Serve points won per 100 above what the rest of the tour manages against the same opponents. A ranking of aces measures the schedule; this measures the server.
Opponent-adjusted. It is the excess over what the rest of the tour manages AGAINST THE SAME OPPONENTS, shrunk towards zero by n/(n+15). Without that adjustment the list rewards whoever never meets a good opponent: a Challenger regular topped the raw table.
Minimum 30 matches with serve statistics. Below that the figure is noise and publishing it would be faking precision. Each family of metrics carries its own counter — 200 matches is not 200 tiebreaks.
Window. Every match since 2024 — that is where serve statistics begin in this database; before that there are none.
One division at a time. A player belongs to the Challenger/ITF division when more than half of their matches since 2024 are played there. Mixing the two ranks schedules rather than players: the raw list of closers was headed by an ITF player at 100% over twenty matches.
Not a betting signal. It describes a player, not a price. Our own prediction engine is measured against the market and loses: 0.632 log loss against 0.593.
