Methodology

Methods

Dixon-Coles Poisson model

A published adjustment to the independent Poisson model of football scorelines (Dixon and Coles, 1997). It corrects the known under-count of low-scoring results such as 0-0 and 1-1, and lets recent matches weigh more than old ones.

Shin de-biasing (Shin, 1993)

A published technique for removing the systematic bias from a set of publicly available pre-match figures before they are read as probabilities, so that the numbers add to one and are not skewed by better-informed participants.

Weighted multi-path fusion

Several independent estimates of the same match — one from history, one from goal-scoring rates, others from publicly available pre-match information — are combined with weights rather than used one at a time, so that a single weak input cannot decide the answer on its own.

Calibration

Calibration asks a narrower question than accuracy: when the model says 70%, does the thing happen about 70% of the time? A model can be well calibrated and still be uncertain, and it can be right often while being badly calibrated.

What this does not tell you

A model estimate summarises what past data suggests about matches like this one. It is not a forecast of one particular match. Football results carry a large random component that none of the methods on this page removes.