A ball that swings without spinning

A cricket ball delivered with the seam angled a few degrees to the line of flight curves sideways through the air, sometimes by a foot or more over twenty metres. The commentary calls it swing, and the explanation offered is almost always the Magnus effect — the ball is spinning, the spin drags air round with it, and the ball goes sideways.

The ball is spinning, and the explanation is still wrong. A conventional swing bowler spins the ball about an axis pointing along its own flight path, deliberately, to keep the seam steady. Spin about that axis produces no Magnus force whatever, because the Magnus force is a cross product and that cross product is zero.

The mechanism is somewhere else entirely, and it is the same one the previous rung was about: where the boundary layer lets go.

The mechanism, in three sentences #

The seam is a raised ridge of stitching about a millimetre proud of the surface. Angled to the flow, it stands in the boundary layer on one side of the ball and not the other.

On the seam side, the ridge trips the layer to turbulence. A turbulent layer carries more momentum near the wall and can climb a steeper pressure rise, so it holds on longer — to something like a hundred and twenty degrees from the front stagnation point.

On the other side, the layer stays laminar and gives up early, at about eighty degrees.

Two different separation angles on the same ball means two different pressure distributions, and two different pressure distributions do not cancel. The residual is a side force.

That is the whole thing, and none of it involves rotation. Rotation appears only as housekeeping: the bowler spins the ball about the flight axis so that the seam stays pointing the same way for the whole delivery rather than tumbling.

The model, and what it assumes #

The site’s solver cannot predict either separation angle — that would require a turbulent boundary layer, and nothing here computes one. So both angles are placed, from measurement, and the calculation is about what follows from them.

The pressure on the attached part of each side is taken as the exact potential-flow distribution for a sphere,

$$C_p = 1 - \tfrac{9}{4}\sin^2\theta$$

which is a genuine closed-form solution of Laplace’s equation and is the same object the site uses everywhere it computes an ideal field. Past separation, both sides take a single common base pressure, on the argument that the wake is one connected region at approximately one pressure.

Two approximations, then, and they are of different weights. The attached distribution is good — a real laminar boundary layer barely disturbs the outer flow, and measured pressures on the front of a sphere follow the potential curve closely, which is the standard justification for the whole boundary-layer picture. The single base pressure is cruder, and it is the one that decides the answer.

There is also a modelling choice hidden in taking $C_p = 1 - \tfrac{9}{4}\sin^2\theta$ at all: it is the three-dimensional sphere solution rather than the two-dimensional cylinder’s $1 - 4\sin^2\theta$. A ball is a sphere, so the sphere’s is right, and the difference is not cosmetic — the cylinder’s suction peak is $-3$ against the sphere’s $-1.25$, so the two would give side forces differing by a factor of more than two. Using the cylinder because it is the shape the site has drawn most often would be exactly the kind of quiet substitution that produces a plausible wrong number.

Two routes to the integral #

The side force is a surface integral of pressure resolved across the flight direction:

$$C_y = -\frac{1}{\pi}\int_0^{2\pi}\!\!\int_0^{\pi} C_p\,\sin^2\theta\,\sin\phi\; \mathrm{d}\theta\,\mathrm{d}\phi$$

It is computed twice. Once by quadrature over both angles, with the $\theta$ range split at each side’s separation angle — the pressure has a step there, and a midpoint rule that straddles a step converges as $1/n$ rather than $1/n^2$; at a thousand intervals that left the two routes $4\times10^{-4}$ apart, which is a tenth of a per cent of a side force and more than enough to hide a real error in one of them. And once in closed form, since the integrals of $\sin^2$ and $\sin^4$ are elementary.

The two agree to $1.8\times10^{-6}$, and the drag from the same integral agrees to $8\times10^{-7}$.

Three assertions run before any figure is drawn, and they are three different failures.

A symmetric ball gets no side force. With both separation angles equal, the closed form returns zero to fourteen decimal places and the quadrature to nine. A model that produced a force on a ball with nothing asymmetric about it would be wrong in a way no picture would reveal, and it is the cheapest possible check on the whole construction.

The two routes agree. That is the arithmetic check, and the interval-splitting above is what makes it a check rather than a coincidence.

The crossover base pressure lies outside the measured range. This one is the physics, and it is the subject of the next section.

The number the answer depends on #

The base pressure is borrowed, and the side force is linear in it. So the honest question is not “what is the side force” but “how much does the answer depend on the number that was borrowed”.

