Analysis
Tennis return statistics explained: what they measure
What tennis return statistics actually count, how the deuce and advantage sides differ, and why return games won beats return points won.
By CricketTaken EditorialPublished Analysis18 min read
A player wins forty two per cent of the points they return across a season, which sounds like a losing position, and finishes the year in the top ten. Another wins forty per cent and cannot break serve. Two numbers, two points apart, and completely different careers.
Tennis return statistics measure how a player performs on the points where the opponent is serving, and the headline figure, return points won, is the least informative of them. The tours publish four return categories: first serve return points won, second serve return points won, return games won, and break points converted. Each answers a narrower question than the others, they are not interchangeable, and the one that predicts results most reliably is the one that most coverage mentions last. Understanding what each actually counts, and what it silently assumes, is the difference between reading a match summary and understanding it.
The core problem is that returning is the only part of tennis where a player's numbers are mostly determined by somebody else. That single fact shapes everything below.
What a return statistic actually counts
Start with the raw definition, because the ambiguity is real.
A return point is any point in which the player in question is receiving. It begins with the opponent's serve and ends however the point ends. Return points won is the count of those the receiver won, divided by the total, expressed as a percentage.
That definition contains an important inclusion that people forget: a double fault is a return point won. The returner did not touch the ball, may have been standing in an odd position, and gets full credit. Across a season this matters, because facing servers who double fault frequently inflates a return figure without the returner having played a shot.
It also contains an important exclusion: nothing about how the point was won. A return winner off a second serve and a twenty five shot rally that ends with the server missing a forehand count identically. Return points won is a scoreboard measure, not a description of the returning stroke, and reading it as an assessment of technique is a category error.
The split into first and second serve versions exists because the two situations barely resemble each other. Against a first serve, the returner is frequently trying to survive the exchange and get the ball back into a neutral position. Against a second serve, the returner often has time to step in and take control of the point immediately. Combining them into one figure blends a defensive statistic with an offensive one and produces something that describes neither.
First serve returns and second serve returns are close to different sports
Consider what the returner is being asked to do in each case.
Against a first serve, the ball arrives at the top of the server's range, usually with a substantial margin of pace over anything else in the sport, often placed at a corner of the box. The returner has a fraction of a second, cannot take a full swing, and in most cases succeeds by making contact in front, blocking the ball deep and neutralising the server's positional advantage. Winning the point outright from that position is rare. The realistic ambition is to get to a neutral rally, which is a coin flip rather than a win.
Against a second serve, the geometry inverts. The server has one delivery left and a double fault costs the point outright, so the second serve is hit with more spin and less pace, landing shorter and sitting higher. The returner can move inside the baseline, take a full swing, and dictate from the first ball. Winning a majority of second serve return points is entirely normal at professional level, which is exactly why holding serve depends so heavily on first serve reliability.
The consequence for reading statistics is that the two numbers should never be averaged in your head. They measure different competencies and they respond to different opponents. A returner who is exceptional against second serves and ordinary against first serves is a specific kind of player: one who punishes any lapse in serving but cannot manufacture pressure against a server having a good day. A returner whose first serve number is unusually high is often doing something structural, standing several metres behind the baseline to buy time, and paying for it in court position later in the point.
The size of that gap is itself a statistic worth attending to. A player with a narrow gap is either unusually good at absorbing first serves or unusually poor at attacking second serves, and which of the two it is can be worked out from the absolute levels. The gap widens against big servers and narrows against players whose first and second deliveries are more similar, which is a genuine tactical choice some players make. Anyone interested in what determines the first ball itself will find the mechanics in the piece on how tennis serve speed is measured.
The deuce side and the advantage side are not the same problem
Every point in a game is served from one of two positions, and the returner stands in a correspondingly different place. The tours do not publish return statistics split by side, but the split exists in every match and it explains a great deal about why some players break more often than their aggregate numbers suggest.
