The wicket goes to the bowler who happened to be there at the end

Dismissals are recorded as single events, which quietly hands one bowler six overs of somebody else's work.

Ninety balls of pressure, and the scorecard remembers one.

That is the bargain cricket made with itself a long time ago and has never revisited. A dismissal is an event with a timestamp and exactly one owner. The bowler who tied a batter down for six overs, made him abandon the sweep, got him playing at the ball he had left alone all morning, ends up with a maiden and a row of dots. The man who replaces him gets the wicket, the better average, and the clip.

I do not think this is a small accounting quirk. It shapes selection. It decides who is asked to bowl the attacking spell and who is asked to hold an end, because holding an end is rewarded in economy rate, and economy rate is worth almost nothing at contract time next to a column of wickets.

There is a real mechanism behind the mismatch. Dismissal probability on any given ball is not a property of that ball alone. It depends on the state of the batter, and state accumulates over deliveries. Someone who has faced twenty dots is playing a different game from someone who has just carved two boundaries. Everything we understand about how batters take on risk says the recent past matters enormously. Our credit system says only the present delivery does.

Attempts to patch this exist. Pressure indexes, dot-ball chains, expected-wicket models that weight the preceding overs. They beat nothing at all, and they share an awkward property: you cannot rerun the innings without the holding spell to see what would have happened instead. That is the honest limit. We are dividing one observed outcome among several causes with no ground truth to check the division against.

So what would I do. Publish the wicket next to the state it arrived in. Balls faced by the dismissed batter. Runs conceded in the previous three overs. Whether the bowler who took it had already bowled at that batter in the innings. None of that needs a model. It needs columns.

The averages we print are not wrong so much as incomplete in a direction that always favours the same type of bowler. Strike bowlers are paid by a number the holders help manufacture. That is a distribution problem dressed up as a statistic, and it is not going to correct itself.