About

I've spent 35+ years working with data, statistics and modelling, in banking, insurance, manufacturing and retail, for small firms and large high-street institutions. Football gives me a far more interesting place to use them.

Why football

Most people learn statistics from coin tosses and dice. Football is better: everyone already has an opinion, the data is plentiful, and the answers are often surprising. Ask how likely a side is to score three, and you have already asked for a Poisson distribution.

Why the examples use Scottish data

I'm Scottish and I live here, so Scottish football is the football I know best. It's also the data I have: full results back to 2000/01 and my own prediction board. The methods work the same for any league, and other competitions come in where a comparison helps.

How the models are built

• Every model says what data went in, what it assumes and where it breaks.
• Predictions are probabilities, not promises. They are judged on log loss, Brier score and calibration, never on hit rate alone.
• Facts, model outputs, hypotheses and opinions are labelled as what they are.
• "There isn't enough evidence to answer this yet" is an acceptable conclusion.

What this site is not

It is not a scores site, a news site or a betting service. Nothing here is betting advice.

Work with me

I can help football clubs answer the questions they usually decide on gut feel: who to sign, which trialists to keep, who marks whom, whether a bad run is bad luck. The help is free for anyone involved in a club, whatever the league.