Why the Model Is Fading Week 1 Passing Props

Across 2023, 2024 and 2025, we graded every Week 1 player prop in the seven core markets against the number the market posted. The market's prices implied the over would hit about half the time. It hit 44%. Weeks 2 through 5, same markets, same method: 49%, right where it should be. By Week 2 the effect is gone.

Passing is where it's most potent.

Week 1 over hit rate, 2023 to 2025

Receiving props ran close to normal. Passing props ran nowhere close. A quarterback priced at a coin flip to clear his yardage line cleared it less than a third of the time.

The lines are anchored to last season. Week 1 numbers get posted off last year's box scores and a summer of storylines. But the first game of a season is played by a different offense than the one that finished the last one: new pieces, new coordinators, a game plan that hasn't been stress-tested. The market prices the quarterback it remembers. The field produces the one that shows up in September.

Week 1 passing has been shrinking, and the lines haven't caught up. This is a recent thing. From 2012 through 2022, starting quarterbacks actually threw for slightly more in Week 1 than their own full-season pace, about 4% more. Since 2023 that flipped. Week 1 starters have thrown for roughly 8% less than their own season pace, and 71% threw one touchdown or fewer, against 52% the rest of the year. The old "openers are shootouts" instinct is priced in. The new reality isn't.

It's a game-level thing, not a player-level thing. In the 37 Week 1 games we have priced, both quarterbacks went under their yardage line in 16 of them. Both went over in 4. When an opener turns into a run-heavy, field-position game, it takes both passing props down with it.

The market doesn't fix it by kickoff. You'd expect sharp money to drag the lines down during the week. It doesn't, or not enough. At closing prices the over still hit 44%. This isn't a "bet it early" edge. It's simply that Week 1 stats fall short of Week 1 lines, and the price never fully reflects it.

It only shows up in the coin-flip zone. The shortfall lives in props priced between 45% and 55%. When the market is confident, say a prop priced at 55% or better to hit, it still hits about 65% of the time. That matters for how the model uses it, which is next.

For Week 1 only, the EP model now runs a Week 1 regime:

• A Week 1 adjustment by prop family, learned from the last three Week 1s. Passing props get the largest, about 11 points of probability against the over. Rushing gets about 5.5, receiving about 5.
• Only near coin-flip lines. The adjustment is full strength on props priced within 10 points of 50/50 and fades to nothing by 20. A lopsided line stays lopsided. The model won't turn "zero passing touchdowns" into a top pick just because it's an under.
• Last season's matchup read is switched off. Usage and matchup data from 2025 is stale by September. Pretending it isn't was costing the Week 1 board.
• The board stays balanced. Unders are capped at 65% of the 90+ band, with the best-priced receiving and rushing overs filling the rest. That's why the vast majority of receiving and rushing props you see at 90+ this week are overs.

In a walk-forward backtest on the 2024 and 2025 Week 1s, this regime graded out at +11.2% on the 90+ band with a 59% hit rate, against roughly breakeven for the standard model on the same two weeks. Then on Week 2, the regime hands back to the regular model, because the effect is gone by then.

Yes, it's nearly every quarterback. This week, 97% of passing props lean under, and 52 of the 54 passing props scored 90 or better are unders. The receiving and rushing overs alongside them are the balance, drawn from the families where the Week 1 shortfall is smallest.

One thing to keep in mind: three Week 1s is three data points. The pattern showed up in every season and every way we sliced it, but it's young, and 2025 was its weakest year. If it repeats at even half its historical size, the unders pay. If it doesn't, Week 1 is a coin flip like any other week.

We'll grade every Week 1 pick and publish results in Model Trends. Week 2, normal model resumes.