Author: blass uri
Date: 10:22:59 12/18/99
Go up one level in this thread
On December 18, 1999 at 13:10:24, Albert Silver wrote: >On December 18, 1999 at 11:25:34, Chessfun wrote: > >>On December 18, 1999 at 05:51:30, allan johnson wrote: >> >>>On December 17, 1999 at 17:10:12, Albert Silver wrote: >>> >>>>I seem to have run across a bizarre problem and wonder if this might not be a >>>>bug in Fritz (5 or 6) or one of the infamous null-move problems. I was analyzing >>>>the following position which was part of some analysis done on >>>>Yakovich-Stefanova, Sevilla op, 1999: >>>> >>>>r4rk1/p3bppp/nn6/1N6/Ppp1P3/1P2BP2/N3K1PP/R2R4 w - - 0 18 >>>> >>>>In this position the indicated move is 18.Nxa7! Fritz never finds this (at least >>>>up to 14 plies) and I decided to check the lines which follow: 18...cxb3 and >>>>here again, Fritz sees nothing for White and the eval drops with each ply; yet >>>>as soon as I feed it the next move 19.Bxb6 its eval shoots up in White's favour. >>>>What gives? >>>> >>>> Albert Silver >>> >>>Albert Rebel C on my Celeron 433 plays Nxa7 within 4 seconds with a score of >>>0.80.The score, however, goes down over the next few moves.(Nxa7 Rxa7 >>>Bxb6 Rb7 Bd4 c3 Rac1 Rc8 f4 Rd7 0.24) >> >>Fritz sees this line as + 1.0 depth 11/11. > >I am amazed. What version of Fritz? I ran this through Fritz 5/16bit, Fritz >5.32, and Fritz 6 and they all prefer 18.bxc4. They never choose 18.Nxa7 at >that depth. > > >> >>>CM6000 plays Nxa7 also but with an initial eval of 0.53.It,too,drops after >>>Nxa7 cxb3 BxB6 bxa2 Rxa2. >> >>This appears the critical line. >> >>>Is it supposed to be a win for white? Fritz 5.32 >> >>Win !! I think not... >> >>>doesn't consider Nxa7 but maybe the move, as Paul says, is not winning. >>>Cheers Allan >> >>Fritz: A bug?. This is a strong heading for this thread. Why not "look at this >>position". >> >>Anyway...clearly Nxa7 is not the best move. Further Fritz's eval after move 19. >>Bxb6 "shoot's up" hmmm that's not what I get after playing the next two moves, > >That is two moves later. I'm talking about after 19.Bxb6. The point is that if >it thinks the position here is at +1.22 after 8 plies, +0.94 after 9 plies, and >+1.09 after 10 plies then it should have said this prior to 18.Nxa7 after 11, >12, and 13 plies respectively, but it doesn't. Before Nxa7 fritz believes that it is in a middle game and cannot see that Nxa7 is going to the endgame. It use evaluation based on the assumption that it will be always in the middlegame. After Nxa7 it understands that it is in the endgame and use the assumption that it is always in the endgame. Fritz does not know that it goes from middle game to endgame based on the search because it is a fast searcher and does not want to waste time to calculate if it is in the middle game or in the endgame. It calculates it only one time when it is in the root position. 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.