Wednesday, May 22, 2013

Wrap Up: Season Long EPL Prediction Contest Results

As an exercise in humility and illustration of the concept of skill in forecasting I have run a few small prediction contests here over the past few years.

Here are the final standings of the 10 participants for 2012-2013:
Elijah 17.4
Megan 17.5
SKILL 17.8
Calvin 19.0
itzik 19.0
n-g 19.3
dave tombs 19.5
keeperusa 23.1
Arthur 23.2
Max 23.9
Pielke 24.0
Congrats to Elijah and Megan, the only two participants whose picks improved upon a naive baseline.  

The numbers above represent the square root of the sums of the mean-squared errors in the predicted standings. The biggest collective errors: West Brom (picked too low) and Newcastle (picked too high). Smallest errors: Man City, Arsenal and Reading.

What does this exercise tell us? The most important lesson, one that is learned in many walks of life, is that a naive prediction methodology can be very hard to beat. Only two entries outperformed an entry based on last year's final table.

The far more difficult question to answer, and one with an answer that has a lot of significance to decision making in areas far more important than sports prognostication -- are those 2 skillful entries showing skill because the forecasts had skill, or just luck?

For a discussion, see this post on hot hands and guaranteed winners.

Thanks all and congrats to the winners!


Post a Comment