A camera angle can invent knee valgus

Two-dimensional video flattens rotation into apparent collapse, so where you stood decides what you saw.

Set the tripod half a metre to the left and the same landing produces a different diagnosis. That should worry anyone screening from video, and mostly it does not, because the footage looks so convincing.

What the camera records is a projection. A knee travelling inward on the screen may be travelling inward in space, or the hip may have rotated internally while the knee stayed put, or the athlete may have turned four degrees on take-off. All three produce the same picture. The measurement is an angle between two lines in an image, and it is only an angle at the joint if the joint happens to be moving in the plane the lens is looking at. Landings rarely oblige.

The error is not random either, which is the part that does damage. Stand slightly off-centre and every athlete in the session picks up a bias in the same direction, so the whole squad looks a little more collapsed, or a little less, than it is. A team that moved its filming position between preseason and midseason has manufactured a change in knee control across the group and will go looking for the training cause. Somebody will find one. There is always a plausible cause available for a trend that does not exist.

Distance matters too. Close in, with a wide lens, the limbs at the edge of frame distort. A tall player filmed from the same spot as a short one is not measured under the same conditions, and taller athletes tend to look worse.

The practical response is not to abandon video. It is to treat the number as ordinal at best. Same camera, same taped mark on the floor, same lens, then compare an athlete only with himself and believe only changes visible without measuring. If you need a protractor on a screenshot to see the difference, the difference is inside your setup error.

And if the screening result feeds a risk model alongside strength and history, it should carry the widest uncertainty in that model, not equal weight with a force measurement that has a calibration certificate.

Video gets treated as the reliable input because it is the one people can see with their own eyes. Seeing something has never been the same as measuring it.