Updated September 15, 2026

How the defenses are ranked

Start the defense whose opponent is expected to score the fewest points. That is the whole method: one number per team from the betting market, no fantasy projections, no team statistics.

The calculation

A defense favored by 8 in a game with a total of 40 faces an opponent implied to score 16. Lower is better. These are implied NFL points, not projected fantasy points.

For each game we pair every US bookmaker's spread and total from The Odds API, compute opponent points per book, and take the median. Equal totals share a rank; alphabetical order only decides display. A gap of half a point is a weak preference.

Why this signal

On 2018–2025 seasons, held out one season at a time, the weekly rank correlation between opponent implied points and actual fantasy D/ST points averaged 0.30. A defense's own scoring history managed 0.13, and adding team statistics to the market number helped only slightly. Correlation says whether the ordering tracks results; it is not an accuracy percentage, and a week's best pick still scores zero sometimes.

The backtest used closing lines, approximate custom scoring, and no special-teams touchdowns. It supports the approach; it does not guarantee any week.

Sources and editions

Each edition is a dated snapshot with its bookmaker inputs, a JSON download and a CSV. Editions are never edited; a correction is a new edition at its own URL. Earlier editions used lines distributed by ESPN, nfelo or nflverse; nfelo's own model spread is never used as a bookmaker line. A game without a posted line is left unranked, not treated as a bye.

How the picks are checked

The accuracy report grades every published edition against actual D/ST points under the labeled scoring: rank correlation, the average of the first five picks, and the average of all completed defenses. Pending games are excluded; zero and negative scores count. The original forecast for each week stays at its own URL, misses included.