Author: Rob
Date: 09:32:07 10/21/97
Go up one level in this thread
On October 21, 1997 at 11:55:20, Amir Ban wrote: >On October 20, 1997 at 13:19:31, Bruce Moreland wrote: > >>This discussion died out. Did you guys arrive at any conclusions, or >>did you just beat on each other until you became exhausted? >> >>Did DB cheat in that game? >> >>bruce > > >We discussed the PV's and evaluations given in the printouts. The object >was to see if they are reasonable. We did reach some conclusions, yes, >and also discovered some interesting things about that position. Here is >a summary: > >The PV given by DB up to ply 10 starting with 36.Qb6 Qe7 37.axb5 Rab8 >38.Qxa6 e4 39.Bxe4 Qe5 evaluated +0.74: > >We found no tactical insights for this line. Positionally, mainstream >opinion is that the eval is way off. Rebel 9.0 evaluates it at +2.00. >Junior at +1.36, Hiarcs somewhere in between. CSTal is a dissenting >opinion with an eval of about +0.50. > >At ply 11 the eval was +0.48 with no PV. If the same line was considered >then the eval is even further off. Rob investigated lines other than >36...Qe7 that may match the eval more closely, and came up with this: > >36. Qb6 Rd8! > >Now the continuation as in the PV is much better for black because Bd6 >is protected. Say: 37.axb5 Rab8 38.Qxa6 e4 39.b6 Qe5 40.Qxc4 e3. This >position is still unclear and evaluated as positive for white on my >computer, but I would feel very uncomfortable as white and would take an >eval of +0.48 as natural. However: > >It took Junior only a few seconds to switch from 37.axb5 to 37.Be4!, >which locks the position, wins a pawn, evals as +1.4 and looks dead lost >for black. However Rob found this interesting line: > >36.Qb6 Rd8 37.Be4 a5! 38.axb5 axb4! 39.Rxa8 Rxa8 40.Rxa8 Qxa8 41.Qxd6 >Qa1+ 42.Kh2 Qf1. > >There's an echo of the final position here, and the result is the same, >although this position is much more complicated. E.g.: > >43.g3 bxc3 44.Bg2 Qd3! 45.b6 c2 draws, since 46.b7?? c1Q 47.b8Q+ Kh7 >48.Qe6 Qxg3+! 49.Kxg3 Qf4# > >or: > >43.cxb4 Qf4+ 44.Kh1 Qxe4 45.Qc5! Qe1+ 46.Kh2 c3 47.Qc4! Qd2 48.b6 c2 >49.b7 c1Q 50.b8Q+ Kh7 51.Qxc1 Qxc1 will draw. Neither 52.Qb6 nor 52.Qb5 >work since white cannot both guard against the perpetual and stop the >e-pawn. > >This variation is very interesting and maybe shows that Qb6 indeed does >not win, but since it's not credible that DB saw that far, I don't think >it's relevant. Except for the last paragraph everything is well summed up. But if you look at the position after Qa1 with Genius5, you will find that in the line mentioned above it fails-low after a relativly short amount of time and afterwards gives a near draw score. If you add the plys which lead to the Qa1 position to the number of plys it took for Genius5 to fail-low you are well within the search-horizon of DB. It is not relevant that in order to correctly assess the position as draw you have to look ahead many plys more. Remember Genius5 and also DB are not perfect ! > >My conclusion: > >1. The DB choice was based on evaluation. >2. That evaluation was "non-standard" for our field. >3. The analysis does not prove cheating, but is no great help in >disproving it either. The status quo remains. > >Amir My conclusion: 1. DB choice might have been based part on evaluation and part on extension-techique, especially singular extensions (Genius5 is one of the few micro-programs rumoured to have implemented SE in one form or the other). 2. That eval seems awfully close to Genius5's eval at one point. Even matching DB's eval EXACTLY at one point ! This might not mean very much, but it shows that it is well inside parameters in "our field". 3. The analysis also does not disprove astrology, UFO's and various conspiracy theories. - Robin -
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.