Author: Robert Hyatt
Date: 08:32:14 08/29/01
Go up one level in this thread
On August 29, 2001 at 11:22:14, Uri Blass wrote: >On August 29, 2001 at 09:33:08, Robert Hyatt wrote: > >>On August 29, 2001 at 08:29:28, Erkki Malkamaki wrote: >> >> >>While the information is both good, and old, it won't help the argument. >>A few are not going to let some facts get in their way when they want to >>bash the machine or the "team" of developers... >> >>:( > >The only facts that can convince me is an anlaysis of the logfiles. > >Can you show me one main line of deeper blue >that the best chess programs need many hours to get? I don't quite know how to do this, for many reasons: 1. We can't see the _entire_ mainline by DB. When the depth is reported as 11(7) we can see the PV from the first 11 plies (maybe more due to extensions) but we can't see the PV from the last 7 plies + the capture search. It is therefore hard to determine what the program "saw". 2. Many moves are made for positional considerations. Often a program will make the right move for the wrong reason, particularly if you try N different programs on the same position. You are almost guaranteed to find one program that will "get it right" but not understand why. If you have seen my post near the bottom of the Deep Blue discussion, you can see why _I_ think DB is far beyond DF/DJ. Because DT was at least as good, and we have good data on the differences between DT and DB2. > >My analysis suggest that deeper blue is 2-3 times faster than >Deep Fritz pIII800 in finding some main lines. > >I am sure that if we analyze every position >that Deeper blue pondered for some hours on fast PC we can find >more positions when top programs find main lines that >begin in the same moves as the main line of deeper blue. > >The interesting question is what is the speed ratio in these cases >If the speed ratio is the same as the speed ratio that I found then >it means that deeper blue is not strong. > >If in most of the cases when top programs find similiar main lines >the speed ratio is at least 100 then it means that deeper blue >was probably stronger then my impression. > >Unfortunately I found only one volunteer to help me to >analyze the positions that deeper blue pondered and >I am not going to use more than 20 hours of computer time >per week to analyze. > >Uri
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.