Author: Howard Exner
Date: 23:10:48 04/30/98
Go up one level in this thread
On April 30, 1998 at 16:52:02, John Stanback wrote: Also, I think that it >is useful have some "easy" positions for use with slow machines or >very fast testing, or for authors of new programs to use in judging >progress. I'm in favor of either keeping all the positions or only >tossing those solved in 1 second or less by the majority of the >commercial programs. I'm not sure how most programs use there first second of time but am guessing that some might do stuff that may restrict the full one second of calculation. For the new programs wanting easy positions they could just keep the 770 positions from ECM98.770(or some name like that). I do now think that 10 seconds is too long. I ran Rebel 8 on the first 100 positions of ECm98.770 for 5 seconds per position(K6-233). It found 88 out of 100. If something like 5 seconds were used it would also be easier to adjust for different hardware speed. Alphas =4, pent266-300 and Amd =5 and so on down the line (just guessing at these ratios but the KKup2 has good data on machine speed ratios). Of course many other programs will find more or less than 88 and these will also be different so out of every 100 maybe 30% will remain, resulting in a suite called ECM98.230 (or however many problems remain after deletions of the easy). It may be easier for testers to report the ones they don't solve in X seconds as that list will be smaller than the solved ones. Anyone not solved automatically stays in the next draft. Now that ECM98 is out there people can do what they will with it but I do believe that the programmers will come to some consensus to further reduce this suite to meet the requirements they originally set out to achieve. Namely, bug free and no easy "WAC-like" problems.
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.