What you'll see on every game.
Every game card carries these five. What they mean:
Home ahead, away ahead, or level: three sides that sum to one. Ties are common enough in a quarter (about one in five) that we carry the level case explicitly rather than letting a normal curve invent a number for it. This is the first-quarter market our back-testing found signal on, and the one we lead with. Where a book quotes it two-way, a tied quarter pushes, so the figure beside the price is the chance we make it given somebody is ahead.
Whether that team scores at all in the first fifteen. The two numbers rarely match: each is that team’s own Q1 scoring rate against the other side’s Q1 defence, shrunk toward the league average. Context only: re-tested over four seasons of walk-forward games, our number on these markets did not clear the margin we treat as run-to-run noise, so neither carries a pick.
Whether the game's first possession ends in points. Until the toss we don't know who'll have the ball, so both cases get carried at even weight. Context only: back-testing found no edge on this market, so it carries no pick.
Points we expect between the two of them in the first quarter.
Whether any points land in the opening quarter, and the scoreless side that is the same number read backwards. We show it for orientation and publish no pick on it: over three seasons of walk-forward testing, our number on this market scored no better than simply quoting the league rate, so there is nothing here we can claim to know that the rate does not already tell you.
Full-game moneyline, spread and total run off the same engine. We don't do player props or halves.
Where the number comes from.
There's no simulation in here and no neural net. Everything starts from something a team has already done in the opening quarter, then gets adjusted for the opponent, the weather and the market. When a number looks wrong to you, you can work out which of those did it.
What we feed it
Per-team Q1 rates from nflverse play-by-play, schedule and scores from ESPN, a kickoff-time weather forecast, and the Pinnacle line.
- How often they score in Q1
- How often they let you score in Q1
- What happens on their opening drive
- How well they move the ball early
- Wind, rain and temperature at kickoff
- How many games we're going on
The opening quarter
Probabilities for the first 15:00. Which side is ahead when it ends, which side gets on the board, how the opening drive ends, and the expected Q1 total and margin.
- Chance each side leads after Q1 (our headline number)
- Points we expect in Q1
- Q1 margin
- Chance each team scores, as context
- Chance the opening drive scores, as context
- Chance anyone scores, as a base rate
Carried to the full game
That Q1 read feeds the full-game projection, so every market on the site traces back to the same engine.
- Full-game win probability
- Projected margin
- Projected total
- How far we are from the book
- Same badges as the Q1 markets
- Moneyline, spread, total only
- Start with what they've already done. How often a team has scored in the opening quarter, and how often they've let the other side score.
- Don't believe a hot start. A team three games into a season doesn't have a real rate yet, so we treat them as close to league-average and let them earn their way off it. Two good weeks isn't evidence.
- Match the offense against that particular defense. A strong Q1 offense against a strong Q1 defense comes out somewhere in the middle rather than at either extreme.
- A scoreless quarter means both sides came up empty. So each team gets its own number first, and then we ask how often neither of them gets there.
- The opening drive is half a coin toss. One team gets the ball and until the toss we don't know which, so we run it both ways at even weight.
- Turn the odds into points. A first-quarter score is a touchdown somewhat more often than a field goal, so a likely scorer is worth 6.7 points on average. Weather and efficiency push it around from there.
Where Pinnacle has already posted a line, we lean toward it rather than override it. They're sharper than we are. The number we publish is the distance between us and them, and it's worth reading it that way.
We take the book's cut out first.
Every book builds a margin into its odds, which means the price on your screen is never the true price. So before comparing anything we take that margin back out and work out what the book would be charging with no cut in it. That's what our number goes up against.
A card reading +5.2% means we put the real chance 5.2 points above the book's true price. It isn't 5.2% better odds; it's 5.2 points of disagreement. On a game card we print it once, on the side we lean, and only when it clears the Lean threshold below — under that the card says no bet rather than naming a side. The full market table on a game page works the other way round: it prints every market from one fixed side, so a number there can be negative, and it can be small.
Three badges, set by the size of the edge.
That's the whole rule. The further our number sits from the book's true price, the bigger the badge. Below the Lean threshold we publish nothing, which happens a lot. Most games don't have an edge in them.
The ones we're most confident in. The ledger stakes them 1u flat like everything else, so a Gold Zone earns its place by hit rate and price, not by sizing.
Flat-staked, unglamorous, and over a season this is where most of the return comes from.
A direction rather than a conviction. Fine as a parlay leg. Don't stake it like a Gold Zone.
An edge over 15 points gets no badge. When we're that far from Pinnacle we read it as something having gone wrong in our own model on that game rather than as value we've found. We got there the expensive way running the baseball version of this, where the huge disagreements with the sharp line lost money and the moderate ones made it. So the edge is a band with a ceiling as well as a floor, and the outliers get dropped along with the tiny ones.
Those three thresholds are a starting point, not a tuned result. We picked them from how the baseball model behaved and we're holding them where they are until there are a hundred-odd settled NFL picks sitting behind them, because a cutoff fitted to forty games mostly just fits the noise in those forty games. If we move them, we'll say so here and date it.
