The Confidence Ladder

Every NFL pick ranked by BEACON™ — the confidence rating from Buildology. One number, 0-100, powering every pick.

Loading NFL Portfolio…
How to read this board

The BEACON score is not a win probability

It is how far our number sits from the market, scaled by how much that gap deserves belief. A low score does not mean the lean is likely wrong. It means the book and the model nearly agree, so there is nothing to bet. Raiders +6.5 at BEACON 4 means our spread and theirs differ by under a point: noise, not a signal against that side.

BEACON 75+Conviction. A full-size edge, four or more games of data, and the scoring model agreeing. The portfolio stakes 2%. Rare by design. BEACON 55-74Strong. A real edge with either the data or the second model behind it. Stakes 1%. BEACON 0-54Skip. The market looks right, or the sample is too thin to say otherwise. Most of every slate lives here. The lean is shown so you can see which way the model tilts; it is not a recommendation. BEACON 0Extreme. Our raw number disagrees with the market by more than any real inefficiency could explain. When that happens the model is the one that is wrong. Never staked, whatever the edge.

What goes into the number

Edge. The gap between our shrunk number and the market's: points on a spread or total, percentage points of win probability on a moneyline. A full edge (score 100 at perfect quality) is 7 points on a spread or total, 15 points of probability.

Shrunk. Early in the season our rating is mostly last year plus a few games, so every number is pulled toward the market in proportion to how little we know. At two games played we keep about 20% of our own opinion. That is why raw and shrunk model numbers both appear on the Edges page.

Quality. Multiplies the edge: 0.35 floor, up to +0.45 for games played (full at four), up to +0.20 when our points-for/against projection, built from different data than the rating, leans the same way against the market. No edge means no score at any sample size.

Line movement. +8 when the spread has moved toward our side since we first saw it, -4 when it has moved away. Spreads and moneylines only.

Reading a row

Model: who we favour and by how many points (shrunk), the home side's win chance (the same number expressed as a probability), and our projected total points for the game. These are projections, not scores.

Absences. Listed starters ruled out or in doubt are taken off our expected margin before the market comparison (NFL: ESPN depth charts, refreshed hourly; CFB: operator-entered, ESPN publishes no college injury data). Out or IR counts in full, Doubtful 85%, Questionable 35%; a QB is 4 points in the NFL and 5 in CFB, other starters half a point to one. The Absences tile shows who and how much; the Matchup page shows the full list.

Spread / ML / Total: the side our number leans to on each market, with its BEACON score. No lean on a moneyline means the model rates both sides below the price the book is offering, which is common when we agree with the market: the book's margin makes both sides negative.

Every game and every market is logged before kickoff and graded after, so these scores are checked against results. That record, not this page, is the proof.

Loading this week's board…
NFL Prove-It Portfolio: a simulated $1,000 bankroll that auto-stakes 2% on every Conviction Pick and 1% on every Strong Signal, per market. Picks are captured Wednesday morning at the price posted then, graded from the final score within hours of the game, and never edited. Live since 18 Sep 2026. The full pregame record of every game, bet or not, is on The Record. Not betting advice. Past performance doesn't guarantee future results. Bet responsibly.