Whet for the NFL.
Scanned every 3 minutes. Graded against real results. Game lines are on the live board. Player props are not.
Game lines are on. Props are not.
Spreads, totals and moneylines have been on the live board since Sep 10. Player props are not on the board. The scanner fetches NFL every 3 minutes and the settle job grades every NFL pick against the result, the same as baseball.
Holding props back is a decision, not an outage. NFL was promoted to the board for a single evening in August and pulled back the same night, because promoting it meant a sport with no record of its own inheriting thresholds fitted on baseball. Game lines went back on Sep 10, on our call. Props stay off — too few graded picks.
Small, and labelled small.
Here is every NFL segment in the pre-registered ledger. The sample floors — 200 graded picks and 14 independent game dates per segment — were written down before any of these results existed, and no segment below them reports a rate. Most of the NFL rows are a long way below.
231 graded picks across 8 segments, on at least 5 game dates. No segment has reached the sample floor yet, so none of them reports a rate.
swipe the table for ROI and the 95% CI →
| Segment | Graded | Win rate | ROI | 95% CI (win · ROI) |
|---|---|---|---|---|
| prop · over | 42 | Too early — 42 of 200 graded picks and 2 of 14 game dates. No rate is published below the floor. | ||
| spread · over | 38 | Too early — 38 of 200 graded picks and 5 of 14 game dates. No rate is published below the floor. | ||
| total · over | 36 | Too early — 36 of 200 graded picks and 5 of 14 game dates. No rate is published below the floor. | ||
| spread · under | 35 | Too early — 35 of 200 graded picks and 4 of 14 game dates. No rate is published below the floor. | ||
| total · under | 34 | Too early — 34 of 200 graded picks and 5 of 14 game dates. No rate is published below the floor. | ||
| prop · under | 24 | Too early — 24 of 200 graded picks and 2 of 14 game dates. No rate is published below the floor. | ||
| moneyline · under | 12 | Too early — 12 of 200 graded picks and 4 of 14 game dates. No rate is published below the floor. | ||
| moneyline · over | 10 | Too early — 10 of 200 graded picks and 4 of 14 game dates. No rate is published below the floor. | ||
Pre-registered basis: every pick the frozen engine issued after 2026-07-14 and graded against the real outcome. Floors — 200 graded picks and 14 independent game dates — were written down before the results existed. Wins and losses both, every segment, no subset.
Spreads, totals and moneylines are on the board. Props are not. These floors decide whether a rate prints, not what goes on the board — that bar is below.
What puts NFL props on the board.
A segment goes on the live board when it has all three: 200 graded picks, a win rate at or above 55%, and a positive return. Per segment — not per sport, and never all-or-nothing. NFL game lines went on Sep 10 ahead of that bar, on our call, with a record still short of 200 graded picks. That is an override of our own bar, not a pass.
NFL props are still issued and still graded, so the record builds where you can check it. You are not being shown a market we cannot yet defend, and you are not being told it is missing.
There is no weather model.
This page previously described wind gates that rejected passing overs above 15mph, live game-time forecasts, precipitation gates, and target-share and snap-count floors calibrated per position. Thresholds for those exist in the code; the data they read has never been connected, so their defaults always pass and no NFL pick has ever been rejected by any of them.
They are off this page. What is left is the position we can actually defend: NFL is scanned, NFL is graded, its game lines are on the board, and its player props are not.
Watch the record build?
Every graded pick, wins and losses, is on the public ledger — including the markets that are not on the board.