The service boxes are mirror images, but the players are not. Against a right-handed returner, the wide serve on the advantage side pulls them off the court to their backhand, while the wide serve on the deuce side pulls them to their forehand. For most right-handers, one of those is a shot they can hurt the server with and the other is one they can only survive. A left-handed returner faces the same asymmetry reversed, which is why left-handers and right-handers can present very different return threats against the identical server.
Then there is the distribution of important points. The scoring system does not spread its decisive moments evenly across the two sides. Game point at 40 to 30 is played on the advantage side. Break point at 30 to 40 is played on the advantage side. Deuce is played on the deuce side and the point after it, the advantage point, is played on the advantage side. The rhythm of a close game therefore concentrates the game deciding moments on one side more than the other, and it is the side that supplies the wide serve into a right-hander's backhand.
The practical implication is that two returners with identical overall return points won can convert pressure at very different rates, simply because one of them is stronger on the side where the pressure lands. A commentator noticing that a player "returns better in the deuce court" is describing something real that no published table contains.
If you want to see this yourself, watch which side a server goes to on break point. Servers and coaches know the asymmetry as well as anyone, and the pattern of serve placement at 30 to 40 tells you what the server believes about the returner's weaker wing. The scoring structure that produces this concentration is set out in how tennis scoring works.
Why return games won is the number that actually matters
Here is the central argument, and it is about the difference between counting points and counting the units the score is made of.
Return points won treats every point as equal. The scoring system does not. A game is won by taking four points with a margin of two, which means points have to arrive in specific combinations to be worth anything. A returner who wins three points in every service game they receive, spread evenly, wins no games at all. A returner who wins none in three games and four in the fourth breaks once. Identical return points won, completely different outcomes.
Return games won measures how often those points actually clustered into a break. It is the statistic that survives the translation from points to score, and it is what appears on the scoreboard at the end of a set.
The gap between the two figures is diagnostic. A player whose return games won is high relative to their return points won is someone who wins points at the right moments: they raise their level at 30 to 30 rather than distributing effort evenly, or they attack second serves in the specific games where the server is under pressure. A player with the opposite pattern is accumulating return points in games that were already lost, chipping away at 40 to love, and producing an impressive percentage that never becomes a break.
There is a further reason to prefer the games measure. Return points won is heavily contaminated by garbage points. In a game the server holds to love, four points are added to the denominator and none to the numerator, and in a game the server wins from 40 to love after a long deuce battle, the returner may add several points won without the game changing hands. Return games won ignores all of that. It asks a single binary question of each service game, which is exactly the question the set asks.
- Even distributor
- Clusterer
Illustrative. Both lines represent a hypothetical player winning the same total share of return points across ten service games received, distributed differently. The point is the distribution, not the values.
Show the numbers
| Item | Even distributor | Clusterer |
|---|---|---|
| Game 1 | 2pts won | 0pts won |
| Game 2 | 3pts won | 1pts won |
| Game 3 | 2pts won | 4pts won |
| Game 4 | 3pts won | 0pts won |
| Game 5 | 2pts won | 1pts won |
| Game 6 | 3pts won | 4pts won |
| Game 7 | 2pts won | 2pts won |
| Game 8 | 3pts won | 0pts won |
| Game 9 | 2pts won | 4pts won |
| Game 10 | 3pts won | 4pts won |
In the illustration above, both players win the same number of return points across ten games. The even distributor reaches deuce repeatedly and breaks rarely. The clusterer loses several games without threatening and takes four of them outright. On the scoreboard these are not close to the same player, and only one of the two published statistics can tell them apart.
Return statistics are opponent statistics in disguise
This is the largest single caveat and it deserves its own treatment.
A player's serving numbers are substantially under their own control. They choose the placement, the pace, the spin and the risk. Their opponent influences the outcome but not the delivery. Return numbers are the reverse: the returner does not choose what arrives, only what to do with it, and the distribution of what arrives is set entirely by whom they played.
