Monday, January 2, 2012

Mid-Season EPL Prediction Contest Update

Before the season started I invited readers to participate in a season-long English Premier League prediction contest. We are now at the mid-season break, which provides a convenient time to evaluate where things stand.

Before getting to the results, let me explain the methodology used to evaluate the predictions. I identified the final standings at the end of the 2010-2011 season as the baseline evaluation of skill. Skill refers to the ability to out-perform a naive baseline. The choice of naive baseline is arbitrary. Specifically, for each place in the table, I take the difference between that place and a predicted place and square it. So if you predicted Man City in 4th and they are in 1st, then the difference is 3, and squared is 9. I then add up the squared differences for all 18 places, and then take the square root of that sum. That gives a measure of the total error in the prediction.

Here are the results at mid-season:

keeperusa 17.0
Arthur 17.3
n-g 17.5
Max 19.0
Roger 19.4
f.a.e 19.5
Elijah 20.1
itzik 20.5
SKILL 20.6
Unknown 22.0
CNY Roger 24.2

You'll note that the standings are tightly packed and that the threshold for skill appears more generous than the Bundesliga contest, which makes sense. n-g won the contest last year and is putting in a good showing at mid-season, but keeperusa and Arthur leading the way.

Right now Bolton, Norwich and Blackburn account for the greatest errors in the predictions. Stay tuned!


Post a Comment