Player-prop expected value starts with the exact prop. This source set does not contain one.
It contains historical game-market observations. Those can describe context. They cannot substitute for a player line, price, projection distribution, or calibration record. The old copy filled every missing layer with confident product claims and called the result FairPlay.
Data limit: no player-level projection or real prop price is attached to this module. The article therefore explains the contract and publishes no player recommendation.
A game market is context, not a player baseline
The Chiefs–Buccaneers source and Jets–Steelers source show that game-market observations reached the historical feed.
They do not tell us how many attempts a quarterback makes, how targets are divided, whether a running back owns the goal-line role, or how a player outcome is distributed around a sportsbook line.
A spread or total can constrain a coherent player model. It cannot allocate production by itself. The shortcut “favorable game environment, therefore over” skips every player-level input.
The prop market needs its own price
Expected value compares the model probability of the exact over or under with the break-even rate implied by the offered price. Each prop therefore needs its own line and price, preferably both sides of the same market.
Borrowing margin from a game moneyline or assuming a standard prop price produces precise-looking nonsense. Strip the margin from the market you are actually betting.
Line shopping starts after those terms are captured across sportsbooks. A list of game markets cannot support the comparison.
The vig guide explains normalization. The rule here is simpler: neighboring markets are context, not substitutes.
The projection must be a distribution
A point estimate is not enough to price an over or under. The model needs an outcome distribution so it can estimate probability above or below the offered line and preserve push mass where the market can land exactly on an integer.
Inputs should be available before the wager. Role, participation, opponent, pace, weather, and injury state need explicit cutoffs. The provenance allow-list should reject information created after kickoff or settlement.
Uncertainty belongs in the output. A projection that barely clears the break-even threshold may not survive ordinary model error. A pass can be the correct result even when the point estimate leans one way.
Calibration decides whether the probability deserves trust
A model that assigns similar probabilities should win at similar rates over a named historical window. Report the evaluation window and n. Do not use a batch timestamp as a time axis, and do not convert a profit figure into a record.
Prop results should be graded on the exact market definition and settlement rules. Voids, pushes, stat corrections, and book-specific terms must remain visible in the ledger.
The Studio can expose the projection assumptions. The track record can grade frozen decisions. Neither should imply a calibrated player model exists because a game spread is present.
FairPlay is a contract, not a feature claim
This article cannot verify that a product ingests a certain number of books, refreshes at a particular cadence, filters by a named tier, or publishes the claimed routes. Those capabilities were asserted without supporting repository evidence in the module.
The defensible meaning of fair play is procedural: use real market terms, pregame data, explicit uncertainty, reproducible calculations, and honest empty states.
If a future product implements that contract, the UI should show the player, stat, line, price, capture time, projection distribution,
The minimum publishable result
A prop EV claim needs the player, market, side, line, price, sportsbook, and capture time. It needs a probability distribution built independently from pregame inputs. It needs a calibration check on comparable historical props.
When every field exists, compare the offered break-even rate with the model probability and keep uncertainty in the decision. When one is missing, name it.
The result for this module is NO PROP EV. The cure is a real player market and a calibrated pregame distribution for the same stat. Game context can wait in the evidence ledger until the player evidence arrives.
Breakeven win rate at recorded American prices
Breakeven probability is calculated only from American prices that were actually captured in the odds-history table.
