The padel games-handicap predictor
Beyond who wins, our padel model now estimates the games handicap — the chance the favourite wins by more than a set number of games. Here is exactly how, and where it stops.
What a games handicap is
A padel match is a race of games inside sets. A games handicap of -3.5 means the favourite has to win by at least 4 games across the whole match for the line to "cover". The underdog covers the +3.5 side if they lose by 3 games or fewer — or win outright.
Bookmakers set that line roughly where covering is a coin-flip, so the handicap is really a question about the margin of victory, not just who wins.
How we estimate the cover chance
We don't fit a tidy formula. We tried — a simple straight-line model — and it failed an honesty check: for heavy favourites it predicted winning margins larger than a best-of-three match can physically produce, which would have meant fake 95%+ cover numbers.
So we use the real data instead. We group thousands of past matches by how big an edge our model gave the favourite, and for each group we count how often the favourite actually beat each margin. Because every number comes from a real, physically-bounded result, the cover chance can never exceed what genuinely happens.
Model versus market
On a padel match page you'll see both: our model's cover chance and the bookmaker's own implied chance for the same line, with the margin stripped out. They usually land close together — which is the honest result.
When they differ, it's almost always our model being a touch over-confident, not a hidden edge. The closing market is sharper, and we say so plainly rather than dressing a gap up as value.
What it can't do (yet)
For very lopsided matches we simply don't have enough history — the heaviest favourites all share one data bucket, so the model won't pretend to tell a 0.92 favourite apart from a 0.78 one. We'd rather be coarse than invent precision.
There's also no over/under games line for padel from our bookmaker source, so we don't show a totals prediction. And like everything padel, the data is thinner than our table-tennis leagues — treat the handicap read as a calibrated estimate, not a certainty.