Tennis serve speed explained: why fastest is not hardest

The number that flashes up after an ace is not the speed of the serve. It is the speed the ball was travelling in the first fraction of a second after it left the strings, and by the time it arrives at the returner's racket a large share of it is gone.

That gap is the whole of tennis serve speed explained, and it is why the fastest serve on a stat sheet is frequently not the hardest one to play. A ball that leaves at 220 km/h and a ball that leaves at 190 km/h do not arrive 30 km/h apart. They arrive much closer together than that, and the slower one may be the one that costs the returner the point, because pace is only one of the four things a serve can do to a person standing 24 metres away.

The other three are time, position and height. Those are the ones the radar cannot see.

What the number on the screen is actually measuring

Two technologies produce the figure, and they do not do the same thing.

A Doppler radar unit transmits a radio signal and listens for the reflection. A ball moving away from the unit returns the signal at a shifted frequency, and the size of the shift gives the speed. It is a mature, cheap and reliable method, and it has one structural weakness that matters enormously in tennis: it measures the component of velocity along its own line of sight, not the true speed of the ball.

That is the cosine error, and it is always in the same direction. A radar reads low, never high. A serve struck straight down the middle at a unit sitting on the centre line is measured almost exactly. A serve struck wide is travelling at an angle to the beam, so only part of its velocity is registered, and the reading understates it. Published analysis of radar measurement in tennis puts the understatement of a wide serve at a few per cent at the moment of delivery, growing considerably by the time the ball reaches the court, because the angle between the ball's path and the beam keeps opening as the ball travels.

The other method is camera-based tracking. An array of synchronised high-speed cameras reconstructs the ball's position many times a second and differentiates that trajectory to get speed. It has no cosine problem, because it knows where the ball is in three dimensions rather than how fast it is receding from one point. It has different weaknesses instead: it depends on calibration, on how many cameras see the ball at any instant, and on how the trajectory is smoothed before the speed is calculated. The camera system that reconstructs a ball's path for line calling is doing considerably more work than the scoreboard number suggests.

Both approaches report the same thing conceptually, which is speed at or immediately after impact. Neither reports the speed the returner faces, and no tournament displays that number, although it would arguably be the more interesting one.

Why the same serve reads differently at different tournaments

Here is the uncomfortable part. Serve speed is not standardised across professional tennis in the way that, say, a stopwatch is standardised across athletics.

The variables are mundane and they compound. Where the measuring equipment sits, and at what angle. Which point on the trajectory the system samples, since the ball is already decelerating in the first metre. How the manufacturer's software processes the raw signal. Whether the event uses radar, cameras, or both with one feeding the display. Whether the equipment was replaced this year with a newer model that samples differently.

None of that is scandalous. It is what happens when a measurement is taken for entertainment rather than for adjudication, and nobody has ever needed the serve speed to be right to decide a point.

The consequence is that comparing a player's fastest serve at one event with their fastest at another is not a like-for-like comparison, and comparing across eras is worse, because the equipment changed underneath the numbers. Any argument that begins with a player's peak serve speed at a particular tournament being higher than another player's peak somewhere else is an argument about instrumentation as much as about tennis.

The sport's own record books admit this. The fastest serve most commonly cited is Sam Groth's 263.4 km/h, or 163.7 mph, struck at a Challenger event in Busan in 2012. The ATP does not treat it as its recognised record, because of questions about how the equipment at that event was calibrated, and instead recognises John Isner's 253.0 km/h, or 157.2 mph. Two figures, both published, both defensible, roughly ten km/h apart, and the disagreement is entirely about measurement rather than about what the two players did.

Serve speed is therefore best read as a rough index rather than a precise quantity. Within a single match on a single court it is a fair comparison. Across tournaments, it is a vibe.

• 6.4Service box depth, metres
• 4.12Service box width, metres
• 0.91Net height at the centre, metres
• 1.07Net height at the posts, metres

Fixed dimensions from the ITF Rules of Tennis. These do not change with surface, event or era, which is what makes them the most reliable numbers in any discussion of serving.

Where the speed goes between the racket and the returner

A tennis ball is a bad aerodynamic object, and deliberately so. The felt nap that makes it grip a string bed also makes it drag heavily through the air, which is what keeps rallies playable. That drag is roughly proportional to the square of the speed, which produces the first counterintuitive result: the faster the serve, the more speed it sheds per metre travelled.

