The pickThe locked winner and its win probability. A priced game's number is built on the market's no-vig price; the model's own margin and total sit beside the market's below, as context.
backtest · not locked
Win chance
78%Pittsburgh SteelersStrong
No moneyline was priced into this call, so it is the model's own number.
Model marginThe model's own predicted margin beside the market's number. Context for reading the game: the model makes no spread pick.
PIT by 10.6market PIT by 14
Model totalThe model's own predicted total points beside the market's number. Context for reading the game: the model makes no over/under pick.
49.5 pointsmarket 44
Backtest, not locked: this game finished before the model ran on it; the calls were reconstructed from pre-game information.
Why the model calls itThe reasons behind the winner call, in plain sentences drawn from this page's numbers.
The model has Pittsburgh Steelers at 78% to win, with the market at PIT by 14.
Quarterback: the pick was locked with Garrett Gilbert starting for Dallas Cowboys, and the model prices that in.
No single matchup area stands out; the call rests on the teams' overall results this season, adjusted for who they played.
Form: DAL has the better last five games by 12.0 points a game (DAL +6.0, PIT -6.0).
Home field is with DAL, which the pick has to overcome.
ContextThe model makes it PIT by 10.6 against a market PIT by 14: 3.4 points of daylight in DAL's favour. The model's total is 49.5 against a line of 44: 5.5 points above it.The model makes no spread or over/under pick: neither has shown an edge over the market in our backtests.
Matchup context: who has the advantageWhich team has the advantage in each area, and how clearly. Context for reading the game, not model inputs.
Model PIT by 10.6 · market PIT by 14
ⓘMost of these areas favor DAL, but the model still picks PIT. The model rates each team on its full body of results, adjusted for who it played, plus the quarterback, rest, home field and weather. Single-area matchups like these are shown for context: over past seasons, when they point the other way the winner has been close to a coin flip, so they have not been a better guide than the ratings.
Passing game
–
—
not priced
No tendency data yet.
Rushing game
–
—
not priced
No tendency data yet.
Third down & red zone
–
—
not priced
No tendency data yet.
Big plays
–
—
not priced
No tendency data yet.
Pass protection & rush
–
—
not priced
No tendency data yet.
Ball security
–
—
not priced
No tendency data yet.
Form (last 5)
DAL
Clear advantage
DAL +6.0 per game over the last five, PIT −6.0.
Market
PIT
Clear advantage
Consensus has PIT by 14; the model says PIT by 10.6.
the market, for comparison
Where the lines sitWhere the market's number and the model's sit among every final margin and total in the league's history; the highlighted bars are the games that landed between the two.
3,893 NFL games since 2012
Marginmarket PIT by 14 · model PIT by 10.6
Mkt PIT 14
Model PIT 10.6
← PIT by 28tieDAL by 28 →
2.6% of NFL games land between the two numbers. The market is on 14, a number 5.1% of games land on exactly (a push). Most common margins: 3 (14.5%), 7 (8.6%), 6 (6.7%), 14 (5.1%).
Totalmarket 44 · model 49.5
Mkt 44
Model 49.5
20 pts4672 pts
13.1% of NFL games finish between 44 and 49.5. The league’s median total since 2012 is 45.
The model alone was backtested walk-forward (each game priced only from what came before it). Since the winner call started following the market, the calibration that counts is the record of locked picks on the results page. Margin and total are context, not picks. Model reads, not personalized advice. For information only, not betting advice. 21+ only. Gambling problem? Call or text 1-800-GAMBLER.