The crossing is at $C_p = -1.1114$, computed by bisection. Measured base pressures on a sphere in this regime run from about $-1.0$ to $-0.2$. So every value measurement supports gives a force towards the later-separating side, which is the observed direction, and the answer is robust to the borrowing rather than being a consequence of it.

Now do the same calculation with the base pressure the site’s own inviscid solver would supply. The natural thing is to take each side’s wake pressure as the potential-flow value at its own separation point: $-0.6875$ on the turbulent side, $-1.1821$ on the laminar one. Doing that gives a side force of $-0.1195$ — towards the earlier-separating side. The sign has flipped.

That is worth being clear about, because it is a rare and instructive kind of failure. The inviscid theory is not slightly wrong here, and it is not wrong about the mechanism. It gets the direction wrong, and it does so by supplying an entirely reasonable-looking value for the one quantity it is worst at — the pressure in a separated wake, where its assumptions have already failed completely.

A figure drawn from the site’s own solver, using nothing borrowed, obeying every one of the site’s own rules about solving rather than sketching, would show a cricket ball swinging the wrong way. The thing that prevents it is a manometer.

The reason the potential value is so bad is not mysterious. Potential flow’s pressure at eighty degrees is the pressure of a flow that goes on to recover on the back of the ball — that is what makes d’Alembert’s paradox a paradox, and it is why the ideal distribution is so deeply negative just past the shoulder. A real separated wake does no such thing: the flow has left, the pressure stops falling at separation and then sits roughly constant, and it sits at a value that has more to do with the wake’s width than with the angle at which the layer let go. Using the ideal value therefore assigns the deepest suction to the side that separates earliest, which is exactly backwards from what a wake does, and it is enough to reverse the integral.

Recorded here rather than buried, because it is the sharpest example this site has of a rule it otherwise applies without argument. Solving rather than sketching is not always the more honest option. Where a solved field’s assumptions have already failed, its number is not a better class of evidence than a measurement — it is a worse one wearing better clothes, and the only defence is to know which quantity is being asked of it.

Reverse swing, which is the same model twice #

Old balls at high speed swing the other way, and the model reproduces it without any new physics.

The reason for the reversal is that the seam has two effects and only one of them is available at a time. On a new ball at moderate speed, the non-seam side is laminar and the seam’s job is to trip it: the seam side wins the separation contest. At higher speeds — or on a ball whose non-seam side has been roughened by twenty overs of use — both sides are already turbulent, so tripping is no longer on offer. What the seam does then is thicken its own boundary layer, and a thicker layer separates earlier. The seam side now loses, and the ball goes the other way.

The model represents this by exchanging which side separates later. Nothing else changes: the same integral, the same base pressure, the same code. The reversal is a consequence of two placed numbers swapping places, which is a satisfying amount of physics to get out of a model this crude.

The window, and why conditions matter #

Every part of the folklore around swing turns out to be a statement about a Reynolds number window, and the model makes each of them checkable in principle even where it cannot supply the numbers.

There is a speed band. Below it, the seam cannot trip a layer that will separate early anyway; above it, both sides are turbulent and conventional swing disappears. Between those two, the seam side is turbulent and the other side is not, and swing is available. That band is narrow, it is why a medium-fast bowler swings the ball more than a genuinely fast one, and it is the same window the previous rung’s crisis defines, read on one hemisphere at a time.

The ball’s condition moves the window. One side polished and one side allowed to roughen is a deliberate widening of the asymmetry, and it works by moving the two sides’ transition Reynolds numbers apart. It is the trip-wire experiment maintained by hand over four hours.

The seam has to stay pointing the same way. A tumbling ball averages the asymmetry away, which is why the spin about the flight axis is applied, and why a ball that wobbles does not swing.

What the model cannot settle is the effect of humidity, which is the most-repeated claim in the sport and the least well established. Damp air is very slightly less dense than dry air at the same temperature and pressure, so its direct effect on the Reynolds number is negligible and in the unhelpful direction; the plausible mechanisms are all about the ball rather than the air — a damp seam standing prouder, or a damp surface roughening differently. This site takes no position, and stating the model and the regime is exactly the discipline that makes it possible to say so cleanly rather than adding a hedge to a number.

What it is not, drawn #

The mechanism this essay refutes is real, is on this site, and is worth putting beside the one that is actually operating.

Two things separate them cleanly.

