Author: Christophe Theron
Date: 01:26:26 09/07/98
Go up one level in this thread
On September 06, 1998 at 14:13:47, Thorsten Czub wrote: >On September 06, 1998 at 06:58:33, Christophe Theron wrote: > >>The marvelous thing about computer chess is that anybody can say anything about >>any program. And it is hard to say if it is true or not, because verifying a >>proposition can take more than a week or a month of test games. >> >>In the case of Chess Tiger, what I can say is that version 11.5 would beat 11.2 >>at any time controls, given a sufficient amount of games. This has been tested >>before version 11.5 was shipped to testers. This kind of test is always done >>before I send a new version. This is the least I can do, so testers don't loose >>their time with an obviously bad version. >> >> >> Christophe > >The problem is that you proofed that 11.5 kills 11.2. But that this does not >tell us or shows us that 11.5 is really stronger. > >I agree completely that only a month or more later, you can know about. >You can find out by doing games against all kinds of opponents, or doing >test-suites , or doing all things together. >But using x vs. x+1 is no good method used alone IMO. I don't use this method alone. But this thread was only about Tiger 11.2 against 11.5. Christophe >Maybe we will learn one day more about evolution-algorithms and than understand >more about this. Different versions of a program are a kind of evolution. >Sometimes it goes up. sometimes down. Difficult to say if x related to x+1 is a >progress. difficult. I don't think that you can use the results played out >between the 2 very good/seriously. These results do not reflect IMO the >strength. they only show you that one version is different than the other and >that the different gives the version xyz an advantage over the other version. > >But is this advantage also a general strength progress ?! I doubt this.
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.