A pick enters the public record the moment it clears the Wager floor above: 4 points, the same number for every market and the same number the badge reads. What a card calls a Wager is what the ledger grades; a Lean is a note on a card, not a call, and is never graded. One ladder, not yet tuned, and it will be tuned in one place.
Rebuilt every five minutes. Every change logged.
Same model each time, just newer information. Every snapshot is versioned, and consecutive snapshots are diffed into the per-game audit log, so you can see what moved and when. Nothing here runs on a publishing calendar.
Slate rebuild
Model reruns on current inputs and Pinnacle is re-pulled and stripped of its margin. A line move lands when it lands.
The Q1 board prices up
Pinnacle opens its 1st-quarter moneyline, spread and total roughly a week before kickoff. Until then those markets carry no price to beat and read as model-only.
Lock price freezes
The moment a pick clears the Wager floor, the price it saw is written once and never rewritten. That number is what the ledger grades, whatever the line does afterwards.
Snapshots stop
Closing price stops being refreshed. Each pick has exactly one deadline — its own kickoff, not a weekly one.
Grading
A 1st-quarter call settles as soon as the quarter does, off the 1st-quarter box, usually within five minutes of the second quarter starting. Full-game markets settle off the final score once the game goes final. Nothing waits on a page view: the same pass grades the public ledger and every subscriber's own logged bet.
What FirstSnap™ will not do.
Re-grade after the fact
Picks lock at kickoff. No going back because the line moved or we meant the other side. Every snapshot we publish stays on the record.
Post bet slips
The picks go up here before kickoff with a timestamp on them, and they stay up afterwards whether they landed or not.
Sell you a "lock of the year"
One price, every week. Gold Zone calls come with any pass and there's no premium tier above it.
Promise you a hit rate
The track record is public, losing weeks included. Anyone quoting you a guaranteed percentage is selling something.
Terms we use on the cards.
How much a play moved a team's expected points. We use Q1-only splits, because season-long EPA washes out what's particular to the opening quarter.
Book odds add up to more than 100%, and the surplus is their margin. Strip it out and you get the price they'd offer with no cut, which is what we compare against.
Did you get a better price than the line closed at? Over enough bets it predicts profit better than your win rate does. We record it wherever we caught both the lock and the closing price.
Questions that come up.
Do you cover the full game, or just the first quarter?
Both. The Q1 engine is the specialty, and it is why most full-game models leave value on the table in the opening 15 minutes. Once FirstSnap™ has modeled the start cleanly, that signal feeds a full-game projection covering total, spread and moneyline.
Is this just a full-game model truncated to Q1?
The other way around. Full-game projections from generic models smooth over the opening 15:00, where pace is faster, defenses haven't tired, kickers haven't worn down and weather hasn't shifted. FirstSnap™ models Q1 directly with its own efficiency, pace and scripted-play priors, then carries that read forward into the full-game lines.
When are picks posted, and how often do they refresh?
Continuously. The slate is rebuilt every 5 minutes, so a line move is picked up as it happens rather than on a publishing calendar. A ★ pick keeps the price it first locked at, and that lock is what the ledger grades. Later refreshes move the live number, never the recorded one.
What's the difference between Gold Zone, Wager, and Lean?
Three badges, set by the size of the edge and nothing else. Gold Zone needs at least 7 points of edge, Wager at least 4, and Lean at least 1.5, which is a directional read without conviction sizing. The edge is a band rather than a floor: at 15 points or more FirstSnap™ publishes no badge at all, since sitting that far from Pinnacle reads as something wrong in the model on that game rather than value. Badging is entering the ledger: a pick is recorded the moment it clears the Wager floor, and a Lean never is.
Do you guarantee profit?
No, and anyone who does is selling you something else. The long-term track record is published in full, losing weeks included. The bet is that the edge over closing line compounds across a season, not that any single Sunday goes your way.
Which sportsbooks do you compare against?
One: Pinnacle, the sharpest line in the market, pulled every refresh and stripped of its margin to a fair price. Better to price against the best number than to average six soft ones. Pinnacle opens its 1st-quarter moneyline, spread and total roughly a week before kickoff, so those carry a real edge once they are up and read as model-only until then. The rest of the Q1 board (which side scores, the opening drive, whether anyone scores at all) is quoted by no book, so it stays model-only always and is labelled that way everywhere it appears.
How does the free pick work?
One game on each slate is open in full: the pick, the number, the edge, the model breakdown, all of it. It is the biggest edge on the board inside the band — the ceiling that drops outliers applies here first — whatever tier that lands in, and it moves as the slate rebuilds. No signup, no email gate, no card. Every other game is modeled the same way but needs a pass to view.
How do I cancel?
There is nothing to cancel. A pass is a one-time purchase, month or season, and it ends when its window does. Nothing auto-renews, no card is kept on file for a next cycle, and you are never asked to call anyone to stop it.
See it on this week's slate.
Month or season pass.
Every game on the slate, same engine behind all of them. One free pick is open on every slate with no signup. The rest opens when you subscribe.