Proportional or power: the two margin-removal methods disagree on uneven markets, and the disagreement is permanent until you test it against something real. The something real is the closing price. The close is the market's most informed number, so the removal method whose fair prices land nearer the close is, on your data, the better model. Here is how to run that test honestly.
The claim you are testing
Each removal method is a claim about where the margin sits. Proportional says it is spread evenly across outcomes; power says it hides more in the longshots, as the no-vig calculator frames the modeling choice, checked 2026-11-07. Both produce fair prices, and on balanced two-way markets they barely differ. The test matters exactly where they diverge: favorites versus longshots.
You are not testing which formula is prettier. You are testing which one's fair prices predict the closing price better on the markets you actually read.
The test set
You need matches with two snapshots each: the price some hours before the close, and the close itself. Historical boards give you both, with the caveat the historical data guide raises first: check the sampling and the timestamp semantics before you trust any archive, checked 2026-11-07. A close that is not actually the last price before the start is a broken benchmark, and the whole test inherits the break.
A few hundred matches is a sensible start. Split them by price level, because the methods' argument is about favorites and longshots, and a pooled answer hides the split you are looking for.
The scoring rule
For each match and each method: remove the margin from the earlier snapshot, and compare each fair price with the closing price on the same side. Score with something boring like the mean absolute gap between fair and closing prices, per method, per price band. The method with the smaller gaps is the one whose margin story fits your markets.
Run it per sport if you can. The answer can differ by league, because where margin hides is a market habit, not a mathematical law.
Read the answer with humility
A test like this ranks the two methods on your data; it does not crown a universal winner. A result that holds across three months of your sports is worth adopting as your house method, and worth re-running when your market mix changes.
What it settles permanently is the habit of choosing by argument. The method question is empirical, and the close is the referee you already have on record.
Two formulas, one benchmark, a few hundred matches. The margin argument ends wherever your own data says it ends.