Author: Dann Corbit
Date: 17:02:08 10/04/99
Go up one level in this thread
On October 04, 1999 at 18:00:21, Sven Reichard wrote: >Dann, >I don't know if I exactly understand the data in that directory. The goal here >is to compare various programs based on their performance on a well-defined >problem set. If that can be deduced from your data it would be nice to provide >sort of a summary, like > >Program: Gary v. Beta1 >Hardware: Atari ST, 8 MHz, 256 Kb, Blitter >Options: 512 Bytes transposition table, no egtb >Time Ctrl: 1/M >--- >Tactics: 14/840, 1.7% >Positional: 192/256, 75.0 % >Endgame: 6/49, 12.?% >Opening (book off): 08/15, ... > >Maybe it's there, but I didn't have the time to read thru 1.2 Mb of data... :) Certainly my goal was different. * Run all commonly occuring EPD test suites at fairly long time controls (12 Minutes of Pentium II 300 MHz equivalent time) 1. Use this data to find out: A. What test suites have completely bogus answers (nearly all of them) B. What kind of problems do computers have a tough time with (sacrifices and positional moves -- anybody surprised?) We have two cases with this data: 0. Test results disagree with suite about evenly divided into: A. Test calculation is wrong (program fails to find) B. Suite is wrong (bug in test suite) If anything, there are more of these. 1. Test results agree with suite (usually reliable but not always). From this set we can create a more reliable (but still not perfect) test suite.
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.