AC Milan v Bologna: 10.3 points of disagreement
The fitted model thinks Bologna is 10.3 percentage points more likely than the best price in the market implies. Here is where that comes from, and why it is not a tip.
What the model is looking at
The rates come from a Dixon–Coles fit on Serie A results with a 240-day half-life, blended with a second fit on shots on target. It knows the two sides' attack and defence, home advantage, and the low-score correction. It does not know the team news, and team news is most of what separates a public model from a closing line.
Where the disagreement could come from
A 10.3-point gap on Bologna is either information the model does not have, or a price that has not caught up. Kick-off is 2026-10-28 17:30 UTC, so the market has time to move. The honest test is not the result: it is whether this price beats the closing line, and that is the number the trading desk records.
Dixon–Coles with time decay, blended with a shots-on-target model. The market's probabilities are the best available prices across bookmakers with the overround removed by the power method.
This is a disagreement, not a tip. Across thirteen leagues the closing line has forecast better than this model in every one of them, and the validation report prints by how much.