That means a return figure is a joint product of the returner and their schedule. A player who spent a season on the Challenger circuit facing modest servers will post a better return points won figure than a player of similar ability who spent it facing the best servers on the main tour. Neither number is wrong. They answer questions about different populations of serves.
The same effect operates within a career. A player who reaches more late rounds faces better servers in those rounds, and their aggregate return numbers get worse as they improve. A player who loses early consistently faces a mix weighted towards the seeds they lose to, which is a different distortion in the other direction.
There is no clean fix available in the published data, but there is a discipline that helps: read return numbers against a specific opponent rather than in aggregate. Return points won against one named server across several meetings is a genuinely informative figure. Return points won across a season is a summary of a schedule as much as of a skill.
The reciprocal caution applies to serving numbers, and it is less severe but not absent. A server's hold percentage also depends on who they served to, and the reason it is less contaminated is that the range of returning ability across a professional field is narrower than the range of serving ability.
The surface effect, and why cross-surface comparison fails
Return numbers move more with the surface than with almost anything else, and they move in a direction that is entirely predictable from the physics.
A slower, higher bouncing court gives the returner more time between the bounce and the contact, and delivers the ball at a more comfortable height. Both make a controlled return easier and reduce the server's ability to win points outright from the delivery. A faster, lower bouncing court compresses the time and forces contact below the ideal height, which makes a neutralising return harder and a punishing one nearly impossible.
Clay sits at one end of that range and grass at the other, with hard courts occupying a wide band in between that varies by venue, by ball and by the specific court preparation. The consequence is that return points won for the same player, playing at the same standard, will be visibly higher on clay than on grass, and the difference is a property of the court rather than of the player.
Two secondary effects compound it. On a slow surface, second serves sit up more, so the second serve return number rises further than the first serve one. And on a slow surface rallies are longer, so a higher share of return points are decided by rally play rather than by the serve itself, which shifts the statistic away from measuring return quality and towards measuring baseline quality.
The practical rule is simple and widely ignored. A return statistic without a surface attached is close to meaningless for comparing two players, because the comparison is contaminated by where each of them chose to play. Two players with identical season figures, one of whom played a clay heavy schedule and one of whom did not, are not equally good returners, and the direction of the correction is knowable. The article on the differences between tennis court surfaces covers the bounce and speed characteristics that drive all of this.
- The serve is struckThe returner's numbers are already partly determined. Pace, placement, spin and the server's identity are inputs the returner did not choose and cannot change.
- The point is classified as a first or second serve returnThis is the split that matters most and it happens before anything else. The two categories are aggregated separately and should never be averaged back together.
- The point is played outEverything about how the point was won is discarded. A return winner and a thirty shot rally count the same, and a double fault counts as a return point won without the returner playing a shot.
- The point is added to a running totalOne point in a denominator that will hold tens of thousands by the end of a season. Nothing records the score at which it was played or which side it started from.
- The game resolves, or does notReturn games won is recorded here, as a binary. This is the only stage at which the clustering of points is captured, and it is why the games measure carries information the points measure cannot.
- The season aggregate is publishedSurface, opponent quality and schedule are all folded invisibly into a single percentage. Reading it without reconstructing those three things is the most common analytical error in the sport.
The chain a point travels through on its way into a published return statistic, and what is discarded at each stage.
Return position: the choice that moves every one of these numbers
Before the serve is struck, the returner has made a decision that will shape their statistics for the whole match, and no published figure records it.
Standing deep, several metres behind the baseline, buys time. The ball has travelled further and lost speed by the time it arrives, and the returner can take a fuller swing at a lower, more comfortable height. The cost is court position: the return is struck from a long way back, it takes longer to reach the server, and the returner begins the rally out of position and running forward. Deep returning tends to lift first serve return points won and suppress the quality of the position the returner ends up in.
Standing on or inside the baseline does the opposite. The ball is taken early, often on the rise, which is a considerably harder shot to execute, but a successful one arrives at the server's feet before they have completed their movement into the court. Aggressive returning tends to depress first serve return points won, because more returns are missed outright, while raising the value of the ones that land.
