Reading a tennis match market
Point-based tennis models turn serve estimates and the scoring rules into match probabilities, with assumptions that need testing.
On this page
A point-based tennis model starts with a sequence of points. In its simplest form, the main inputs are the probabilities that each player wins a point on serve. Run those estimates through the scoring rules and the model can produce game, set and match probabilities from one distribution.
Why the scoring system amplifies everything
Move a player's serve-point probability from 65% to 68% and their game-hold probability moves further, because a game is a race that the server keeps restarting from an advantage. Move the hold probability and the set probability moves further still. By the time the calculation reaches the match, the change can be larger than the point-level input change.
This is why a point-based model can be sensitive to a small input change. The scoring system can amplify the change, but that does not by itself show that a market move is correct or excessive.
Where to check consistency
Set betting, total games and handicaps can all be calculated from the same point distribution. If the probabilities implied by those markets cannot be reconciled after allowing for margin and settlement rules, the board may be internally inconsistent. The comparison does not reveal how the bookmaker built each price, so it should be treated as a diagnostic rather than an edge.
What moves the inputs
Surface is one important input, and it is worth its own piece. After that: recent injury, altitude, ball type, and whether a best of five is being played by somebody whose fitness is in question.
What this does not tell you
The simplest point model assumes a constant service-point probability and independent points. Research has found departures from both assumptions. Score state, fatigue and matchup effects can matter, so the model is an approximation whose calibration must be checked rather than assumed.
Common questions
How can tennis match probabilities be calculated?
A common point-based model estimates each player's chance of winning a service point and runs those estimates through the scoring rules to produce game, set and match probabilities.
Why can a small change in serve move the model so much?
The scoring system compounds point-level differences through games and sets, so a small input change can produce a larger change in match probability.
Are set markets priced from match odds?
A set market and match market can be calculated from the same point model, but a bookmaker's method is private. Comparing both with one distribution is a consistency check, not proof of how either price was made.