Author: Uri Blass
Date: 00:09:03 12/12/00
Go up one level in this thread
On December 11, 2000 at 20:21:38, Christophe Theron wrote: >On December 11, 2000 at 14:18:59, Enrique Irazoqui wrote: > >>On December 11, 2000 at 13:48:38, Christophe Theron wrote: >> >>>On December 11, 2000 at 07:45:46, Enrique Irazoqui wrote: >>> >>>>1r4qk/4Q1p1/5p1p/4pP2/pP5P/8/1P4P1/3R2K1 w >>>> >>>>Analyzing this position posted by Chen, Gambit Tiger sees the drawing b5 at ply >>>>13, -0.22, but 2 iterations later it switches to Rd7 even if it evaluates this >>>>move as -1.10, worse than b5. I have seen this other times too. Why is Gambit >>>>Tiger doing that? >>>> >>>>Enrique >>>> >>>>00:00:00.3 -0,34 7 123745 Rd7 Re8 Qc5 Qb3 Qc3 Qxc3 bxc3 Rc8 >>>>00:00:00.5 -0,34 8 169859 Rd7 Re8 Qc5 Qb3 Qc3 Qxc3 bxc3 Rc8 Ra7 Rxc3 Rxa4 >>>>00:00:01.0 -0,60 9 500265 Rd7 Re8 Qc5 Qb3 Ra7 Qxb2 Rxa4 Qb1+ Kh2 Qxf5 >>>>00:00:06.4 -0,60 10 1645814 Rd7 Re8 Qc5 Qb3 Ra7 Qxb2 Rxa4 Qb1+ Kh2 Qxf5 Ra7 >>>>00:00:09.5 -0,58 10 2530646 Qc7 Rc8 Qd7 Qc4 Qe6 Qc2 Ra1 Rd8 Qe7 Rd2 >>>>00:00:12.7 -0,86 11 3461939 Qc7 Rc8 Qd7 Qc4 Qb7 Kh7 Rd7 Qc1+ Kh2 Qf4+ Kh3 Qxf5+ >>>>g4 Qf1+ Qg2 >>>>00:00:13.6 -0,82 11 3722276 Rd7 Re8 Qc5 Qb3 Ra7 Qxb2 Rxa4 Qb1+ Kh2 Qxf5 b5 Rc8 >>>>00:00:26.5 -0,96 12 7284762 Rd7 Re8 Qc5 Qb3 Ra7 Qd1+ Kh2 Qg4 g3 Rd8 Qc2 Qxb4 >>>>Rxa4 >>>>00:01:23.6 -1,06 13 22947585 Rd7 Re8 Qd6 Qa2 Qd2 Qb1+ Kh2 Qxf5 b5 Qg4 g3 Rb8 Rd5 >>>>Rb6 >>>>00:03:17.0 -0,22 13 54439711 b5 Qb3 Rd8+ Rxd8 Qxd8+ Kh7 Qe8 Qe3+ Kh1 Qc1+ Kh2 >>>>Qf4+ Kg1 Qxf5 h5 Qb1+ Kh2 Qxb2 >>>>00:04:42.9 -0,22 14 78462887 b5 Qb3 Rd8+ Rxd8 Qxd8+ Kh7 Qe8 Qe3+ Kh1 Qc1+ Kh2 >>>>Qf4+ Kg1 Qxf5 h5 Qb1+ Kh2 Qxb2 >>>>00:07:34.9 -1,10 15 126099869 Rd7 Re8 Qd6 Qc4 Rc7 Qxh4 b5 Qd4+ Qxd4 exd4 b6 d3 >>>>Rd7 Rb8 b7 d2 Rxd2 Rxb7 Kf2 >>> >>> >>> >>>Well... I don't know! Apparently b5 has a fail-low at ply depth 15, and Rd7 is >>>chosen instead. >> >>I have seen Deep Fritz doing the same thing now and then. Paranoid >>interpretation: programs don't show their real evals anymore. > > >:) > >It's not a problem of paranoia. > >It's just that the way we deal with "fail lows" (moves that are suddenly found >to lead to a disaster). > >Instead of spending time in evaluating how worse things are going to be if it >plays the move, Tiger tries to find another one as quickly as possible. > >It's only if there is no better move that Tiger will try to find out the real >value of the move. I can understand that tiger prefers -1.10 for white and not fail low after -0.22 for white but I still think that the information is misleading. I guess that Tiger saw the fail low without solving it before finding Rd7 and it does not give information about it. The question is what is the best evaluation that is needed in order to force tiger to try to solve the fail low. I think that this evaluation should be written as the evaluation of b5 and not the number -0.22 There should be no main line after b5 when the program decided not to solve the fail low. This is exactly the case with Junior and when it has a fail high or fail low I can see the move without main line and there is a change in the evaluation by 0.3 if it fails high or -0.3 if it fails low(at least this is the case with all the Junior version before Junior6). using always 0.3 seems illogical to me and I do not know if Amir Ban improved it. Uri
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.