Author: Dann Corbit
Date: 18:49:07 10/05/01
Go up one level in this thread
On October 05, 2001 at 17:29:28, Slater Wold wrote: [sniup] >>Ideally, a test problem should: >>1. Not be stupendously easy. 3 ply solutions are not worth the bother. >>2. Not be stupendously hard. All day solutions are not pragmatically useful. >>3. Have a single *clear* answer that is found after a fairly difficult search >>of at least one second but not more than ten minutes. (There could be special >>case sets that take days to solve, but they won't help you to improve your chess >>program much. Only to know that you have improved it after the fact). >> >>The very best problems are those that can be proven -- the solution move wins >>and other moves lose. Probably, there are very few of these problems known. > >Dann if you believe all this, then I think it's time to move on from WAC. > >80% of WAC's problems can be solved in 3 seconds. 19% can be solved in 30 >seconds. And 1% cannot be solved in 100,000 seconds. For most engines, 95% can be solved in under 5 seconds. For the best engines, it is less than ten problems that cannot be solved in 5 seconds on a 950 MHz machine. Yet *most* of them are not what I would call stupendously easy. >Perhaps that 19% should be focused on, while the others are of no meaning, to >most anyway. The fairly easy ones also serve a purpose. WAC is a quick tactical check to see if you have screwed up your eval. And for the few tough problems at the top end, there is still a bit of a technological challenge. >GCP and I just went through his EPD, and any problem that had 2 solutions, was >removed. Any problem that had a transposition into the winning line, was >removed. There are probably still some left, but most have been removed. I would like to get a copy of the finished test set. >If anyone ever gets serious about it, I would donate my computer for 14 hours a >day to crunch positions, to come up with a "standard" EPD. WAC is not a >challenge anymore, and perhaps those developing chess programs now with the goal >of solving all the WAC problems, are not striving hard enough. It's like >running a mile in under an hour. That's *not* an accomplishment. I would like to get a copy of the EPD output your computer gets when you solve epd problems. For instance, if you issue: epdpfga foo.epd foo.out email the foo.out file to me. Your machine can produce excellent data that would be highly useful for the CAP project.
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.