeit (Lag: Av=0.36s, max=0.5s) 1} 1-0
`[D]3r1rk1/2q1bpp1/p3pn1p/1ppp1n2/3P1P2/P1PQPNP1/1P2R1KP/R3BN2 b - - 0 19

[Tony Thomas]

I think its most likely an operator error. Look at his rating to begin with, Rybka is much stronger than 2100 even on a hardware that slow.

[Mike S.]

Look at his rating to begin with, Rybka is much stronger than 2100 even on a hardware that slow.

Since Rybka 3 is new, it is possible that the low rating is from another much weaker engine and he has just started to use Rybka 3, and/or the Celeron lost against the Core 2 quads a lot, before that game. So, the rating itself does not point me to a problem.

But I think you are right: We find extremely small depths in this game. Maybe the hash size either was much too small, or much too big and the computer swapped. This could also explain the loss on time. I tried other Rybkas at bullet time control in the engine room, and I do not recall one single loss on time.

I don't understand why the performance info gives 13.0 ply when the depths in the game are between 4 and 9.

[Tony Thomas]

Look at his rating to begin with, Rybka is much stronger than 2100 even on a hardware that slow.

Since Rybka 3 is new, it is possible that the low rating is from another much weaker engine and he has just started to use Rybka 3, and/or the Celeron lost against the Core 2 quads a lot, before that game. So, the rating itself does not point me to a problem.

But I think you are right: We find extremely small depths in this game. Maybe the hash size either was much too small, or much too big and the computer swapped. This could also explain the loss on time. I tried other Rybkas at bullet time control in the engine room, and I do not recall one single loss on time.

I don't understand why the performance info gives 13.0 ply when the depths in the game are between 4 and 9.

As I have stated many times, if someone plays Rybka vs Rybka matches on different hardware they are not doing anything different to a normal person competing in special olympics. I know that there are few people who are trying to refute opening lines, and trying to find new (better) lines, they are excluded from the comment above. I do not own Rybka, so I cannot comment how many plys it gets on a celeron. I have a celeron, and due to problems its running most of the time at around 1400 before it turns off, but Romichess gets around 12-18 plys in similar time controls. Isnt a Celeron M, the mobile version of Pentium? It's possible that the dude was playing from a laptop and dropped it while looking at pictures of Aria Giovanni. Sorry for comparing goat vs Fish.

[Ovyron]

I think its most likely an operator error.

Indeed, Rybka adds extra 75MB of RAM usage, probably the operator set more than its system had and it went overboard, causing memory swap.