Then the ball hits the ground, and the bounce takes another slice. The contact lasts a few thousandths of a second, during which the ball deforms, slides or grips depending on the surface and the spin, and loses forward speed to friction and to the energy absorbed in squashing and un-squashing itself. On most surfaces something in the region of a fifth of the horizontal speed goes here.

Then there is a short additional flight to the returner, and more drag.

• Speed
• km/h

Constructed illustration with invented round numbers, not a measurement of any real serve. It assumes drag removes a quarter of the speed on the way to the bounce, the bounce removes a fifth of what is left, and a little more is lost in the short flight afterwards. The real fractions vary with surface, ball, spin, altitude and temperature, but the shape of the loss is always this.

Show the numbers

Worked example: what happens to a 200 km/h serve

Item	Speed	km/h
Leaving the racket	200
Crossing the net	168
Just before the bounce	150
Just after the bounce	120
Reaching the returner's racket	112

56%25%15%

• Still there when the returner strikes it112
• Lost to air resistance before the bounce50
• Lost in the bounce itself30
• Lost to air resistance after the bounce8

The same invented 200 km/h serve, shown as a composition. The parts add to the number on the screen. Constructed for illustration and not a measurement of any real serve.

Show the numbers

Worked example: where a serve's speed ends up

Item	Value
Still there when the returner strikes it	112
Lost to air resistance before the bounce	50
Lost in the bounce itself	30
Lost to air resistance after the bounce	8

That is the first reason the fastest serve is not automatically the hardest one to return. A serve struck 15 km/h faster does not arrive 15 km/h faster, because the extra speed is precisely what the drag term punishes. The gap narrows. It does not close, and the faster serve still arrives first, but the advantage bought by raw pace is smaller at the receiving end than it looks at the transmitting end.

The returner is not short of speed, they are short of time

Reframe the whole question in the unit that matters to the person receiving.

A returner's problem is not that the ball is fast. It is that the interval between the server's contact and their own contact is short, and everything they have to do has to fit inside it: recognise the toss and the swing path, read the ball's line and spin, decide, move, set the feet, and swing. Speed only matters through its effect on that interval.

The court is 23.77 metres from baseline to baseline. A server contacts the ball slightly inside their own baseline and a returner standing behind theirs may be a metre or two further back, so the ball's journey is roughly 24 to 26 metres depending on where both players choose to stand.

Now do the arithmetic on an invented serve. Suppose that over the whole journey, including the bounce, the ball averages 150 km/h, which is 41.7 metres per second. Twenty-five metres at that average takes about 0.60 seconds. Raise the serve speed enough to lift the average to 160 km/h, or 44.4 metres per second, and the same journey takes about 0.56 seconds.

Four hundredths of a second. That is what a substantial increase in serve speed buys, and it is a real gain, because at that timescale four hundredths is a meaningful fraction of a returner's preparation. But it is a small gain compared with what happens if the returner is standing in the wrong place, or has to hit the ball above their shoulder, or has guessed the wrong side.

This is why professional returners will stand two or three metres behind the baseline against a big server. Standing deeper adds distance, which adds time, at the cost of having to cover a wider angle and having to hit the return from further out of the court. It is a straightforward trade of court position for milliseconds, and where a player chooses to stand tells you what they think their problem is.

Standing further forward is the opposite bet: less time, but the ball is caught earlier in its bounce, before it has risen to an awkward height, and the return is struck from inside the court where the angles are better. Both are defensible. Neither has anything to do with how fast the ball is going in absolute terms.

Spin is the variable that beats pace

If a returner's problem is time and position, then a serve's job is to attack both. Pace attacks time. Spin attacks position, and it does so more reliably.

Spin works through the pressure difference a rotating ball creates in the air around it, which pushes the ball perpendicular to its direction of travel. Topspin pushes it downwards, which means a ball hit with topspin can be struck harder and aimed higher over the net and still come down inside the box. Sidespin pushes it sideways, which curves the flight. Most serves carry some combination of both, and the axis of rotation decides the mixture.

Three families of serve come out of that.

The flat serve carries the least spin and therefore has the least help getting down into the box. It is the fastest through the air, it stays lowest after the bounce, and it has the smallest margin for error of any serve in tennis. It is a weapon that works by removing time and nothing else.

