Phase splits stopped being comparable the moment the fielding rules changed

Middle overs numbers from different rule eras sit in the same database column and quietly refuse to mean the same thing.

The middle overs are not a fixed object.

They are whatever the fielding restrictions of the day made them, and those restrictions have been revised repeatedly: how many fielders outside the circle, for how long, with what mandatory catchers, in which format. Each revision changed the expected run rate for the same batting method against the same bowling. Yet the phase splits sit in one column, labelled overs eleven to forty, and get compared straight across.

I would call that the single most common unlabelled error in cricket data, and it is invisible because the column heading looks stable. The name of the phase did not change. The physics of the phase did.

Take a straightforward case. A restriction allowing one more fielder on the boundary during a stretch of overs reduces the value of an aerial shot to that region, which changes which shots batters attempt, which changes the dismissal profile, which changes both the run rate and the wicket rate for reasons that have nothing to do with the players involved. Compare a bowler's middle overs economy either side of that change and you are measuring a rule, not a person.

The same applies to bat sizes, boundary rope positions, ball manufacturers and the number of balls used per innings. All of them shift the baseline. Any of them can produce a trend line that looks like a change in playing standards.

The practical answer is boring and effective: era adjust everything, the way baseball has done for decades by expressing figures relative to the league mean of the same season. Runs per over means little on its own. Runs per over relative to the competition average that year means a great deal, and it survives rule changes because the baseline moves with them.

Cricket resists this because the raw figures carry emotional weight. An average of fifty is a landmark, and nobody wants a relative index instead. Fair enough for the pub. Not fair enough for recruitment, where somebody is paying real money on the basis of a comparison across two rule regimes.

There is one thing I would insist on immediately, and it costs nothing. Stamp the rule set onto the record. Every phase split should carry the fielding restriction configuration it was accumulated under.

Then at least the incomparability would be visible, rather than hiding inside a column heading that has not changed since the eighties.