Neither is correct in general, and the best returners change position within a match and often within a game. The tell is that a player will retreat against a first serve and step in for the second, converting the same body of technique into two different jobs. Watch a returner's feet at 30 to 30 in a tight game and you will usually see the choice being made explicitly.
The statistical consequence is that return points won conflates a skill with a strategy. A player whose numbers are modest may be returning aggressively and losing points on purpose in exchange for winning better ones. That trade shows up in return games won, which is another reason the games measure carries information the points measure does not.
What return statistics cannot see
The published figures record an outcome and discard everything that produced it. Four things in particular are invisible.
Depth is the first. A return that lands a metre inside the baseline and a return that drops in the middle of the service box are the same event as far as return points won is concerned, and they are entirely different events as far as the point is concerned. The shallow return usually loses; the deep one usually neutralises.
Direction is the second. Returning crosscourt, down the line, or at the server's body are three different tactical decisions with three different risk profiles, and no published category separates them. A returner who is exceptional at going down the line against a wide serve has a weapon that will never appear in their percentages.
Contact height and timing are the third. Two returners with identical numbers may be taking the ball at completely different points in its flight, which determines what they can do next and how much the server has to respect them. This is closer to the actual skill of returning than anything the percentages capture.
The fourth is the effect on the server. A returner who is known to attack second serves changes the second serve that arrives: the server takes more risk, hits closer to the lines, and double faults more. That double fault appears in the returner's column as a point won without a shot played, which understates the returner's influence rather than overstating it. Return statistics credit a returner for what they did with the ball and give them nothing for what they made the server do before it was struck.
Ball tracking systems capture much of this, and teams work with it internally. Very little of it is published, which is why the public numbers remain a summary rather than a description.
From return games won to sets won
The reason return games won deserves priority over the points measure becomes clearest when you follow it one step further, into the score.
A set is decided by games, and the two players do not contribute to it symmetrically. The server is expected to win most service games at professional level, which means the ordinary path through a set is a sequence of holds, and the set turns on the small number of games where that expectation fails. One break, held to the end of the set, is usually enough.
Return games won is the direct measure of how often a player creates that event. A player who breaks in roughly one service game in six is a completely different competitive proposition from one who breaks in one in twelve, and the gap between those two rates can be produced by a difference in return points won of only a few percentage points, because the conversion from points to games is steep near the middle of the range.
That steepness is the structural reason the two statistics diverge. Winning slightly more return points does not produce slightly more breaks; it produces disproportionately more, because a small increase in the chance of winning each point compounds across the four points a game requires. The same arithmetic works in reverse, which is why a returner having a marginally poor day can look as though they have stopped competing entirely.
It also explains a familiar viewing experience. A set can feel close, with the returner reaching deuce in half the service games, and finish six games to three. The points were close and the games were not, and the games are what the scoreboard counted.
Sample size, and how little one match tells you
A tennis match is a small sample and a set is a tiny one. A returner facing a set won six games to love receives twenty four points at the arithmetic minimum, and a competitive three set match rarely supplies more than a hundred and fifty.
At those counts, ordinary variance dominates. A shift of five percentage points in return points won across a single match can be produced by three or four points landing differently, and three or four points can turn on a net cord, a line call, a gust of wind or a mistimed ball toss. The number is real but it is not evidence of anything about ability.
The same problem is worse for the split statistics. Second serve return points won within one match is computed over the number of second serves the opponent hit, which against a good server might be thirty or forty points. A single break of serve can move that percentage by several points.
There is a specific version of this trap in commentary. Late in a close match, a broadcaster will note that a player is winning some striking share of second serve return points, and the audience takes it as a description of how the player returns. Over the twenty or so relevant points in that match it is a description of what happened, and nothing more. The same player's season figure is a description of how they return, and it will be a different number.
