BGonline.org Forums

My apologies to Snowie

Posted By: Fabrice Liardet
Date: Saturday, 22 August 2009, at 9:33 a.m.

Some weeks ago I stated that Snowie's assessment "player A was a xx%-yy% favourite" was misleading, disturbed that I was with the fact that the values were obtained by aggregating all the errors in MWC, which could lead to someone being more than a 100% favourite over one match, while in that case 100% would clearly be a more accurate assessment.

But I was wrong. It has been answered to me that when the number of match grew larger, the average Snowie assessment would tend to the real match winning chances, under the assumption that Snowie's evaluation were perfect. I can now see the simple reason why this is true : since only the errors and the dice noise affect the MWC, the average match result equals the sum of all errors, plus the sum of all dice noise, divided by the number of matches. Since the expectation of the noise is zero by definition, the result will tend to the sum of all errors divided by the number of matches, which is exactly the average of Snowie's assessment. Sorry if I have somewhat confused the matter and made it sound more convoluted than it really was.