Author: Robert Hyatt
Date: 18:51:47 11/15/00
Go up one level in this thread
On November 15, 2000 at 16:51:58, Uri Blass wrote: >On November 15, 2000 at 16:33:55, Robert Hyatt wrote: > >>On November 15, 2000 at 13:04:05, Uri Blass wrote: >> >>>On November 15, 2000 at 12:56:21, Robert Hyatt wrote: >>> >>>>On November 15, 2000 at 12:41:36, Peter Skinner wrote: >>>> >>>>>On November 15, 2000 at 11:16:24, Uri Blass wrote: >>>>> >>>>>>On November 15, 2000 at 11:12:30, Robert Hyatt wrote: >>>>>> >>>>>>>On November 15, 2000 at 09:52:03, Eduard Nemeth wrote: >>>>>>> >>>>>>>>This position is real from a game ( but not from Tiger ) ! >>>>>>>> >>>>>>>>[D]8/8/8/6Np/2N5/Pk6/6K1/8 w - a3 0 1 >>>>>>>> >>>>>>>>In this position played a old version from Fritz 5 (Fritz-Nemeth) Ne5? >>>>>>>>And the game was draw ! :) >>>>>>>> >>>>>>>>I give now this position Tiger 13.0 for test an Tiger 13.0 played Ne5 ? >>>>>>>> >>>>>>>>(With tablebases Tiger how about better) >>>>>>>> >>>>>>>>Eduard >>>>>>> >>>>>>> >>>>>>>Ne4 is a mate in 26, of course. But a program that doesn't >>>>>>>support tablebases is probably not going to have a clue here. >>>>>> >>>>>>Junior5.9 without tablebase have no problem to find Ne4 with +4.08 score in a >>>>>>short time. >>>>>> >>>>>>Uri >>>>> >>>>>Engine version : Rebel Century 2.01 >>>>>Hash table size : 28 Mb >>>>> >>>>>8/8/8/6Np/2N5/Pk6/6K1/8 w - - >>>>> >>>>>00:00 01.01 3.13 1.Ne4 Kxc4 >>>>>00:00 02.00 3.13 1.Ne4 Kxc4 >>>>>00:00 02.15 3.13 1.Ne5 >>>>>00:00 02.16 3.13 1.Nd6 >>>>>00:00 02.17 3.13 1.Ne3 >>>>>00:00 02.20 3.13 1.Nb6 >>>>>00:00 02.21 3.13 1.Nb2 >>>>>00:00 03.00 3.35 1.Ne4 Kxc4 2.a4 >>>>>00:00 04.00 3.05 1.Ne4 Kxc4 2.Kh3 Kb5 >>>>>00:00 04.01 3.05 1.Kg3 >>>>>00:02 05.00 3.05 1.Ne4 Kxc4 2.Nf6 h4 3.a4 >>>>>00:03 05.01 3.05 1.Kg3 >>>>>00:03 05.02 3.12 1.Ne6 Kxc4 2.Nf4 h4 3.a4 >>>>>00:05 06.00 2.71 1.Ne6 Kxc4 2.Nc7 Kb3 3.Nb5 h4 >>>>>00:06 06.01 2.85 1.Ne4 Kxc4 2.Nd2+ Kb5 3.Kg3 Ka4 >>>>>00:07 06.03 2.85 1.Kg3 >>>>>00:11 07.00 2.91 1.Ne4 Kxc4 2.Nd2+ Kb5 3.Kg3 Ka4 4.Nc4 >>>>>00:27 08.00 2.90 1.Ne4 Kxc4 2.Nd2+ Kb5 3.Kg3 Ka4 4.Nc4 Kb5 5.Nd6 Ka4 >>>>>01:15 09.00 2.79 1.Ne4 Kxc4 2.a4 Kb4 3.Nc3 Ka5 4.Kh3 Kb4 5.Kg3 >>>>>02:24 10.00 3.66 1.Ne4 Kxc4 2.a4 Kd5 3.Nc3 Kc6 4.Kh3 Kd7 5.a5 Kd6 >>>>>02:26 11.00 4.12 1.Ne4 Kxc4 2.a4 Kd5 3.Nc3+ Kd4 4.Kh3 Ke5 5.a5 Kd6 6.a6 >>>>>02:31 12.00 4.80 1.Ne4 Kxc4 2.a4 Kd5 3.Nc3+ Kd4 4.Kh3 Kc5 5.Kh4 Kc4 6.Kxh5 >>>>>Kb4 >>>>>02:40 13.00 4.81 1.Ne4 Kxc4 2.a4 Kd5 3.Nc3+ Kd4 4.Kh3 Kc5 5.Kg2 Kb4 6.Kh3 Kc5 >>>>>7.Kg2 Kb4 8.Kh3 >>>>> >>>>>Rebel Tiger 13.0 >>>>> >>>>>Move: Nd2+ (3/22) depth=16 >>>>>Score: 6.62 depth: 16 >>>>>Ne4 Kxa3 Kh3 h4 Ng6 Kb4 Nh7 Ka3 Ne5 Ka4 Nf6 Ka3 Nfg4 Ka4 Nd7 Kb4 >>>>> >>>>>Tiger takes nearly 1.5 mins to find Ne4, yet Rebel 2.01 analysis engine in the >>>>>Rebel 11.0 package finds Ne4 instantly. >>>> >>>> >>>>Try them against a tablebase program to see if they can _win_ the position. >>>>It is not easy. The eval of +4 shows they likely don't understand... >>> >>>Evaluation of +4 is usually a win and the opponent is not important. >> >>Not when one side has two knights and the other has a single pawn. Many >>programs don't call KNNK a draw. they call it +6 or more. And I can >>guarantee you it won't win. :) >> >> >> >>> >>>Some programs with no tablebases know that KNN vs KP is often a draw so they >>>have no problem not to evaluate KNN vs KP as a big advantage for white. >> >>The game he posted had +4. That isn't close to a draw, nor is it close to >>a mate. Which means the program really really doesn't know what it is into, >>most likely. > >programs that believe that KNN vs K is a win evaluate the original position as >+6 and not +4. > >The fact that some programs without tablebases can see +4 suggests that they >have no problem to win. > >Uri I disagree here. the tough ending is KNN vs KP. NN = +6, P often = -2 for overly optimistic programs... So +4 would be pretty natural here... I tried it on Crafty and with no tablebases I got +4.2 myself. I used to have code to recognize this, but with the tablebase available, several complained that the eval was sometimes wrong...
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.