Author: Dann Corbit
Date: 11:22:05 06/28/00
Go up one level in this thread
On June 28, 2000 at 14:01:20, Daniel Chancey wrote: >Final results of the 60 min Chessmaster tournament. > >1.Bendorz +07 -02 =05 9.5/14 >2.Test6 +07 -03 =04 9.0/14 >3.CM5600 +05 -03 =06 8.0/14 >3.CMQueen+ +06 -04 =04 8.0/14 >5.CM5500JS +06 -05 =03 7.5/14 >6.CMNimzo732 +04 -04 =06 7.0/14 >7.CMTest +04 -06 =04 6.0/14 >8.CMEllwein +01 -13 =00 1.0/14 > >CMEllwein's only win was in the very first game of the tournament against Test6. > >Bendorz and Test6 can perform as well as CMNimzo732 and CMQueen+ and should be >considered as a possibility of winning the upcoming Chessmaster Personailty >Championships. > >There is a Chessmaster Rookies tournament going on involving CMQueen+, >CMQueen++, CMQTest, and CMNimzo732 Here are the current standings after Round >2 of 10. > >1.CMQueen++ +3 -1 =2 4.0/6 >2.CMQueen+ +2 -2 =2 3.0/6 >2.CMQTest +2 -2 =2 3.0/6 >4.CMNimzo732 +1 -3 =2 2.0/6 Let's imagine that these personalities can increase (or decrease!) the ELO of a program by as much as 50 ELO over the standard personality. You might find it amusing to run a mathematical simulation of 50 games and see what kind of a spread you get under assumptions of +/-50 ELO. I think you will be surprised. Put another way, if you have a set of controls where each personality is identical, and you change nothing but the "names" of the identical personalities, it would not be at all surprising to see a result such as the above. I am only pointing this out as a reality check. It takes a huge number of games between peers to establish a sensible ELO rating. Just for fun, feed your games to ELOSTAT and see what the spread is. I guarantee that you will find they are all completely uncertain as far as which one is stronger (well within one standard deviation).
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.