Ask the league how referee crews get assigned to games and you get a paragraph about fairness and rotation. Ask seven seasons of schedule data and you get actual rules. We pulled every crew assignment from 2019 through 2025, 1,956 games, and tested whether the madness has a method. It does.
Three rules nobody wrote down
First, a season cap. A crew almost never works the same team three times in one season. Since 2022 it has happened exactly once. Twice is routine; about a quarter of crew and team pairings meet twice a year. Three times is close to forbidden.
Second, spacing. When a crew does see a team again, the re-meetings keep their distance. Only 4 percent happen within four weeks of the first meeting. Most land six to ten weeks apart. If a crew just called your team’s game, you will not see them again for about a month and a half.
Third, the math of the roster. Seventeen crews cover roughly sixteen games a week, so nearly every crew works every week and one rotates off. Before any modeling starts, the guessing pool is already smaller than it looks.
How far the rules get you
We turned those constraints into a predictor and backtested it against the 2025 season. Guessing at random picks the right crew about 6 percent of the time. The elimination rules alone nearly double that. Add one more idea, ranking which crews are most overdue to see a given team, and the predictor names the exact crew 15 percent of the time and has the right crew in its top three 44 percent of the time.
| Predictor | Picks the exact crew | Right crew in top 3 |
|---|---|---|
| Random guess | 5.9% | 17.6% |
| Elimination rules only | 10.7% | – |
| Rules plus overdue ranking | 15.1% | 44.1% |
That is a lunch break heuristic beating chance by two and a half times. A serious model has room beyond it. The weekly slate is really a matching problem, since each crew works exactly one game, and the league visibly places its top graded crews on the prime time windows. We think the ceiling sits somewhere between one in four and one in three on the exact crew, and above 60 percent for top three.
The honest part
Two caveats belong in print. Our backtest knew which seventeen crews were working each week, a small information leak in the predictor’s favor. And the league announces the real assignments midweek anyway. So the prediction game lives in the early week window, the days when fans are guessing and the schedule data already knows more than they do.
That is also why the right shape for this is probabilistic, not a hard pick: your team’s three most likely crews, each with its penalty history against your team. When the real assignment drops midweek, it is in the app within hours either way, next to what a crew report should actually tell you. And if you want the official version of how assignments work, we wrote up the league’s own process; this piece is what the schedule says the process really is.
The whistle has a record. It turns out the schedule does too.