Yorker execution rate is a human opinion typed into a spreadsheet

Somebody sitting in a booth decides whether the ball was a yorker, and two analysts watching the same over will disagree.

Somebody has to press the button.

Every execution rate you have ever seen quoted at the death rests on a person, usually a young analyst on a laptop, deciding within a couple of seconds whether the delivery counted as an attempted yorker. Not whether it landed as one. Whether it was meant to be one. That is an inference about intention, made from a chair, at speed, several dozen times an innings.

Nobody publishes the agreement rate between coders. In fields that take measurement seriously, inter-rater reliability is the first thing you report, because a variable that two trained observers code differently a fifth of the time cannot support the conclusions people want to draw from it. Cricket analytics almost never reports it, and I suspect the reason is that the numbers would be uncomfortable.

Think about the boundary cases, which are not rare. A ball lands as a low full toss. Attempted yorker that slipped, or a deliberate wide yorker that went wrong, or a slower ball misjudged? The coder decides. A ball lands on a hard length. Failed yorker or an intentional back of a length change up? Decides again. Depending on which way those go, the same bowler in the same over can finish the night with an execution rate of sixty per cent or forty.

Then that figure moves through the system as if it were read off an instrument. It goes into the coaching review. It appears in a recruitment note. Three months later somebody cites it in a selection argument without any idea that a person made a judgement call to create it.

The fix is not exotic. Double code a sample of matches, publish the agreement, and report execution rates with a stated tolerance. Where the tolerance is wide, say so. Teams that do this internally are quietly ahead of everyone quoting cleaner looking figures.

There is a better version available, which is to record intent from the bowler or the captain before the over and code against that. A few setups do it. It is more work and it produces a number that actually means what its name says.

Until then, treat execution rate the way you would treat a scout's opinion, because that is what it is. A structured opinion, stored in a database, wearing the clothes of a measurement.