Tennis first serve percentage: why higher is not better

A player walks off having landed 74 per cent of first serves, a figure that would have been read aloud approvingly at any point in the previous forty years, and they have lost in straight sets. Across the net someone landed 56 per cent and held serve all afternoon. The tennis first serve percentage column, taken alone, has just told you the exact opposite of what happened.

That is not a freak result. It is the ordinary behaviour of a statistic that measures reliability and gets read as though it measured quality. Landing more first serves is only an improvement if the serves you land are still worth something, and the usual way to land more of them is to hit them slower and closer to the middle of the box, which makes each one worth less. The percentage rises and the serve gets worse. Both things are happening at once, and only one of them appears on the graphic.

The right way to think about the number is as one term in a trade, never as a target in itself. The whole of professional serving is the management of that trade, and the arithmetic that governs it is not complicated.

What the percentage measures, and what it quietly leaves out

The calculation is as plain as it looks. Every point begins with a first serve. Count how many of those land in the correct service box. Divide by the number attempted. A let is replayed and disappears from both the top and the bottom of the fraction, which is one of several ways the let rule quietly reshapes the statistics without anybody noticing.

What the fraction contains is a single yes-or-no fact about each attempt. In or out.

What it does not contain is everything that decides the point. It has no room for pace, for spin, for placement, for whether the serve found the corner or sat up in the middle of the box, for whether the returner was pushed off the court or stepped in and struck a winner. An ace and a serve that is chipped back to the middle of the baseline are recorded identically. So are a body serve that jams the returner and a floated delivery that lands on the service line at three quarters pace.

That is a very large amount of missing information, and it is missing in a specific direction: everything the fraction ignores gets worse as the fraction gets better. The adjustments a player makes to land more first serves are precisely the adjustments that make each landed serve easier to deal with. A statistic whose omissions are correlated with its value is not a neutral summary. It is a misleading one.

• 2Serve attempts allowed on each point
• 2Faults required to lose the point outright
• 1Service boxes a serve may legally land in
• 0Points a let costs the server

Features of the two-serve rule as set out in the Rules of Tennis, not match data.

The number that actually settles the argument

There is a metric that answers the question people think the percentage is answering, and it is already on every official match page. It is total service points won: the share of all points on a player's serve that the server won, counting first serves and second serves together, aces and double faults and everything in between.

That figure is the only one that cares about the outcome rather than the process. It is also the one that determines whether the player holds serve, which determines whether they win sets, which determines whether they win the match. Nothing in the serving columns matters except through it.

Write it out and the structure becomes obvious. If a player lands a share of first serves and wins some proportion of those points, and loses the rest to the second serve where they win a different proportion, then:

Service points won = (first serve in rate × first serve win rate) + (first serve miss rate × second serve win rate).

Every serving decision a player makes moves at least two of those four terms, usually in opposite directions. Hit the first serve harder and flatter, and the in rate falls while the win rate behind it rises. Take pace off, and the in rate rises while the win rate behind it falls. The second serve terms barely move at all, because a second serve is a compulsory shot whose quality is set by technique rather than by strategy on the day.

So the sum has a maximum somewhere in the middle, and the maximum is not at the right-hand edge. That is the entire argument of this piece, and everything below is a working out of why the peak sits where it does.

A constructed illustrative model. First serve win rate is set to fall as the serve is made safer, second serve win rate is held constant, and total service points won is the resulting sum. The shape is the point, not the exact values.

Show the numbers

The trade-off has a peak, and it is not at the top

Item	Total service points won
40% first serves in	64%
50% first serves in	65.5%
60% first serves in	66.1%
70% first serves in	65.7%
80% first serves in	64%
90% first serves in	61%

Why one hundred per cent would be a catastrophe

Push the logic to its edge and the case makes itself.

A player can guarantee a first serve percentage of one hundred. Stand at the baseline, hit the ball at half pace into the middle of the box with a comfortable margin over the net, and it will go in every single time. The column will read 100 per cent. It will be the highest first serve percentage ever recorded at professional level.

It will also be the worst serving performance ever recorded, because every one of those deliveries is a free ball. The returner steps inside the baseline, hits through it, and takes control of the rally on every point. A player serving that way would not hold serve once in a match.

The reason this sounds absurd is that everybody already understands, intuitively, that the serve is a weapon and that reliability bought with pace is not free. The absurdity is useful because it locates the argument. If 100 per cent is bad and 0 per cent is obviously catastrophic, the best value is interior, and the only question left is where.

Nothing about that reasoning depends on the numbers being professional ones. It applies with equal force at club level, where the returner is worse but so is the serve, and the same interior optimum exists a long way below the top of the range.