The workable threshold is a season, or at least a surface swing within a season. A career figure is more stable still but blends together periods of very different ability, which is its own problem for a player whose game changed.
The composite ratings, and what they buy
The tours publish an aggregated return rating built from the four published categories, and similar composites exist elsewhere. They have a real use and a real limitation.
The use is ranking players against each other on a single axis, which is convenient and which the raw categories cannot do, since a player can lead one category and trail another. Aggregation forces a decision about relative importance and produces an ordering.
The limitation is that the decision is hidden. Adding percentages together implies a weighting, and the weighting implied by simple addition is arbitrary rather than derived from how much each component actually contributes to winning matches. A composite that adds a percentage of points to a percentage of games is adding two quantities measured in different units and treating them as commensurable.
None of this makes composites useless. It makes them a starting point rather than a conclusion. When a composite and its components disagree, meaning a player rates highly overall while trailing badly in one category, the components carry the information and the composite is smoothing it away.
The related caution is against using return numbers to predict a specific match. Return statistics describe a distribution of past opponents. A specific opponent is one point in a distribution, and if that opponent is unlike the average of the ones who produced the number, the number will not transfer. The ranking systems that determine who plays whom, and therefore what each player's return sample looks like, are covered in how the ATP and WTA rankings work.
How to read a return column during a match
A few habits turn the on-screen statistics from decoration into information.
Read the two serve splits before the aggregate. If the first serve return number is low and the second serve number is high, the returner is waiting for a lapse and the match will turn on the server's first serve reliability. If both are moderate, the returner is competing in rallies rather than punishing serves, and the match will be long.
Compare return games won against return points won as the set develops. A player accumulating points without breaking is losing the games that matter, and the pattern usually continues unless something tactical changes. A player breaking more often than their point share suggests is playing the important points better, which is a durable trait rather than luck.
Watch which side the pressure points are being played from, and where the server aims on them. That tells you what the server thinks of the returner's two wings, which is usually more reliable than any published number because the server has had a whole match to test it.
Discount everything by the surface before comparing two players. A returner posting decent numbers on grass is doing something considerably harder than a returner posting the same numbers on clay, and no broadcast graphic will tell you so.
And treat a single set as noise. The statistic that means something at the end of a match is the games column, because that is the one the scoreboard was built to count. The tiebreak, where the return points come thickest and the margins are smallest, has its own peculiar arithmetic set out in how tennis tiebreaks work.
More on serving, surfaces, scoring and rankings sits in our tennis section, and every explainer we have written across all our sports is indexed on the blog.
Common questions
What does return points won actually measure?
The share of points a player wins while receiving serve, counted across every point they returned. It is usually split into first serve return points won and second serve return points won, because those two situations are close to different sports.
Why is return games won a better indicator than return points won?
Because games are the unit that changes the score. Return points won treats every point identically, while return games won reflects whether those points arrived together and at the moments that finish a game. Two players with the same return points won can have very different break records.
Do return statistics distinguish the deuce and advantage sides?
The tours do not publish that split, but every point is played from one side or the other and the sides are not equivalent. The advantage side supplies the wide serve into a right-hander's backhand and carries most of the game deciding points, so a player's return can be materially stronger on one side.
Which surface produces the best return numbers?
Clay, as a rule. A slower, higher bounce gives the returner more time and a more comfortable contact point, so return points won rises. Grass does the opposite. Comparing raw return figures across surfaces without adjusting for this is the most common error in tennis analysis.
Does a low first serve return points won number mean a player returns badly?
Not on its own. It largely reflects who they played and how well those opponents served. Return numbers are opponent-dependent in a way that most serving statistics are not, which is why they should be read against the specific opponents faced.
How many return points are enough to judge a player?
Far more than one match. A single match usually supplies fewer than a hundred return points, which is small enough that a couple of net cords move the percentage by a visible amount. A season is a reasonable sample; a match is an anecdote.
Filed under Tennis·tennis · analytics · statistics · atp · wta · match analysis