The Magnus force needs spin perpendicular to the flight direction. Lift without a wing is the exact statement: the force is $\rho \mathbf{U} \times \boldsymbol{\Gamma}$, and a ball spinning about its own flight axis has $\boldsymbol{\Gamma}$ parallel to $\mathbf{U}$, giving zero. A swing bowler’s ball spins about the flight axis on purpose.

Swing needs a seam, and Magnus does not. A perfectly smooth ball with backspin produces Magnus lift and no swing. A seamed ball with no spin at all, held steady, produces swing and no Magnus force. Both are demonstrable, and they have been demonstrated in wind tunnels since the 1980s.

There is a third case, which is where the confusion is honestly earned: a spinning ball at subcritical Reynolds number can be pushed against the Magnus direction, because the spin trips the layer on the retreating side and moves the separation points asymmetrically. That is the reverse Magnus effect, it is separation-driven, and it is this essay’s mechanism intruding on the other essay’s.

Late swing, which needs no late anything #

The most repeated claim about a swinging ball is that the good ones swing late, and the usual inference is that something about the force changes near the end of the flight. Take the side force computed above, hold it constant for the whole delivery, and see what the trajectory does.

A constant sideways force gives a constant sideways acceleration, and a constant acceleration gives a deviation proportional to the square of the time elapsed. So of the total lateral movement, a quarter has happened by the halfway point and three quarters is still to come; forty-four per cent of it occurs in the final quarter of the flight. A ball under a rigorously constant force therefore appears to travel almost straight and then dart, and the appearance is arithmetic rather than aerodynamics.

It is worse than that from the batter’s point of view, because the judgement has to be made early. The decision about where to put the bat is taken from the first part of the trajectory, which is exactly the part in which almost none of the deviation has yet accumulated. A constant force is not merely compatible with the folklore; it predicts it.

That said, there is a mechanism that would make swing genuinely late, and this essay’s own window supplies it. The ball decelerates through its flight, so its Reynolds number falls steadily from release. A delivery bowled a little above the top of the window starts with both sides turbulent and no asymmetry worth the name, and drops into the band partway down the pitch — at which point the seam side and the other side part company and the force switches on.

Both accounts are available and they are not exclusive. The useful discipline is that the first costs nothing, so an observation of late swing is not by itself evidence of the second.

Where the model stops #

Both separation angles are placed. They come from measurement. Nothing here predicts them, and they are the input the entire result is proportional to.

The wake has one pressure. It does not. A real wake behind an asymmetric body is itself asymmetric, and the single-pressure assumption is the crudest thing in the calculation. What defends it is the crossover computation: the answer’s sign survives any plausible value, which is a weaker claim than getting the magnitude right and is the claim the essay actually makes.

The magnitude is not to be trusted. The model gives a side-force coefficient of 0.295 and a drag coefficient of 0.358. Measured swing forces are smaller and measured sphere drag at this Reynolds number is nearer 0.5. The model is right about the sign, the mechanism and the reversal, and roughly right about the order — which is what a two-parameter model of a separated flow is entitled to be.

Nothing here is unsteady. A real ball’s wake oscillates, the side force fluctuates, and the trajectory is not the smooth arc a steady coefficient implies.

Who found it, and when #

Swing was bowled long before it was explained. The systematic wind-tunnel work belongs to Rabindra Mehta from the late 1970s onward, whose measurements established the separation-angle mechanism, the Reynolds-number window in which conventional swing operates, and — later, and more contentiously — the reverse-swing mechanism at higher speeds. Before that the standard published explanation in coaching literature was the Magnus effect, and it persists in commentary today.

The physics has a much older parallel that nobody connected at the time. Prandtl’s 1914 trip wire on a sphere is a seam applied symmetrically; a cricket seam is Prandtl’s wire applied to half the ball. The two experiments are sixty years apart, from entirely separate communities, and are the same experiment.

Where the ladder goes #

This anchor closes with a pattern worth naming, because it recurs whenever a field runs out of solvable flow.

Both rungs computed something exactly on top of something borrowed. The drag crisis is a measurement and the falling force is arithmetic; the separation angles are measurements and the pressure integral is arithmetic. In both cases what the site contributes is a consequence, and the honest reporting is to say which half is which.

The field turns next to a case where nothing needs to be borrowed at all. A sailing boat’s performance, to the order that matters, is two drag angles and a trigonometric identity — no areas, no coefficients, no measurements — and the identity says that a boat can sail nearly three times faster than the wind driving it.