Author: Bill McGaugh
Date: 11:59:26 08/24/01
Go up one level in this thread
I have sympathy for Christophe in his attempt to explain the problem... A few games mean almost nothing... An example...chessmaster personality mg1 versus the default chessmaster at game in 5 on a p3-933: the first 100 game set: 49 to 51...in favor of chessmaster default... so...is chessmaster better? the next: 51 to 49...in favor of mg1... the next: 57.5 to 42.5 then: 47-53 then: 53.5 to 46.5 then: 55 to 45 then: 54 to 46 then: 58.5 to 41.5 then: 59.5 to 46.5 then: 50.5 to 49.5 the totals: 529.5 to 470.5 if I would have accepted the first 100 game set as the "truth", I would have concluded that mg1 is worse than cm...and imagine, if I had the time to go through all the pgns...I am sure there are some long strings of wins in a row for both sides...if any of those subsets were the one that a person viewed first, they would have been completely deluded about which program version was stronger... how can people possibly view a 9 game tournament as anything definitive...especially people that have played many computer chess games... among the high level programs, it is close to a coin flip...and a lot of flips are necessary to figure which side of the coin it is biased toward... addendum: 1000 games mg2 vs. chessmaster default..game in 5... 55.5-44.5 62.5-37.5 61-39 55.5-44.5 59-41 53-47 57-43 52.5-47.5 53.5-46.5 58-42 total: 567.5-433.5
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.