Updated Sep 7, 2026 · Week 1 board.
Three books, one number, both sides of the ball
Kickoff is 8:20 p.m. ET, Sep 9. Pinnacle, DraftKings and Bovada all post the same spread and the same total on their latest check.
| Book | Spread (home, Seattle) | Total | Moneyline (home / away) | Last checked |
|---|---|---|---|---|
| Pinnacle | −3.5 (+101 / −114) | 44.5 | −179 / +156 | 8:52 p.m. UTC, Sep 7 |
| DraftKings | −3.5 (−102 / −118) | 44.5 | −175 / +145 | 10:07 p.m. UTC, Sep 7 |
| Bovada | −3.5 (−105 / −115) | 44.5 | −185 / +160 | 10:07 p.m. UTC, Sep 7 |
That is a tight three-book market. The same −3.5 and the same 44.5 sit at every book we track, and only the prices around them move. Bovada pays the home side at −185 and the away side at +160. DraftKings is shorter at −175 and +145.
Take the same side at three books and you pay three different levels of vig (the book’s cut) for the same 59% implied chance the home team wins. One published edge number is one input on a market like that.
We already published two of the three markets here
Two picks on this game are already live. The over at 44.5 is one. Our number there is 45.48, an edge of 0.98 points, and our confidence in it is 0.568 out of 1. The moneyline on the home team is the other.
Our number is 0.59, that same 59% implied win chance, against an edge of 0.09 at a confidence of 0.59. Both edges are small. Under a point on the total, under ten cents of probability on the moneyline, and both cleared the bar we use before we publish anything. The spread cleared nothing, and the 36 numbers below are why.
36 spread models, one line, no published number
We ran 36 models on the spread, one number apiece, keeping the latest run wherever a model ran twice. Thirty-four of the 36 take New England against the market line. Two take Seattle. Thirty-four price against the current 3.5.
Two are still working off a 4.5 last run on May 16, before the market moved. The numbers run from −7.8343 up to 12.916, a little over 20 points of disagreement on one game.
| Model | Model number | Pick & line | Run on |
|---|---|---|---|
| NFL Spread (MARS) | 12.916 | home 3.5 | 2026-09-04 |
| House NFL Bridge - Preseason Prior (Elo) - Spread | 5.7004 | home 3.5 | 2026-09-04 |
| NFL Elo Power Ratings Spread (XGBoost) | 1.2344 | away 3.5 | 2026-09-04 |
| Simple Elo Spread Model (Bayesian) | −3.5849 | away 3.5 | 2026-09-04 |
| Pressure & Efficiency Model (LightGBM) | −7.8343 | away 3.5 | 2026-09-05 |
Set that against the two markets we did publish. The total models packed in close enough around 44.5 to clear a sub-1-point edge. The moneyline models packed in close enough to clear a sub-10-cent edge. On the spread we hold one number that likes Seattle by 12.9 points and another that likes New England by 7.8, out of the same 36.
That is the sourced reason no spread pick has appeared. Two of the 36 are working off a stale line. NFL Market Efficiency Spread (XGBoost) last ran May 16 at 5:06 p.m. ET with −4.7183 against a 4.5 line.
NFL Minimal Signal Spread (XGBoost) ran the same minute with −5.5189, also against 4.5. Neither carries a current line or price, so neither has been rerun since the market came from 4.5 down to 3.5. Take those two out and the other 34 still run from −7.8343 to 12.916. The stale pair isn't what makes the range wide.
Fixing it would still leave 20 points between our own numbers.
What would turn this into a published number
Our internal bar for publishing a pick is not something we'll guess at in print. What you can watch is whether the spread numbers start bunching the way the total and moneyline numbers already did. Fewer models sitting on old run dates. The outliers, the 12.916 and the Preseason Prior model at 5.7004, dropping out or coming in toward the −3.5-to-−4.5 cluster where 34 of the 36 already sit.
And a spread pick actually landing on the board. Until that happens, the spread here is a watch and not a bet.
Sources
We checked New England at Seattle against our own spread models on the Week 1 board, with every run time printed above. What that is worth is a watch on the spread, and nothing more, until those numbers narrow the way our total and moneyline sets already have. Our live picks on this game sit on /picks/nfl, and the three books sit on /odds. If you want the mechanics, we wrote up how the spread models are built and how we handle a stale line.
Those pages move when the market does. The numbers above don't.
