Why does a ball suddenly swerve? Because air drag collapses at a critical speed, and a spin-free ball feels a seam-dependent side force. Inspired by the 2026 FIFA World Cup (USA, Canada, Mexico), where the thinner air of Mexico City changed how far shots fly.
The science (equations, data sources, honest simplifications)
The ball is a point mass with a spin vector. Newton's second law is integrated with a 4th-order Runge-Kutta step (dt = 1/480 s):
m a = m g + F_drag + F_Magnus + F_knuckle
F_drag = -1/2 rho A C_d(Re) |v| v
F_Magnus = 1/2 rho A C_L(S) |v|^2 * unit(w x v)
Re = rho v D / mu S = w r / v
A = pi r^2, r = 0.11 m, m = 0.43 kg (size 5), mu = 1.81e-5 Pa s
The drag crisis. Behind a sphere the boundary layer separates early (laminar, Cd about 0.45-0.5). When the layer turns turbulent it clings to the ball longer, the wake narrows and Cd drops to about 0.1-0.25 (Achenbach 1972; Mehta 1985). For a size-5 ball this happens at Re of roughly 2-3 x 105, i.e. about 10-20 m/s at sea level. Rough surfaces and deep panel seams trip the layer earlier, so the crisis speed falls (Asai et al. 2007; Hong & Asai 2014). This lab uses a smooth logistic step in log10Re centred on Rec = 3.0 x 105 - 1.5 x 105 k, with the "roughness" k in [0, 1] shifting Rec, a floor Cd,min = 0.10 + 0.14 k and a gentle post-critical rise. This is a fit to the shape of published curves, not a measurement of any real ball.
Magnus lift. A spinning ball drags air around one side, so the wake deflects and the ball is pushed sideways (Robins 1742, Magnus 1852). The empirical lift coefficient used here is CL = 1 / (2 + 1/S) (the form popular in soccer trajectory fits, e.g. Goff & Carré 2009/2010). Very close to the drag crisis and at small S, wind-tunnel data show the lift can reverse sign ("reverse Magnus"); a narrow phenomenological window around Rec reproduces that.
The kick. A foot hitting the ball off-centre by b gives it angular impulse m v b. With a thin-shell moment of inertia I = (2/3) m r2 the ideal spin is w = 1.5 v b / r2. Real feet slip, so only 60 % is kept (eta = 0.6). Spin decays with a time constant of about 14 s at sea level (scaling inversely with air density). A banana kick needs 8-10 rev/s, which is only about 2 cm off-centre.
The knuckleball. With almost no spin (below about 1 rev/s, S below about 0.03) the ball's seams stay in roughly the same orientation relative to the flow, the boundary layer separates asymmetrically, and a side force acts that changes as the ball slowly turns (Asai et al. 2007; Goff & Carré 2010). Here: Fk = 1/2 rho A v2 Ak [ f(phi) elat + 0.6 f(phi + 2.1) eup ] w(Re) w(S), with f(phi) = [sin 3phi + 0.5 sin(5 phi + 1.1)] / 1.5, plus an Ornstein-Uhlenbeck random part (correlation time 0.18 s) standing for the unsteady wake. phi advances at the spin rate. The sinusoidal dependence and amplitudes are a phenomenological model that gives the right qualitative behaviour (random-looking swerve that fades once the ball spins faster than about 1.5 rev/s); they are not derived from the Navier-Stokes equations. The seed fixes the random numbers so a kick can be repeated.
Simplifications. No ball bounce (flight ends at the first ground contact), no foot-ball contact dynamics, a fixed spin-decay law, no wind gradient, no wake or Reynolds history effects, goal posts end the flight, and the keeper and wall are static silhouettes.
Checks built into this page. (1) With Cd = CL = 0 the numerical range equals v2 sin 2theta / g (Experiment 3, typically below 10-9 %). (2) Experiment 1 recovers the critical speed from the simulated deceleration and compares it with vc = Rec mu / (rho D).
Sources: Achenbach, J. Fluid Mech. 54 (1972); Mehta, Ann. Rev. Fluid Mech. 17 (1985); Asai, Seo, Kobayashi, Sakashita, Sports Eng. 10 (2007); Goff & Carré, Am. J. Phys. 77 (2009) and Eur. J. Phys. 31 (2010); Hong & Asai, Sci. Rep. 4 (2014). The ball model is a generic 32-panel design: no brand or tournament artwork.