What a missing star costs a team

The question

"How much does losing him hurt tonight?" is the question behind most matchup questions, and a team rating built from a season's results cannot answer it, because the season did not know he was going to sit. This study asked how much of the rating's gap to the betting market lives in exactly those games, and what knowing the lineup at tip is worth.

What was tested

Six NBA seasons (2020-21 through 2025-26) and six NFL seasons (2020 through 2025), every regular-season game with a closing moneyline. A star's absence was read after the fact from the box score: in the NBA one of a team's top three players by season minutes with no line or no minutes that night, in the NFL the season's top passer by attempts. Games were sorted by whether the home star, the away star, both or neither sat, and by how far into the season they fell, and the rating's forecast error against the close was compared bucket by bucket.

A second run priced the absence. A synthetic injury report defined each team's regulars from the season so far and charged each absence as the missing player's value times a weight fitted on the earlier seasons and read once on the newest, in points per game, plus-minus per game or minutes for the NBA and a flag for the NFL. A third run rated the players instead of the team and summed the lineup, with the same knowledge of who dressed.

What turned up

Star absence is where the market's edge concentrates. In NBA games where a home star sat, the rating's error is 16.0% worse than the market's and it over-rates the home side by 11.8 points of win probability. With no star out it is 5.1% worse. In NFL games where the home team's top passer sat, the rating is 31.4% worse than the market and carries the same 11.8-point home bias, against 4.7% worse with everyone in.

Knowing who sits at tip is worth about 1.5 points of Brier in both sports. The NBA gap falls from 6.4% to 4.8%, at a weight that makes a 25-point scorer's absence worth about four points of margin. The NFL gap falls from 8.0% to 6.5%, with a missing top passer worth about four points. Plus-minus per game, the other currency tried, carries almost nothing. That closes about a quarter of the gap on a team rating. Rating the players and summing the lineup with the same knowledge closes about half: 2.8% behind the market on the holdout, against 4.4% for the team rating with the lineup patched in.

The gap also moves with the calendar. In the NBA it widens past 62 games into the season, to 12.1%, where absence is planned rest rather than injury. In the NFL it is worst at four to six games in, at 10.9%. MLB is flat near 1% all season.

What it means when you ask

A rating that ignores the injury report over-rates the depleted side, and by a lot when the missing player is a star. The NBA Power Rating now sums the players expected on the floor and drops anyone the report marks Out, so an answer names the lineup it assumed and who is missing. The MLB and NFL ratings read the season, so an injury question in those leagues gets the rating beside what is currently true of the roster, and the four-point figure above is the size of the thing the rating cannot see.

What it does not mean

The market prices an absence before any rating can, so none of this is an edge. The four-point figures are fitted constants from a synthetic report, not a player's value. Nothing here was measured for MLB, where the analog is the starting pitcher, for the playoffs, or for questionable and probable designations.

The record

Measured 6 September 2026 with diagnose_power_rating_market.py, backtest_availability.py and backtest_player_rating.py. The absence flag is a post-hoc proxy from the box score, the ceiling of a real feature rather than a feature, because no history of a live injury feed exists before August 2026. The player-level result belongs to Power Rating v2, which is built and awaiting release. The full portfolio and its testing rules are on the decision metrics page.