The favourite loses nearly one tennis match in three

The favourite loses nearly one tennis match in three. That held over 178,000 ATP and WTA singles matches since 2022, at the Grand Slam tournaments as much as on the ITF circuit, and it holds for the heavy favourites too: a player the bookmakers price at 90% still loses about once in 23. What the market does get right is the shape of the risk. When it says seven in ten, the player wins about seven times in ten, and if anything a touch more often.

Nearly one in three, at every level

The favourite loses at much the same rate wherever the match is played. At the Grand Slams it is just under three in ten. On the Masters and WTA 1000 level it is a third. The ITF circuit is where the favourite loses least often, and even there it is nearly three in ten.

Heavy favourites lose less often, and still not rarely: a favourite priced at 80% or more loses about one match in nine.

Grand Slam main draws are the safest place for a favourite: from the second round on it loses about one match in four. In qualifying it loses nearly one in three, as often as on the Challenger Tour.

What that means with a match on

The odds tell you how often the favourite loses, never which time. Odds of 1.25 say four wins in five, and over the whole sample that is what happened, band by band. A favourite priced at three in five won three in five, and a shade more.

So the risk is the right size. Nobody knows the outcome, the bookmaker included; what the price gives is the rate at which favourites of that price have lost.

Favourites beat their price, longshots fall short

Favourites win a touch more often than their price says, and longshots win less. In every five-point band the favourite beat its price, and the gap is widest among the near-certainties. Take the favourites priced just above nine in ten. The market gave them 92.2% on average. They won 95% of the time.

The same matches read from the other side show where the difference goes. Players priced at one in five or less were given 13.8% on average. They won 11.3% of their matches. At the very bottom the gap is wider still: those players won about two of every three matches their price promised.

Some of that gap is the bookmaker's cut rather than the crowd's optimism. The two prices of a match, turned into probabilities, add up to more than one, and the excess is the bookmaker's margin. At a Grand Slam the median margin is 3%. On the ITF circuit it is 6.9%, so the price of a forecast rises as the level drops.

Every figure here has the margin stripped out, by a method that takes more of it off the favourite than others do. Part of the favourite's edge may be that arithmetic. Either way, a bias of under three points is not one a bettor could live on.

Where the market struggles most

The market misses most in the middle tiers. A Brier score shows it: it charges each forecast the square of the gap between its probability and the result, then averages the charges. Zero is perfect foresight. A forecaster who calls every match a coin flip scores 0.25.

The closing line scores 0.201 over the whole sample. Backing the favourite blindly scores 0.312, worse than the coin, because certainty is punished in full every time the favourite loses.

The line is sharpest at the two ends of the pyramid. At the Grand Slams and at ITF level it scores 0.192. In between, at the Masters and WTA 1000 events, the rest of the main tour and the Challenger Tour, it scores 0.205 or worse.

The ends are not cleverer, only more lopsided. Slam and ITF fields hold more heavy favourites who duly win, and the average falls without the forecast improving.

What the data cannot say

The odds are Pinnacle's closing price and nobody else's; a consensus of books or an exchange would give slightly different probabilities. The opening odds sit unused in the files, so nothing above describes how the market moves between the draw and the first ball. Every match counts once, whatever was staked on it.

Coverage has gaps. The closing odds begin with the 2022 season and stop on 25 September 2026. The ITF rows for 2024 hold a third of a normal season's matches, and Davis Cup and Billie Jean King Cup ties are left out. The 620 lines that were exactly even have no favourite; they count in the Brier scores and not in the favourite tables.

Why FifthSet keeps score against it

FifthSet's Numbers strand publishes probabilities and grades each one on the forecast ledger once the result is in, by the method on the method page. The closing line is the yardstick because it is public, fixed before the first ball and set by a bookmaker who pays when it is wrong. A Brier score of about 0.2 over a season is the mark to beat.

The figures in one table