The slice serve is struck with the racket travelling across the ball, generating sidespin. It curves in flight, skids low off the bounce and moves away from the returner. On the deuce court a right-hander's slice drags the returner off the side of the court; on the ad court a left-hander's does the same to a right-handed opponent, which is the single most valuable geometric asset in the professional game and explains a great deal about why left-handers over-perform in tennis.

The kick serve is struck with the racket brushing up and across, producing a spin axis tilted enough to give both topspin and sidespin. The topspin makes the ball dip steeply into the box despite being hit with a large net clearance, then makes it leap upwards and sideways off the surface.

The kick serve is the direct refutation of the idea that speed is what matters. It leaves the racket slower than a flat serve. It arrives slower still. And it is, for most returners, the harder ball to attack.

The reason is contact height. A ball that bounces to shoulder height or above forces a returner into a stroke they cannot drive. Below roughly waist height a player can swing up through the ball and impart their own topspin; at shoulder height they are reaching, the racket path is short, the body cannot rotate into the shot properly, and the ball comes back slower and higher than they wanted. A returner faced with a good kick serve has three bad options: take it early on the rise with almost no time, take it late from well behind the baseline having surrendered court position for the rest of the point, or block it back and hope.

Published measurement of serve spin puts typical rates in the low thousands of revolutions per minute, with the split between topspin and sidespin varying enormously with the axis. A serve can carry a great deal of total spin and very little of the topspin component that produces the bounce, which is why two serves that look similar on a broadcast can behave completely differently on the second bounce.

There is a strategic consequence that goes beyond the individual point. A kick serve to the backhand, especially on a slow surface, does not just win points outright. It starts almost every point with the returner deep, wide and hitting from above shoulder height, which means the server takes the initiative in the rally without having to hit a spectacular serve at all. Consistent positional advantage over four hours beats a handful of aces, and the way that advantage shows up in the statistics a return of serve actually generates is more informative than any speed column.

Placement, and the geometry the box imposes

The service box is 6.40 metres deep and 4.115 metres wide. That is a small target at the end of a long flight, and every metre per second of extra pace shrinks the margin around it.

Within that target there are three meaningfully different places to aim, and they do different jobs.

Down the T, close to the centre service line, gives the ball the shortest path, the lowest part of the net to clear, and the least sideways deviation to control. It is the highest-percentage place to hit a fast serve. Its weakness is that it leaves the returner close to the middle of the court, so even a weak return arrives from a decent position and the server has not opened any angle.

Out wide, close to the singles sideline, has to travel further and clear a higher part of the net, because the net rises from 0.914 metres at the centre to 1.07 metres at the posts. It is therefore the lower-percentage serve. In exchange it drags the returner off the side of the court, which opens the whole of the opposite side for the next ball. The wide serve is not trying to win the point. It is trying to win the point after next.

Into the body is the option that broadcast coverage almost never discusses and that professionals use constantly. A ball aimed at the returner's hip gives them nowhere to put their arms. They cannot extend on either wing, they have to make a decision about which side to take it on while moving, and the resulting return is usually short. A body serve does not need to be fast. It needs to be accurate to within about half a metre, which is a much easier task than hitting a line.

The rule that governs all three is the same one: aiming closer to a line increases the value of the serve and increases the probability of a fault, and the exchange rate between those two is not linear. The last twenty centimetres of accuracy costs far more in faults than the first twenty centimetres bought in advantage, which is why professional targets are areas rather than points, and why the best servers are not the ones who paint lines but the ones who consistently miss by a small amount on the safe side.

Height gives an angular advantage, not just a power advantage

Tall servers dominate the speed lists, and the usual explanation is leverage. That is part of it. The larger part is geometry, and it can be worked out from the court dimensions alone.

Consider the limiting case of a completely flat serve, travelling in a straight line with no spin and no time to fall under gravity. It leaves the racket at height h somewhere near the baseline, 11.885 metres from the net. It has to pass above the net, 0.914 metres high at the centre. It has to land within the service box, so within 6.40 metres of the net on the far side, which is 18.285 metres from the server.

A straight line from contact to the far edge of the service box passes over the net at a height of h multiplied by 6.40 divided by 18.285, which is about 0.35 times h. For that to clear the net at all, h must exceed 0.914 divided by 0.35, which is about 2.61 metres.