The second serve's penalty structure is what sets the optimum

Here is the part that makes tennis different from almost every other risk decision in sport.

Missing a first serve does not cost a point. It costs the difference between playing the rest of the point from a first serve and playing it from a second. That is a much smaller penalty than losing the point outright, and it is the reason a player can afford to miss so many.

Quantify it in the loosest possible terms and the shape is clear. Behind a landed first serve, a professional wins a clear majority of points, because the returner is defending and the server is dictating. Behind a second serve, the same player wins something much closer to half, because the returner can step in and attack a slower ball. The gap between those two rates is the price of a fault.

Notice how the incentive follows. If the penalty for a miss were losing the point, the correct first serve would be a very safe one, because each fault would be catastrophic. Because the penalty is only the drop from first-serve conditions to second-serve conditions, the correct first serve is an aggressive one, and it is correct to miss it a great deal.

This is also why the optimum is personal rather than universal. A player with a heavy, high-bouncing second serve that wins close to as many points as a modest first serve has a small penalty for missing, so their optimum sits lower down the reliability scale and they should be swinging harder. A player whose second serve is a liability, sitting up in the middle of the box at low pace, has an enormous penalty for missing, so their optimum sits much higher and they should be taking pace off the first ball to avoid ever facing the situation.

Two players can therefore be serving perfectly while producing first serve percentages twenty points apart. Ranking them by that column is meaningless.

1. 1The server picks a target and a level of riskThis is the only real decision. Everything the percentage column later records is a consequence of how much margin the server left over the net and inside the lines.
2. 2The first serve landsThe server is now in the strongest position available in tennis. This is the branch that produces aces, unreturned serves and weak replies, and it is the reason the first serve is worth taking risk on.
3. 3The first serve missesNothing is lost yet. The point continues from a worse starting position, and the size of that loss is the entire cost of the fault. It is much smaller than losing the point.
4. 4The second serve must landRisk is now capped by the double fault. The server takes pace off, adds spin, and moves the target away from the lines, which hands the returner a ball they can step into.
5. 5The second serve landsAn ordinary rally begins on close to level terms, or slightly in the returner's favour if the second serve was weak. The server has lost most of the advantage that opening the point should confer.
6. 6The second serve missesDouble fault, point over. This outcome is rare enough that its direct cost is small, but the fear of it is what shapes every second serve that does land.

The decision structure behind every service point, and where the percentage column is set. The point-win rates referred to are patterns, not quoted figures.

Why a double fault costs less than it looks and more than it should

Double faults occupy a strange position. They are the most conspicuous failure in the sport, they are recorded in their own column, and they are far less important than the attention they receive.

The direct cost is one point. A handful of double faults across a match is a handful of points, and a player who wins two thirds of the points they serve can absorb several without much damage. Measured on the scoreboard, the double fault column is one of the smaller line items in a match.

The indirect cost is the one that matters, and it does not appear anywhere. Because the double fault is so visible and so embarrassing, players and coaches over-insure against it. The second serve gets slower and safer than the arithmetic justifies. And because the second serve is now weaker, the penalty for missing a first serve grows, which pushes the player to serve a safer first ball, which raises the first serve percentage and lowers the value of every serve they hit.

That chain runs in one direction and it explains a great deal of ordinary club and junior tennis. Fear of the double fault produces a soft second serve, the soft second serve produces a cautious first serve, and the cautious first serve produces a player with an excellent-looking percentage who cannot hold serve against anybody decent.

The correct response is unintuitive. A player who never double faults is almost certainly serving their second serve too conservatively, in the same way that a poker player who never gets caught bluffing is not bluffing enough. Some double faults are the price of a second serve worth having.

The percentage moves for reasons that have nothing to do with the server

Even taken as a pure reliability measure, the figure is contaminated by things outside the server's control.

The returner's position changes what the server attempts. A returner standing several metres behind the baseline invites a wider, more aggressive first serve because the angle is available and the cost of missing is unchanged. A returner standing on the baseline pushes the server towards the body and the safer target. Two identical serving performances against two different returners will produce different percentages.

Conditions do a great deal of the work. Altitude thins the air and lengthens the flight of the ball, so serves that would land inside the line at sea level go long, and percentages fall. Heat does something similar by livening the ball. Cold and heavy air do the reverse. Wind is the most direct of all, because the ball toss is the least controllable part of the motion and a crosswind moves it.

The balls matter. A newer, faster ball travels further for the same swing, and a heavier, softer ball drops earlier. Tournaments differ in their ball supply, and the change is enough to move a player's serving numbers by a visible margin without any change in what they are doing. The same factors run through why the same serve reads differently at different tournaments, and they contaminate the percentage exactly as they contaminate the speed gun.

