Decision map
Every club placed by luck per game and change in strength. The shaded zones are the verdict rule; click a dot to focus that club.
Luck map
Actual points vs expected points (xPts) for every club so far. Above the band, results flatter the team; below it, the team deserves more.
Points vs expected, game by game
Cumulative points against cumulative xPts, with each game's luck underneath. A widening gap is luck accumulating, and history says it does not last.
Underlying strength (Elo)
Elo rating after each game, against today's top-4 and drop-zone levels. This is the signal a board should watch: it predicts the rest of the season better than points do.
Where the season ends
Distribution of final points from 5,000 simulations of the remaining fixtures, using current Elo ratings.
Table, with the luck stripped out
Click a column to sort, a row to focus a club. Strength rank is where Elo alone would place the club; projection and probabilities come from the simulation.
Why luck doesn't last: the evidence
Luck in the first half vs the second
Each dot is a club-season. If luck were a skill, the averages would climb along a diagonal. They stay flat at zero.
What predicts the second half?
Share of variance in second-half points per game explained (R²) by each midseason signal.
Is the model honest?
Predicted vs observed home-win rate on matches the model never saw (2019-20 onward).