Author: Terry Ripple
Date: 12:55:47 08/16/00
Go up one level in this thread
On August 16, 2000 at 13:49:40, Dann Corbit wrote: >A tournament of approximate peers is largely a crap-shoot. > >Anyone can win, including a program that isn't the best. > >Considering the recent CCC tournament where crafty won by a landslide, we might >consider crafty a favorite. > >But Shredder and Junior have had recent and very impressive results. > >Ferret has had some near misses. > >Who can discount Fritz? > >In other words, nobody knows who will win. That's why we go ahead and run the >contest. > >The stronger your program is, the higher probability it will win. So there is a >clear correlation between strength and outcome. But you could have a bad book >line come up by pure random chance and cost you the tournament, even though your >program is the strongest by 100 ELO. > >We often think of machines as completely deterministic -- like a light switch or >something of that nature. But they are not. IRQ's happen, memory gets paged, >all sorts of quasi-random events occur which can change the outcome of a >calculation. > >All the competitors are worthy and I think any of them can win it, though some >will have a much higher liklihood than others. Also, i would like to add that there are hardly enough games being played to determine a true winner. You could hold another tournament with the same opponents playing each other and the outcome can be different in the end! Regards, Terry
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.