A spread can break a quarterback tie. It cannot build the ranking for you. This board contains team prices, not player projections, dropback rates, injury reports, or league settings. So the Week 1 call is narrower than the title sounds: when two quarterbacks are otherwise close, prefer the healthier starter attached to the stronger game environment. Do not bench a materially better player because one early book posted a field goal.
The top of the board is a tiebreaker, not a depth chart
Provenance tier: live_pregame; source table: game_odds. The Chargers were listed as 4.5-point home favorites against the Rams . Miami was a 4.5-point home underdog to Atlanta . Those were the widest NFL prices in this snapshot, which makes them the cleanest places to ask a game-script question.
That question is not “which quarterback scores more?” The spread does not answer it. It asks whether a close lineup decision gets a small nudge toward the favored offense and away from the underdog offense. A mobile quarterback chasing points can beat that rule. A low-volume passer protecting a lead can lose it. Use the number after the player evidence, never instead of it.
The field-goal tier is mostly noise for start-sit
Provenance tier: live_pregame; source table: game_odds. Tennessee was minus 3 against Chicago . Philadelphia was minus 3 against Cincinnati , and Indianapolis was minus 3.5 against Detroit . Those prices describe competitive games. They do not separate fantasy starters from benches without player-level inputs.
The sensible read is to leave established starters alone and use the spread on streaming decisions. If two waiver options have similar roles, the favored side gets the nod. If one quarterback runs, throws more, or has a materially stronger supporting cast, that evidence wins. The model-versus-consensus guide is useful precisely because it asks for a real model before claiming a gap. Make the final call on the start/sit tool once that model exists.
The narrow favorites deserve even less confidence
Provenance tier: live_pregame; source table: game_odds. Baltimore was minus 2.5 against Washington , and Las Vegas was minus 2.5 against San Francisco . Buffalo was minus 1.5 against Pittsburgh , Kansas City was minus 1.5 against Seattle , and Dallas was minus 1.5 against New Orleans .
A one-score price is not a command to start the home quarterback. It is the market saying the teams are close. In that range, rushing work, red-zone role, offensive line health, and your scoring rules can swamp the spread. The number belongs in the final column of the decision sheet, not the first.
The underdog board is a warning label
Provenance tier: live_pregame; source table: game_odds. The Jets were 3-point home underdogs to the Giants . Cleveland was a 2.5-point home underdog to New England . Carolina was a 2.5-point home underdog to Houston , Green Bay was plus 2.5 against Arizona , and Jacksonville was plus 1.5 against Tampa Bay .
Those prices are reasons to inspect the quarterback, not reasons to erase him. Trailing can create pass volume; it can also create sacks, turnovers, and empty possessions. A start-sit column that pretends the spread settles that tradeoff is selling certainty it does not have.
The honest Week 1 rule
Set the player first. Check health, starting status, rushing contribution, and league scoring. Then use this board as a modest script adjustment. The Chargers side gets the strongest positive nudge in the cited snapshot; the Miami side gets the strongest caution. Everything clustered around a field goal is close enough that player evidence should dominate.
Every cited line came from one book, so even the market context is provisional. The cure is concrete: compare another current quote and revisit the call after injury news. That is less dramatic than “start every favorite.” It is also how a spread-only column stays useful without pretending it ran a quarterback model.
NFL ATS cover-margin distribution
Bars count completed NFL schedule rows by closing-spread cover margin using the repository canonical home-margin grading convention.
Model calibration from graded predictions
Calibration points render only when a verified source binds prediction probabilities to settled outcomes for the same observations.
