Updated Sep 7, 2026 · Week 1 board.
The market’s own line versus our own model
Every pick we publish keeps three numbers. The market line at the time. Our own number. The distance between them, which is what we call the edge. On Dallas at New York, spread market: market line −2.5, our number 1.47, edge 3.97. Do it yourself, it's subtraction: 1.47 − (−2.5) = 3.97. Same game, total market: market line 48.5, our number 53, edge 4.5. And again: 53 − 48.5 = 4.5.
One convention, read correctly
Here's the part that trips people. Both of those numbers use a convention where a positive figure means the home side is the favorite. That's backwards from the −X you see on a live odds board, where the minus sign marks the favorite. So a market line of −2.5 in our own note means the home Giants were priced as 2.5-point underdogs when we published. Our number, 1.47, has the Giants winning outright by 1.47 points. Read that swing end to end, from a 2.5-point market underdog to a 1.47-point projected winner, and you've got where the 3.97 comes from. Read the convention backwards and this looks like a mild disagreement instead of the wide one it is.
A worked example, in one table
| Market | Line | Our number | Edge (points) |
|---|---|---|---|
| Spread (home, Giants) | −2.5 | 1.47 | 3.97 |
| Total | 48.5 | 53 | 4.5 |
The board has not moved — only the book you check has
We keep tracking this same spread pick under a model we call “House NFL Spread.” It was generated Aug 21, 2026, and on our Sep 7, 2026 pass it still carried the same 1.47. Its reference book now is Bovada at −3 in our convention. First read, that looks like half a point moved against the Giants since we published. It didn't. Live odds today have Pinnacle quoting the Giants at +2.5, with DraftKings and Bovada both at +3. That's ordinary board notation, where a plus sign means the home team is the underdog. Flip them into our convention by multiplying by −1 and you land on −2.5 at Pinnacle and −3 at Bovada. Those are the two numbers you've already read. The −2.5 is what our published pick referenced. The −3 is what our tracker reads today. The half-point you spotted is Pinnacle against Bovada. It isn't August against September.
| Book | Today’s spread (home, standard notation) | Same number, our convention |
|---|---|---|
| Pinnacle | +2.5 | −2.5 |
| Draftkings | +3 | −3 |
| Bovada | +3 | −3 |
From a points gap to a probability gap, with the assumption stated
A 3.97-point edge isn't a probability, and turning it into one takes an assumption I'd rather hand you than bury. The common rule of thumb says a point of NFL spread is worth roughly 3 points of win probability near a given number. That's a rule of thumb. Our own numbers don't prove it. Run it anyway: 3.97 points at 3 points a point comes to about 11.9, call it 12 percentage points. So we're claiming something like a 12-point swing against what the market's spread implies, and every bit of that rests on the borrowed rule rather than on arithmetic we did ourselves.
What continuous tracking adds, and what it still does not prove
Holding the same number from Aug 21 to Sep 7, 2026, at 1.47 both times, tells you something real. The projection's been stable. Nobody quietly revised it. It's still a pre-game estimate with no settled result behind it, and the count of graded picks sits in the sourcing note at the bottom. A number that hasn't moved is still a number that hasn't been tested.
Want the mechanics of turning one price into a fair, no-vig probability instead of comparing our number to a line? That's our companion piece, worked on a flat New England-Seattle board. Prices for this game, book by book, sit on the NFL odds desk next to NFL picks, and the rest of our Week 1 board is there with them. The bankroll question that comes after an edge estimate is the Kelly criterion explainer (how much of your bankroll to bet). If Bovada moves off −3 before kickoff, refresh the desk and recheck this edge.
The Receipts Drawer
Same number twice, 1.47, two and a half weeks apart. All that proves is we didn't flinch while the board around us got noisy, which is a different thing from being right. Turn the 3.97-point edge into a probability and it lands near 12 percentage points, but only if you carry that borrowed rule of thumb along with it. Nothing on this pick has been graded. Watch Bovada's −3 tonight: if that line moves, this 3.97-point edge moves the same distance, and what you're holding is worth less than it looks.
