Author: Joshua Lee
Date: 05:47:58 04/26/01
Awhile back i decided to see if fritz could solve all the test suites and if not how much time is needed to solve most. I found this link which has the main ones i am sure all of you have http://www.cs.cmu.edu/afs/cs/project/airepository/ai/areas/games/chess/test/0.html but it also contained some you definately don't for instance if you compare these with the ones at Dan Corbits site he has some that aren't in this file of 6000 and his are in epd format already so i hope you type fast or i may have them still on my computer. i put them into a big cbh file that my main computer is working on right now. Well this file has positions from Dvoretsky , Think like a GM , Jeno Ban, Kopec, Botvinnik's Computers chess and Long rang planning, My System, and others. so with these 6000 plus the mega's and the epd's from corbit this time around i manage to remove atleast 90% of the duplicates and went from 17,000 to around 10,800. The bad news is that there is this one test set from corbit that i guess was used for Zchess to learn from and i get more unsolved after 3 minutes than anything else. I intend on finishing this time around last time everything was wiped out by accident so that sucked but then again it was good because there were way too many doubles thanks in part to alot of the same positions are in several different files Mega , Bench3 , so on and so forth. I am at 5500 and with any luck will have all unsolved within a week or too and then i will run those at 9minutes and then 27minutes and by then i should be down to under 100 unsolved but i wouldn't hold my breath. To top it off there are quite a few that fritz wouldn't solve in hours but where other programs because they look at the positions differently they will get the move no problem. This is weird as one position which i figure Fritz to see that one move (the correct one) is better than the one it likes it would have to look 12ply plus the 15ply from that point so 27ply to find the correct solution where junior finds this quickly like Fritz but doesn't change it's mind after a minute i guess that's just how it is.... Well if you are wondering the point of all this i am trying to find out with fritz and other programs which few used in tandem are enough to solve almost all the problems and on which Processor would you need to have to find these moves in 3min. Right now it is looking like somewhere over 584Ghz hehe ahhh that would be nice.
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.