The score changes the strategy within the match. A server facing a break point will typically land more first serves than usual, because the cost of a fault is concentrated at that score. A server at 40-love can afford to swing freely. So the match's overall percentage is a weighted average over score lines the player deliberately treated differently, which is one reason the break point columns need reading with care as well.

None of these is an error in the statistic. They are simply reasons why comparing one player's percentage against another's, or against their own from a different week, is a weaker exercise than it appears.

Surface changes where the optimum sits

The trade-off is the same on every court, but the size of the terms in it is not, and that moves the peak.

On a fast, low-bouncing surface the returner is short of time on both serves. A landed first serve is worth a great deal, because it frequently ends the point or produces a defensive reply. The second serve is comparatively exposed, because the returner can step in and use the pace of a slower ball on a court that rewards flat hitting. Both effects raise the penalty for missing a first serve, and both push the optimum towards higher reliability.

On a slow, high-bouncing surface the picture reverses. The returner has time on the first serve, which reduces what a landed one is worth. The returner also has to deal with a kicking second serve that jumps above shoulder height, which reduces how badly the second serve is punished. The penalty for missing shrinks, and the optimum drifts down the scale towards more aggression.

That is the tactical reason clay-court specialists often show unremarkable first serve percentages and still hold serve comfortably, while the same figure on an indoor hard court would signal a problem. The differences between the main court surfaces do not just change how rallies look. They change the correct answer to a purely statistical question about how often to miss.

Indoor conditions sharpen the effect further by removing wind and sun, which raises achievable percentages for everybody and makes the serve a bigger share of the match. A percentage that would be respectable outdoors in a breeze is mediocre in a still hall.

The optimum is not constant inside a single match

Treating the first serve as one setting for two hours is another way the column misleads.

The right amount of risk depends on the score, and it depends on it in a way that follows directly from the leverage structure of tennis scoring. At 40-love the server can lose the next two points and still be favourite to hold, so a fault is cheap and a free point is valuable. At 30-40 a fault hands the returner the most winnable point available at the worst possible moment, so the same fault is expensive.

Good servers act on that, and the acting on it is visible if you look. First serve percentage tends to rise at the score lines where a fault is costly and fall where it is not. That is not inconsistency. It is a player solving a different arithmetic problem at each score, which is exactly what they should be doing.

The consequence for the statistic is that a match figure is an average over a set of deliberately different decisions. Two players with the same match percentage may have distributed it in opposite ways, one serving conservatively throughout and the other alternating between free swinging and caution. The first is worse and the column cannot tell you so.

The same logic governs risk decisions across sport wherever a failed attempt costs less than the naive reading suggests, which is the insight behind the strokes gained framework in golf: value the outcome against the position it leaves you in, not against a pass-or-fail line.

Men's and women's tennis: same structure, different numbers

The trade-off is identical on both tours because the rules are identical. The magnitudes differ, and the differences change where the optimum sits rather than whether one exists.

Serve speeds are lower in the women's game, and the gap between what a first serve and a second serve achieve is generally narrower. A narrower gap means a smaller penalty for missing, which by the same arithmetic argues for taking more risk on the first ball rather than less. The optimum on the women's tour sits at a comparable place on the reliability scale, but the reason it sits there is a different balance of the same terms.

The other structural difference is match length. Women's singles at every level is best of three sets, while the men's Grand Slam draws are best of five. Longer matches admit more fatigue, and fatigue erodes the first serve before it erodes almost anything else, because the serve is the most demanding single action in the sport. A five-set match will often show a first serve percentage that decays across sets, and that decay is physical rather than tactical.

Neither tour supports a universal target. The percentage that suits a player with a 190 kilometre-per-hour flat serve is not the percentage that suits a player whose weapon is a heavy kicker into the backhand, on either tour.

Why coaches still shout at players to get the first serve in

If the optimum is interior, why is the most common instruction in tennis an unqualified demand for more first serves?

Because most players, most of the time, are not at the optimum, and the error is asymmetric. The player who is serving badly is usually missing a lot of first serves without hitting them well enough to justify the misses. Their first serve win rate is not high enough to pay for the volume of second serves they are producing. In that state, telling them to get the first serve in is correct advice, and it works.

The instruction is also easy to follow, which matters. A player cannot execute an expected-value calculation between points. They can execute "more margin over the net", and that single adjustment fixes the common failure mode.

Where the advice goes wrong is when it is applied to a player who is already at or above their optimum. A big server landing serves in the fifties and winning a very high share of the points behind them is doing the arithmetic correctly. Asking that player to reach the sixties by taking pace off is asking them to convert their best asset into a neutral one. It looks like an improvement on the graphic and it is a downgrade on the scoreboard.

