An interesting article on prediction models

I found this PDF on how to predict the result of association football matches this week.
You can download a copy here.

Its a great insight into how you might go about implementing fuzzy models and refine the results.

The basic premise is that you categorize results into big win, little win, draw little loss and big loss.
You then group these outputs with the results of the previous matches between the actual teams that are due to play and that gives a prediction model.
Smart tuning methods are then implemented on top of this model and your away in a hat!!

Although the whole point of the approach is to simplify analysis I think it could be worth incorporating some home and aways as weighting.

But what is REALLY interesting in this paper is the last paragraph on the first page.

"In the practice of prediction making the football experts and fans usually make good decisions using simple reasonings
on the common sense level."

This gave me the idea that it could be worth investigating if the model could be used to predict betting trends instead of actual results.

Watch this space!