Author: Dann Corbit
Date: 09:12:27 08/17/99
Go up one level in this thread
On August 17, 1999 at 03:35:25, blass uri wrote: >On August 17, 1999 at 03:05:18, Jouni Uski wrote: > >>I ran complete 210 position marathon test in my 450 Mhz >>AMD with 1000s time limit. (Suite is from Fritz CD.) Results: >> >>LG2000 192 >>Fritz5.32 187 >> >>So we have new tactical reference! Most stunning LG2000 is in finding >>long drawing (check) combinations. Sometimes it finds 30 - 40 ply combis >>in seconds, when Fritz is totally helpless. >> >>Also average NPS is now a little higher than Fritz's: 550kN/S vs. 450kN/S. >> >>Only possible menace in tactical area is CM6000, but who runs whole >>marathon with it? >> >>Jouni >> >>PS. Of course LG2000 loses matches against Fritz, but now I am talking >> about tactics... > >I do not know if this tactical test proves which program is best at tactics. > >I think a better test is to find many tactical mistakes of computers from the >ssdf games and to do a test suite when the target is to avoid the errors. Uri's comment is very valid in this sense: "The proof of the pudding is in the eating." But LG2000 does appear to have a very special ability to find tactical things faster or better than many other programs. This too is undeniable. Before believeing any tactical result, make sure the test suite is debugged. Almost none of them are. Some of them have about 30% errors. Sometimes, you could even have a program score worse but in reality do better.
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.