The tell is which direction the points are going. If a player's serve is going in more often and they are winning fewer service points, they have overshot.

Variety is the term the percentage cannot see

There is a third variable in the trade that the two-term description above still leaves out, and it works against the safe server as well.

A returner does not react purely to the ball. They anticipate, and anticipation is worth more than reaction time on a fast serve. If a server has three targets available from the same toss and the same swing, the returner has to cover all of them and cannot commit early to any. If a server has effectively one target, because they have taken enough pace off to be sure of landing it, the returner can lean.

So making the first serve safer degrades it twice over. The ball itself is easier to handle, and the ball is easier to predict. The second effect compounds through the match, because a returner accumulates evidence. By the third set a returner facing a player who has been serving to the same spot at the same speed is guessing correctly far more often than they were in the first game, and none of that shows in a percentage that has been flat all afternoon.

This is why the best servers deliberately hit a proportion of first serves they do not expect to be their most effective. A body serve at 40-love, a wide one on the deuce court to a returner who has been cheating to the middle, a slice that the opponent has not seen for two sets. Some of them are wasted points. All of them are payments towards the value of the serves that come later, and the ledger they belong on has no column anywhere on the match page.

Judging a serve over a season rather than a match

Stretch the sample out and the percentage becomes more stable, but it does not become more meaningful, because the thing it omits is systematic rather than random.

A season-long first serve percentage is a reliable estimate of how much margin a player habitually leaves. That is a real fact about them and it is worth knowing. What it still cannot tell you is whether the margin they leave is the right amount, because the right amount depends on their second serve, their returner population, the surfaces they play most, and the altitude and ball type of their usual tournaments.

The comparison that does work over a season is a player against their own history. If a player's percentage has climbed four points across a year while their first serve points won has fallen five, something has changed in the motion or in the confidence, and it is almost certainly not an improvement. If both have risen together, the serve has genuinely got better. That second pattern is rare, because the two figures pull against each other by construction, and it is the clearest single sign of a player who has done real work on their delivery.

Cross-player comparison over a season is the exercise that stays broken at every sample size. The percentage answers a question about method, and the leaderboard presents it as an answer about quality.

Reading the serving block on a stats page properly

The official match page carries several serving lines. Read them in this order.

Total service points won, first. This is the only figure with a direct claim on the result. If it is high, the serving day was good, whatever the percentage says. If it is low, the serving day was bad for exactly the same reason.

First serve points won, second. This is the quality measure the percentage lacks. A player landing a modest share of first serves and winning a very high proportion of them is serving well and knows it.

First serve percentage, third, and only as context for the two above. Ask what it explains rather than what it scores. A high percentage next to a mediocre first serve win rate is the signature of a player who has taken too much pace off.

Second serve points won, fourth. This is the number that sets everything else, because it fixes the penalty for a fault. A player winning close to half the points behind their second serve has earned the right to swing at the first one.

Double faults, last and lightly. A small number is normal and a zero is suspicious.

• First serves in
• First serve points won
• Total service points won

Aggressive server557967

Cautious server746662

A constructed illustrative pair, built so that the player with the lower first serve percentage wins more service points. Figures are chosen to demonstrate the arithmetic and are not taken from any match.

Show the numbers

Two serving days that the percentage column ranks backwards

Item	First serves in	First serve points won	Total service points won
Aggressive server	55	79	67
Cautious server	74	66	62

What to watch for in the next service game

The statistics arrive after the fact. The decision is visible in real time, and it is one of the easier things to read from a seat or a sofa.

Watch where the first serve lands rather than whether it lands. A first serve that catches the line on the wide side of the box, or that jams the returner at the hip, is doing its job. A first serve that lands two-thirds of the way into the box at moderate pace is a serve the player has already surrendered on, and it will show up in the same column as the good one.

Watch the returner's feet on second serves. If they are stepping several feet inside the baseline to meet the ball, the second serve is not holding them off, and this player needs a higher first serve percentage than a player whose returner is pinned back. That single observation tells you more about where the optimum sits for this individual than any season-long average.

Watch what happens to the percentage after a break of serve. A player who has just been broken frequently tightens, adds margin, and starts landing more first serves while winning fewer points behind them. The graphic at the end of the set will show the tightening as an improvement.

And when the summary appears, resist the number in the largest font. Reliability is a means. Points on serve is the end, and the two come apart far more often than the presentation admits.

More on serving, scoring and the statistics the broadcast gets wrong is collected in the tennis archive, and every explainer across the other sports sits in the blog index.