How a cricket rating system separates a good innings from an easy one

What a raw average leaves out

A batting average divides runs by dismissals, which is a clean summary and a poor description of how difficult those runs were to accumulate. The same total can be assembled against a depleted attack on a flat surface or against a strong one where scoring is genuinely hard. Treating those as identical is the fundamental limitation, and it is why two competitors with similar averages can differ substantially in value.

The average also ignores the state of the match, so runs scored when the result is settled count exactly as much as runs that changed it. None of this makes the average useless, and it makes it a starting point rather than a conclusion.

Pricing the opposition

A rating system adjusts by asking how the same performance compares with what the same opposition typically concedes to others. That comparison requires a reliable estimate of the opposition level, which in turn depends on ratings derived from the same pool of matches. The circularity is resolved by iteration, and the resulting estimates are stable when the schedule connects teams densely enough to support them.

In formats where certain sides meet rarely, the connections thin out and the adjustment rests on limited evidence. The adjustment is therefore strongest exactly where it is least needed and weakest where the raw numbers are most misleading.

Conditions as a measurable variable

Surface behaviour changes what constitutes a good score, and the standard approach estimates that behaviour from the aggregate scoring in the same match. Using the match itself as the reference is elegant because it automatically accounts for the surface, the weather and the ball on that occasion. It also introduces a problem, since a dominant individual performance inflates the reference and therefore discounts itself.

Excluding the performance being assessed from its own baseline corrects this, at the cost of a smaller and noisier reference sample. The tension between an accurate baseline and an independent one recurs in every context-adjusted measure across every sport.

Match situation and the value of a run

Runs are not equally valuable at every point, because the same total contributes differently depending on how much of the innings remains. Systems that weight for situation reward performances that arrive when the position is difficult and discount those that arrive when it is not. That weighting encodes a view about what the sport values, and reasonable people disagree about how strongly situation should count.

A heavy weighting can reward competitors who happen to bat in a position where difficult situations occur more frequently. Reporting the adjusted and unadjusted figures side by side lets the reader see how much of the assessment depends on that choice.

Reading an adjusted rating honestly

An adjusted rating is a model output, and like every model output it is only as good as the assumptions used to build it. Small differences between competitors sit inside the uncertainty the adjustments introduce and should not be presented as a ranking. Large and persistent differences survive most reasonable variations in method, and those are the ones worth taking seriously.

Checking whether a conclusion holds under several plausible adjustment schemes is more informative than refining a single scheme further. A rating that changes order whenever a coefficient moves slightly is telling you that the underlying evidence does not separate the candidates.