Ball tracking rounds off the exact thing that beats the batter

The deliveries that deceive are decided in centimetres of dip and drift, which is roughly where tracking systems run out of resolution.

Dip is the part nobody can see and the part that does the work.

A doosra, or any well-flighted delivery, takes its wickets in the last third of its flight. The batter commits to a length based on the first two thirds. What arrives is shorter, or wider, than what was read. The gap between the two is small. Sometimes it is a matter of a few centimetres of extra drop over the final metre before pitching.

Now think about what a tracking system does with that. It samples the ball at some frame rate, fits a smooth path through the detections, and reports the result to a viewer as a clean coloured line. The fit is the point. A fitted trajectory is a model that assumes the ball behaves as ballistics say it should, and a model that assumes smoothness will always shave the sharpest part of the curve. The deviation that fooled the batter is precisely the deviation the smoothing treats as noise.

This is not a scandal. Fitting is how you get any usable path out of noisy detections, and the systems are good. But we have started reporting derived quantities off those fits as if they were measured: degrees of turn, metres of drift, revolutions. Those are outputs of a model with assumptions, quoted to a precision the underlying capture cannot support.

The practical consequence shows up in how spinners get compared. Two bowlers can post nearly identical turn and drift profiles and get wildly different results, and the temptation is to attribute the difference to something vague like presence or nous. I would look at the measurement first. If your instrument cannot separate two deliveries that batters clearly play differently, the instrument is not describing the mechanism. It is describing whatever survived the fitting process.

What would help is uncertainty. Every reported turn figure should come with an error band, and the band should widen when the ball is close to the pitch, when detections are sparse, when the light is poor. Broadcast will hate that.

Meanwhile the batter is doing something no system has yet quantified, which is deciding at around the halfway mark on evidence that has not finished arriving. We measure the ball. We do not measure the moment the read was made, and that interval is where the contest actually happens.