Author: Timothy J. Frohlick
Date: 14:00:07 04/16/99
Go up one level in this thread
On April 15, 1999 at 10:55:10, Steve Coladonato wrote: >I have noticed that many computer chess tournaments are run on differing CPU's. >I don't understand how the results of ProgramX runnning on a P200 against >ProgramY running on a P90 are meaningful. I think the hardware environment >affects the results quite substantially. For example, given Fritz5 running on a >486 66mz and Fritz5 running on a Pentium 450, I would think that a game analysis >done on the Pentium 450 at 30 seconds per move would be more thorough than a >game analysis done on the 486 at 3 minutes per move. > >Am I way off here? No, you are not way off. Most of us have two or more computers and there is usually a substantial difference in speed. For instance, I have a PII 133, a PII 166 and a PII 333. The 2.3 or so doubling in speed should allow a rating increase of at least 80 rating points. I don't know if this applies in all positions or at all time controls. In science you want to reduce as many variables as possible. Not much on this site is real science even though a large number of our contributors are real scientists. A certain number of our posters will post a string of games that give their pet program a large advantage either by giving the "pet" a faster processor or by screwing with the opponents book. Some statistically minded folks will then attempt to rationalize the outcome. If you play a series of twenty games and switch the processors you still can't make a case as to which program is better. That would take at least hundreds of games. Who has the time? Of most interest are test positions which are rather subtle. You can then use the same processor and node count to determine which program gets the correct answer in the shortest period of time. Tim Frohlick
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.