Author: blass uri
Date: 05:34:37 10/08/98
Go up one level in this thread
On October 08, 1998 at 08:28:10, Robert Hyatt wrote: >On October 08, 1998 at 04:35:57, blass uri wrote: > >> >>On October 08, 1998 at 04:28:13, Kai Lübke wrote: >> >>>On October 08, 1998 at 02:03:56, blass uri wrote: >>> >>>> >>>>On October 07, 1998 at 23:33:53, James T. Walker wrote: >>>> >>>>>I gave Fritz 5.16 a position /p4K/P////k// W. I think that I have that figured >>>>>out right. Anyway, Fritz announces #14 after about 16 sec. But when you go >>>>>through the moves it announces #13 twice, #10 twice,#8 followed by #9 etc. It >>>>>jumps from #6 to #4 no matter what move I make in that position. After that it >>>>>counts on down to mate properly. It ends up with about a mate in 17. What >>>>>causes this phenom ? Crafty with the same position announces #17 on the first >>>>>move then while pondering it finds mate in 15. After that it counts down by one >>>>>to mate. That seems fairly normal. >>>>>Jim >>>> >>>>Fritz never knew to count >>>>It was designed to play chess and not to count the number of moves to mate >>>> >>>>I think that this is because of hash tables >>>>Maybe fritz found that some position leads to mate and remember it in the hash >>>>tables as mate without the number of moves >>>> >>>>After it go to the same position again in the search it evaluates it as >>>>checkmate without number of moves and this is the reason that it cannot count >>> >>>The "mate in 8, then on the next move mate in 9" stuff is something you often >>>see in Fritz engines (Hiarcs and Junior have the same problems sometimes). >>>Crafty has a special code that assures that a "mate in N" is never followed by >>>a "mate in N+k" where k>=0. >>>I'm just waiting for someone to find a position where Fritz will not be able to >>>mate because of this... :-) >>> >>>--- >>>Shep >> >>The main problem with fritz is that often the mate in N is only an illusion of >>fritz. >> >>It is a bad idea to use a special code for fritz telling it to look after mate >>in N only for mate in N-1 or less before solving this problem because the result >>may be: no mate found. >> >>Uri > > >I don't believe that fritz is ever *wrong* when it announces a mate. It might >say mate in 10 when there is a mate in 6 that can be played. But I have *never* >seen a program announce mate in N when there is no mate there, unless it is a >new program with bugs. I did not see it announce a mate when there is no mate there but I saw it announce mate in N when there is mate only in more than N I remember that I tried a simple position of king and rook against king (white king h1 rook g2 black king f3) and it announced mate in 10 when I think that there is only a mate in 18 Uri > >But *if* you find a mate in N, after your opponent plays a move, it will not >hurt a thing to search until you find a mate in N-1... even if you can't find >it time will stop you eventually, and in 99.99% of the times the N-1 mate is >there and can be found within the time limit...
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.