The one thing the market might undervalue: which teams simply kept their guys

After ruling out every version of 'we can predict the game better,' one lead survived — and it's not about talent, which the market prices efficiently. It's about continuity: not how good a roster is, but how much of it came back.

Early in the season, before any 2025 games exist to judge a team by, we bet the side that returned more of its production from the prior year. In 2023 and 2024 that won 55 to 56% of the time against the spread — genuinely past break-even, and notably, a signal our own model wasn't already using. The market, it seems, may over-react to roster churn: it gets excited about splashy transfers and new faces and slightly under-credits the boring team that just ran it back.

Why it would make sense — and why we're not celebrating yet

The logic is clean. In September there's no film on the new season, so everyone leans on last year plus preseason hype. Continuity is quietly predictive in that window because a team that returns its quarterback, its line and its skill players is more knowable than the market's hype-weighted number admits. By October, once real games pile up, the edge evaporates — form replaces roster guesswork, exactly as you'd expect if the effect is real.

So we're holding it honestly: a mechanistically sensible, not-yet-proven hint that the one inefficiency left on the main line isn't in how good teams are, but in how much the market discounts the ones that simply stayed together.