Dixon–Coles is a statistical model for football scores published in 1997 by Mark Dixon and Stuart Coles. It builds on the idea (Maher, 1982) that each team's goals follow a Poisson distribution driven by its attacking strength, the opponent's defensive strength and a home-advantage factor. Dixon and Coles added two things. First, a correction (often written ρ) for low-scoring results — 0–0, 1–0, 0–1 and 1–1 — which a plain independent-Poisson model gets slightly wrong. Second, time-decay weighting, so recent matches count more than old ones. From the predicted chance of every scoreline you can add up home-win, draw and away-win probabilities. Limits: the model sees results, not injuries, line-ups or motivation, and newly promoted teams need special handling (WinSight uses a cold-start treatment). See /how-it-works. Dixon, M. J. & Coles, S. G. (1997), Applied Statistics.
What is the Dixon–Coles model?
18+ (21+ where required)
WinSight figures are model estimates for information only. They are not betting advice and do not guarantee any result. Football is unpredictable, so please treat these numbers with care.