Author: Mike Hood
Date: 04:39:35 03/26/03
Go up one level in this thread
On March 26, 2003 at 04:49:19, Jouni Uski wrote: >With current hardware and excellent programs almost all test positions posted >here are solved by some programs - and in many cases stunningly fast even if >original poster says something like "impossible" or "very difficult". >My question: is it still possible to find test positions which no program is >able >to solve in lets say 60 minutes?! Let's forget Nolot positions and collect at >least >30 new "unsolvable by 2003" positions. > >Jouni I'm sure it is theoretically possible to find 30 such test positions, but there are practical problems in verifying the list. Assuming that 25 representive chess programs are used to test the test positions -- and it is feasible that more than 25 may be used -- it would take 750 hours, ie just over a month, to verify the list. Actually longer, because it would be sensible to run each analysis for longer than 60 minutes in order to check if a not-yet-released faster processor can solve the problem in less than 60 minutes. And if a test position fails, ie one of the programs solves it, a new test position has to be offered as a replacement, meaning another 25 hours or longer of testing. And every new version of a chess program means another 30 hours of testing. Maybe the time can be reduced by adding a second PC, but as you can see, putting together the list is a strenuous, time-consuming labour of love.
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.