That is a constructed calculation from published dimensions, and it deserves the caveats. It ignores the ball's radius, it ignores the fact that the ball does fall under gravity during the flight, which relaxes the constraint slightly, and it assumes the server contacts the ball exactly on the baseline when in practice the contact is usually a little inside it, which makes the requirement harder rather than easier.

The conclusion survives all of that. A perfectly flat serve is not available to most human beings. Below a contact height somewhere in that region, the geometry simply does not permit a straight line that clears the net and lands in the box, and every serve must therefore be curved downwards by topspin.

This is what height buys. It is not that a taller player hits harder, although they generally do. It is that a taller player has a legal window where a shorter player has none, so they can serve closer to flat with a margin of error rather than relying on spin to bring the ball down. They also see a wider range of angles into the box, because a higher contact point can aim more steeply and still land in.

The corollary is the part people forget. A shorter server is not disadvantaged at serving. They are disadvantaged at serving flat. The spin-heavy serve that height makes optional is available to everyone, and against a returner who cannot handle a high bounce it is the better serve anyway.

The first and second serve are one decision, not two

A server gets two attempts, the second with the entire point at stake. That turns serving into an optimisation problem rather than a skill test: how much risk should go on the first serve, and how much on the second?

The arithmetic is worth doing properly, because the intuitive answer is wrong.

Take an invented player. Their first serve lands in 60 per cent of the time, and when it does they win 75 per cent of those points. Their second serve lands in 90 per cent of the time, and when it does they win 55 per cent of those points. All four numbers are made up for the example, but they are the right shape for a professional.

Playing the conventional way, they win 0.60 times 0.75, which is 0.45, plus the 40 per cent of the time the first serve misses multiplied by 0.90 times 0.55, which is 0.198. Total: 0.648, or 64.8 points per hundred service points.

Now suppose they hit two first serves. They win 0.45 on the first attempt, plus 0.40 multiplied by 0.45 on the second, which is 0.18. Total: 0.63, or 63.0 points per hundred. Worse, but by under two points per hundred.

Now suppose they play two second serves, the maximally cautious approach. They win 0.90 times 0.55, which is 0.495, plus 0.10 multiplied by 0.495, which is 0.0495. Total: 0.5445, or 54.5 points per hundred.

First serve, then second serve64.8per 100

Two first serves63per 100

Two second serves54.5per 100

Constructed arithmetic on invented inputs. The imaginary player lands 60 per cent of first serves and wins 75 per cent of those points; lands 90 per cent of second serves and wins 55 per cent of those. No real player's numbers are used. The point is the size of the gaps, not the values.

Show the numbers

Worked example: three serving strategies compared

Item	Value
First serve, then second serve	64.8per 100
Two first serves	63per 100
Two second serves	54.5per 100

Three things fall out of that, and all of them are more useful than a speed reading.

The conventional pattern wins, but the margin over hitting two first serves is small enough that for some players, on some days, on some surfaces, it flips. That is why the idea keeps resurfacing and why nobody has ever quite dismissed it.

The cautious approach is a disaster. Ten points per hundred is an enormous quantity in tennis, more than enough to turn a comfortable server into a broken one. A second serve that avoids double faults by giving away the initiative is losing far more points than it saves.

And the second serve, not the first, is where matches are decided. A player's second-serve points won is the single most predictive service statistic there is, because it captures the whole trade between risk and reward under maximum pressure. The relationship between how often a first serve lands and how much aggression it carries is the subject of what a first serve percentage does and does not tell you, and it is a more revealing column than anything the radar produces.

1. 1The toss and the swingThe returner is already reading. Toss position, shoulder alignment and the path of the racket leak information about direction and spin before contact, which is why a server with a consistent toss for every serve is worth several points a set.
2. 2Impact, and the number on the screenThe ball leaves the strings. Radar or camera tracking samples it in the first metre or two and reports that speed. This is the only moment at which the displayed figure is true.
3. 3Flight to the netAir resistance begins removing speed immediately, and removes it faster the faster the ball is going. Spin starts bending the trajectory, downwards for topspin and sideways for slice.
4. 4Over the netThe ball clears 0.914 metres at the centre and up to 1.07 metres nearer the posts. Serving wide costs both extra distance and extra height, which is why it is the lower-percentage target.
5. 5The bounceThe ball deforms and either grips or slides depending on surface and spin. Forward speed drops sharply, and the spin decides how high and in which direction the ball leaves the court.
6. 6The riseTopspin sends the ball up steeply, slice keeps it low and skidding. This is the stage that decides the returner's contact height, and contact height decides what kind of shot is even possible.
7. 7The returner's decisionStand deep for time and hit from out of the court, or stand up for position and hit with almost none. The choice was made before the serve was struck and cannot be changed mid-flight.
8. 8ContactThe return is struck at somewhere near half the speed the scoreboard advertised, at a height and a court position that the spin and placement chose. This is the number that decided the point, and nobody measures it.

