Author: Robert Hyatt
Date: 14:04:04 10/26/97
Go up one level in this thread
On October 26, 1997 at 13:32:45, Jari Huikari wrote: >This happened with an early version of 'Nero.' I played against it >and got it's queen trapped in d8. My Knight would have taken the >queen in next move. > >But Nero decided to move Qd8xc8! Taking it's own black bishop. >( It is cheaper than give away a queen for a knight, isn't it. :-) > >After the game I had to teach the program not to take it's own >pieces... > > >Does anyone else have these kind of funny experiences, when >developing a chess playing program? > > Jari I think the most amusing thing that happened to me was with Cray Blitz at the Mississippi State Championship in 1984. I had just re-written the parallel search code to use what is now called the PVS parallel search algorithm. We were running on a Cray XMP with 4 processors. After playing well for a couple of rounds, out of the clear blue sky, Cray Blitz played the move BxP check and mate. I looked at the move dumbfounded. My opponent did to, calmly played PxB and went on to win easily. What had happened was we reached a position with 3 legal moves. 3 of the processors searched the legal moves and found the values for those moves, while the fourth said "I'm mated". Unfortunately this score was backed up to the root. :)
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.