Soccer as a Markov process: modelling and estimation of the zonal variation of team strengths.

This article focuses on modeling soccer games as Markov processes by discretizing the pitch into zones and defining states based on ball location, possession, and score. Using play-by-play data from the 2015 Japan League Division 1 season, the study develops three Markov models with increasing pitch discretization levels—one, three, and nine zones—and employs log-linear models to estimate team strengths in scoring, conceding, and possession transitions while accounting for home advantage. The findings indicate that the most detailed model (nine zones) provides a good approximation of real game dynamics and reveals that home advantage primarily arises from increased possession time near the goal rather than enhanced goal-scoring ability. The study highlights Sanfrecce Hiroshima’s superior offensive and defensive strengths as an example and suggests that this modeling approach can offer nuanced insights into team characteristics and inform strategic analysis in soccer.