Author: Uri Blass
Date: 01:52:43 04/01/04
Go up one level in this thread
On April 01, 2004 at 02:37:00, Andreas Stabel wrote: >I'm working in the Norwegian meteorological institute where we have a very >powerful multiprocessor / multimachine computer system to do weather >simulations. The system has been idle a couple of months now while new >models were developed, so I installed a chess program I have made and have now >been running it for seven weeks. > >The system is just an attempt at analysing the initial position to try to >develop alternate openings to established theory, but it has a complete >chess engine with all 5-man EGTBs and most of the generated 6-man too. > >I have been looking for an opportunity like this, so I have made my system >capable of distributing the search over multiple CPUs and machines and I >have tested the program extensively. > >Today when I arrived at work the program had just stopped with a draw score ! >I don't understand this - it had so far 1.e4 c5 as the best opening with a >0.37 advantage for white, but now it seem to have just stopped with score = 0. >The strange thing is that it has searched very deep and have hit the EGTBs a >lot - something which I wouldn't have thought possible. Perhaps the 20TB >hash table has been part of it. > >Can anybody help me explain this ? >All suggestions will be appreciated. > >Best regards >Andreas Stabel Yes You have a bug in your program. It probably prunes a lot of lines It espacially happen when the date is 1.4_:( Even without having the date 1.4 it is not hard to search to the end of the game if you do a lot of pruning and search at most 2 legal moves at every ply when cases when you search only 1 legal move are even more common. It means that you can get final result for every position but it does not mean that the final result is correct. 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.