What is "Model vs Market"?
Model vs market is the gap between two probabilities: the one our model produces for an outcome, and the one implied by the bookmaker’s odds. When the gap is meaningful, the match is flagged for a closer look.
Implied probability, in one paragraph
Any bookmaker price has an implied probability baked into it. A price of 2.00 implies a 50% chance for that outcome. 1.50 implies 67%. 3.00 implies 33%. Margins (vig) push the implied probabilities for the three 1X2 outcomes to add up to a little above the 1.00 baseline — that is the bookmaker’s profit cushion.
Our model produces its own probability for each outcome from form, schedule, head-to-head and pre-kickoff signals. When we say "model probability for the home win is 56%" we mean exactly that — our number, not a guarantee about reality.
What "model vs market" actually flags
It flags a disagreement larger than ~3 points. If the market says 50% and our model says 58%, that 8-point gap is the kind of pattern that may be worth a closer look. We never wrap it as "value" because the market is right most of the time — that’s exactly why a real disagreement is interesting.
There are three honest readings of any flagged disagreement: 1) the model is seeing something the price has not absorbed yet (often happens with late lineup or weather news the model already had); 2) the market has information the model lacks; 3) it’s just noise within the model’s typical error band.
How to use the flag
Treat it as a starting point, not an instruction. The match report shows both numbers (model and implied) so you can see the gap at a glance, and the data-quality badge so you know how strong the model’s inputs are.
Pair the flag with the odds-movement read. If the market has been drifting toward the model’s side all day, that’s a stronger reading than a snapshot disagreement that was already closing.
What it never is
Not a guaranteed pick. Not a value bet in the formal sense. Not a tip. We publish the disagreement and the inputs so you can decide for yourself.