A star being ruled out does not create free money. It creates a new offense. Touches have to go somewhere, minutes have to be reassigned, and the market has to decide which teammates inherit creation, finishing, rebounding, and defensive work. The edge, when there is one, lives in that redistribution rather than in the headline.
NBA load management prop edges are therefore a timing and role problem. The injury status must be real, the lineup assumptions must be explicit, and the market snapshot must come from after the news. Anyone comparing a post-scratch projection with a pre-scratch line is not measuring an edge. They are measuring a stale screen.
Start with the status, not the rumor
Record the official designation, the source, and the time it changed. A player moving from questionable to out is different from a reporter speculating that rest is likely. Team announcements and league injury reports belong in the settled input set. Unconfirmed social chatter belongs in a watch queue.
Do not overwrite the earlier status. Preserve the sequence so the model can be reconstructed later. That history explains whether a price moved before confirmation, after confirmation, or after the starting lineup was announced. It also prevents the backtest from granting itself news that was not public yet.
Redistribute jobs, not box-score averages
When a high-usage player sits, the missing possessions do not flow evenly to the next names on the roster. One teammate may become the primary ball handler. Another may gain shots without gaining assists. A reserve may absorb minutes but spend them standing in the corner. The role tree matters more than the season scoring table.
Build the adjustment from lineup evidence: who initiates when the star is off the floor, who shares the court with the replacement, which actions the offense keeps, and which actions disappear. Use recent coaching choices as evidence, but do not assume the last emergency rotation is permanent. If several replacements are plausible, model scenarios or pass.
Minutes are the first prop input
Every counting-stat projection starts with expected playing time. A teammate can gain usage and still miss an over if the rotation becomes deeper. A reserve can become a starter while keeping roughly the same total minutes. A blowout-sensitive role can gain early work and lose the closing shift.
Write the minutes assumption next to the projection and show what would change it. Starting lineup confirmation, a second inactive, foul risk, a minutes restriction, or a returning player can all break the original allocation. The model should react to those facts, not cling to a stale point estimate.
Different props inherit different pieces of the vacancy
Points depend on shot volume, free-throw opportunities, and efficiency in the new role. Assists depend on initiation and whether teammates finish the chances. Rebounds depend on position, lineup size, shot profile, and who is removed from the floor. Three-point attempts can rise even when total usage barely changes.
That is why a universal star-out bump is bad practice. Translate the absence through the mechanism for each market. A secondary creator may be the best assist candidate. A center may gain rebounds when the missing player normally cleans the defensive glass. A scorer may draw harder coverage and become less efficient despite more touches.
The market timestamp is part of the feature set
Capture the exact line and price available after the status change. If the book suspends the market, there is no actionable price and no honest edge. When the market returns, note whether the projection used the confirmed starters or only the out designation. Those are different information states.
Line shopping can reveal disagreement, but the comparison must be contemporaneous. Do not pair the softest number from one time with the sharpest benchmark from another. A historical study should store each snapshot rather than reconstructing the board from a final close.
Correlation can make several good projections one crowded bet
Teammate props from the same absence often share the same game script. More creation for one guard may also mean more finishing chances for a wing, but both can fail if the game slows or the rotation changes. Treating each over as independent understates the risk.
Group exposures by the assumptions that drive them. If several positions all require the same player to handle the ball and the same game to stay competitive, the portfolio has one thesis wearing several jerseys. The cleanest response may be choosing the market that expresses the role change most directly.
Grade the alert honestly
Keep the status time, projection time, available line, closing line, and result. Grade whether the projected direction was right, whether the entry beat the later market, and whether the role assumption held. Those questions can disagree.
Any published record needs a defined window, sample, and ATS win rate in wins, losses, and percentage form. Props that are not ATS should be described by their graded win rate without fabricated standard pricing. If the sample or price history is incomplete, say so and withhold the performance claim.
When the right answer is no play
- The absence is not confirmed by a reliable source.
- The replacement lineup has several credible versions.
- The market is suspended or the captured price is stale.
- Another questionable player can rewrite the same roles.
- The model has no comparable lineup evidence.
- The available prop already reflects the full projected shift.
Late news rewards speed only after the model earns the right to be fast. Confirm the status, rebuild the roles, match the price to the information, and decline the rest. The scratch is news. The edge has to be proved.
Breakeven win rate at recorded American prices
Breakeven probability is calculated only from American prices that were actually captured in the odds-history table.
