How does Hawk-Eye work?
Multiple synchronised cameras track the ball, software reconstructs its three-dimensional flight path, and the system predicts where it went — including, critically, where it would have gone had something not been in the way.
The method. Typically six to ten high-speed cameras are positioned around the venue at surveyed locations. Each frame, every camera identifies the ball's position in its own two-dimensional image. Combining views from cameras at known positions allows the ball's 3D coordinates to be calculated by triangulation, many times per second.
Those points are then fitted to a trajectory, using a model of ball flight that accounts for gravity, drag and spin. The output is a smooth path rather than a series of dots.
Why the distinction between measurement and prediction matters. In tennis, Hawk-Eye reports where the ball actually bounced, including the compression of the ball on contact with the surface — which is why the graphic shows an elliptical mark rather than a point. This is close to measurement.
In cricket's LBW decisions, the system must do something fundamentally different: predict where the ball would have travelled after striking the pad. That is an extrapolation, and it carries genuine uncertainty — which is precisely why the umpire's call provision exists. If the predicted path only marginally clips the stumps, the on-field decision stands, because the margin is within the system's acknowledged error. This is widely misunderstood as inconsistency and is actually an honest treatment of measurement uncertainty.
Where it is used: tennis line calling, now fully automated in some tournaments and replacing line judges entirely; cricket, for LBW and for tracking; football goal-line technology and offside; and various other sports.
Its accuracy is quoted in millimetres for tennis, and independent testing has broadly supported the figures.
Its limitations: it needs enough cameras with an unobstructed view, so heavy occlusion degrades it; it depends on accurate calibration; and prediction-based applications inherit the uncertainty of the model.
The cultural effect is worth noting — automated calling removed a whole category of dispute from tennis, and with it some of the sport's theatre.