Every stage removes something from the ball or adds something to the returner's problem. The speed reading is taken at the second stage and describes none of the rest.

Surface changes the value of every serve, and it changes it differently

The bounce is where a surface does most of its work, and the serve is the shot most exposed to it because it is the only one that must bounce before the opponent can play it.

A low-friction, fast surface takes less forward speed out of the ball and returns it at a lower angle. Serves stay fast and stay low. The flat serve is worth most here, because the thing it does best, arriving early and staying beneath the returner's comfortable strike zone, is amplified. The kick serve is worth least, because the surface will not throw the ball up to the height that makes it awkward.

A high-friction, slow surface does the opposite. It removes more forward speed and converts more of the ball's rotation into vertical rebound, so the ball sits up. A flat serve loses much of its advantage because the returner gets the time back. A kick serve becomes a genuine weapon, because the bounce is exaggerated into something a returner has to deal with above their shoulder every single point.

The governing body classifies court pace on a measured scale rather than by material, which matters because two hard courts can be further apart than a hard court and a grass court. The full account of what separates one tennis surface from another works through the two variables that produce the whole spectrum.

Air does its own share. At altitude the air is thinner, so drag is lower and the Magnus force that bends a spinning ball is weaker. Serves arrive faster and spin serves curve and kick less, which pushes the advantage back towards the flat server and makes control noticeably harder for everybody. Heat thins the air too, on a smaller scale, and a new ball is faster than a worn one because the nap has not yet been raised. None of this shows in a speed reading, and all of it changes what the serve does.

Why serve speed records are a poor guide to serving quality

Everything above accumulates into one conclusion, which is that the speed columns are close to the least informative public data about serving.

They are inconsistently measured across events. They describe a single instant, at the start of the ball's journey, before the two processes that remove most of the speed have happened. They ignore spin entirely, and spin is the variable that decides the returner's contact height. They ignore placement, and placement is what decides whether the returner is in the court at all. And they are peak values, so they describe a player's single best effort of a match rather than what they did on the other hundred serves.

A serve's actual value is the probability it wins the point, and that is measurable from published data without any radar at all. Service points won. First-serve points won and second-serve points won, separately. The proportion of serves the opponent failed to return. Games and break points saved. Every one of those describes the outcome the serve was for.

Compare the two lists at any tournament and they rarely match. The fastest server in a draw is very often a player whose serve wins fewer points than someone thirty km/h slower, because the slower server hits a spot, varies the spin, and never gives the returner the same ball twice. What makes a serve unplayable is uncertainty plus an awkward contact height, and neither has a unit.

Serve speed is genuinely fun to watch, and there is no reason to stop enjoying it. It is simply not the measurement anyone thinks it is, and the wider archive of analysis and explainers on tennis is mostly an argument that the interesting numbers in this sport are rarely the ones on the screen.

What to watch instead of the speed reading

Four things will tell you more about a serve than any radar figure, and all of them are visible on a normal broadcast.

Watch where the returner stands, and whether it changes. A returner who moves two metres back between the first and second serve is telling you what the kick serve is doing to them. A returner who steps in is telling you they have read the server's second-serve pattern and intend to punish it.

Watch the contact height on the return. A return struck at waist height is a return the player can drive. A return struck at shoulder height is a concession, whatever the speed of the serve that produced it. Count how often a server forces the second of those.

Watch the second serve's landing spot, not its speed. A second serve that lands consistently in the same half of the box is being read, and the returner will start attacking it within a set. Variation matters more than pace, and it is visible without any instrumentation.

Count the unreturned serves rather than the aces. An ace is a serve the returner could not touch. A serve that comes back short and floating has done almost the same job and gets no credit for it. The proportion of serves that produce either outcome is the honest measure of a serving performance, and it correlates poorly with the fastest number anyone hit all afternoon.