Do referees and umpires change the game?
It is one of the most repeated claims in betting: this referee calls more penalties, those games go over. The
claim is testable, and the data here covers every NFL, NRL and NBA official with at least 25 games.
Why points per game is the wrong number on its own
An official whose games average more points may simply be assigned bigger fixtures — prime-time matches
between high-scoring teams. Ranking crews by raw scoring measures the schedule as much as the whistle. So this
page shows the posted total for those same games next to the average, and the difference between them.
That difference is the only part that could be attributable to the official, because the market had already
accounted for who was playing.
What the comparison shows
Very little. The gap between the highest and lowest scoring officials looks dramatic, but it shrinks sharply
once you measure against the posted line instead of the league average, and the remaining spread is consistent
with sample size. Fewer officials clear the 52.4% over break-even than chance alone would produce.
Why the leader is almost certainly noise
Comparing thirty officials is thirty tests. At a hundred games each the standard error on an over rate is near
five percentage points, so somebody will always sit at 60%. The figure printed above the table is the probability
that the actual leader looks that good by luck given how many are being compared — and on this data it is
usually above 90%.
AFL umpires
Not available. No umpire is recorded against any AFL match in the data, so the AFL is reported as unavailable
rather than shown as an empty table. NFL, NRL and NBA coverage all exceed 90% of games.
Is this the whole crew or one official?
The named official is the lead referee or
crew chief. Full crews are not held, and crews change composition between seasons, so a crew-level claim could not
be supported by this data even if the lead-official one were.
Why a 25-game minimum?
Below that the standard error on an over rate exceeds ten
points, which is wider than any effect being looked for. Officials under the threshold are excluded rather than
listed with a number that cannot mean anything.