Author: Dann Corbit
Date: 10:33:10 04/08/05
Go up one level in this thread
On April 06, 2005 at 18:18:25, George Tsavdaris wrote: >On April 06, 2005 at 17:52:09, Dann Corbit wrote: > >>On April 06, 2005 at 13:03:04, Jorge Pichard wrote: >> >>>Deep Junior 8 is Shredder 9 Uci Killer, but still weaker than Shredder :-) >> >>You should not try to conclude too much from a single matchup. >> >>After 1000 games between the two, you will know for sure. > >He will never know for sure.......:-) > >> >>After 25 games, it is not enough. > > Not even after 1000 as you suggest. Only after an infinite number....... >So here comes the matter of when someone is happy with the result. What are his >expectations........ > Jorge's expectations may stop at 25 games with a big +- margin of error. Your >expectations may stop at 1000 games with a smaller margin of error. But the >certaincy comes after an infinite number of games. So someone that is a >perfectionist, would never know if Shredder 9 UCI is better than Crafty >8.11......... After 1000 games, you will have an error bar of around 25 Elo. So at least you will know if it is probably stronger, probably weaker, or about the same. After 25 games, you really cannot tell much. >>At more than 30 (which is something of a magical number statistically) the >>results start to shape up. > >Actually there isn't any magic number..........:-) http://www.google.com/search?hl=en&q=%22at+least+30+observations%22
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.