Projections & Methodology

Every fixture now comes with a win probability and every season with a projected table. Where do those numbers come from, how much should you trust them, and why do two reputable models forecast the same league so differently?

These pieces open up the machinery: the inputs, the simulations, the assumptions and the uncertainty. The framing is analytical throughout — this is about understanding models, not following betting markets.

There is a running argument in these four pieces, and it is worth stating up front: a forecast is not a prediction of what will happen, it is a summary of what a specific set of assumptions implies. Change the inputs from results to expected goals, change how quickly the model forgets last season, change whether home advantage is one number for the league or one number per club, and the same fixture comes out at 58% instead of 64%. None of those choices is obviously wrong. That is exactly why two careful models disagree, and why a projected table is more useful as a description of uncertainty than as a claim about the future.

So the pieces here work outward from the machinery. Win-probability models covers the in-match version, where a single scoreline swings the number more than most viewers expect. Elo and SPI takes apart the ratings that feed the season-long simulations, and why projection models disagree traces the divergences back to the specific assumptions that caused them. Preseason form is the negative result of the set: an input everyone reaches for that turns out to carry very little.

What you will not find is a tipping page. Probabilities here are how the analysis expresses uncertainty, and nothing in this category is a recommendation to do anything with money.