In this position
Rybka plays h7 or d3 with evaluation draw.
Rybka did not see : 1 h7 a4 - 2 h8 (Knight promotion) b4 - 3 Cg6 !
Is it a rybka bug ?
8/5p2/5P1P/pp1p1P2/3P4/kp6/1p1P4/1K6 w - - 0 9
Rybka plays h7 or d3 with evaluation draw.
Rybka did not see : 1 h7 a4 - 2 h8 (Knight promotion) b4 - 3 Cg6 !
Is it a rybka bug ?
It seems that Rybka doesn't find it. I tried up to depth 42.
Since I was testing these kind of positions in IDeA, I had it analyze this position with Rybka 3 1-cpu. It found the knight promotion and a mating score in a few seconds.
Since I was testing these kind of positions in IDeA, I had it analyze this position with Rybka 3 1-cpu. It found the knight promotion and a mating score in a few seconds.
Very curious: on my Core 2 Duo, only with MaxCPUs=1 does Rybka 3 32-bit find it!
Is it another form of the EGTB bug? (but the workaround doesn't help, here...)
Is it another form of the EGTB bug? (but the workaround doesn't help, here...)
But a guess R3 finds 10. (2. but your FEN start with move no. 9) h8N after h7 b4?
Analysis by Rybka 2.3.2a 32-bit :
10.h8N a4 11.Ng6 fxg6 12.f7 gxf5 13.f8Q f4 14.Qf7 f3 15.Qxf3
+- (6.07) Depth: 9 00:00:00 1kN
+- (18.01) Depth: 13 00:00:05 629kN
8/5p1P/5P2/p2p1P2/1p1P4/kp6/1p1P4/1K6 w - - 0 10
Analysis by Rybka 2.3.2a 32-bit :
10.h8N a4 11.Ng6 fxg6 12.f7 gxf5 13.f8Q f4 14.Qf7 f3 15.Qxf3
+- (6.07) Depth: 9 00:00:00 1kN
+- (18.01) Depth: 13 00:00:05 629kN
You're right.
With only one cpu it is working.
I think the problem may be bad transfer information between process or thread core.
I don't forget that rybka is always not able to under promote with bishop and the under promote knight was not here on the first Rybka releases.
With only one cpu it is working.
I think the problem may be bad transfer information between process or thread core.
I don't forget that rybka is always not able to under promote with bishop and the under promote knight was not here on the first Rybka releases.
Powered by mwForum 2.27.4 © 1999-2012 Markus Wichitill
