Every score projection built from drive-level play-by-play data: field position curves, two-round opponent adjustment, and pace-based possession modeling. Priced before Monday's lines open.
Each component addresses a distinct source of predictive information that the betting market systematically underweights.
Drive expected points are not linear. The value of starting at the opponent's 30 is disproportionately greater than starting at the 35. An exponential decay curve, calibrated on four seasons of nflfastR play-by-play, converts starting field position into precise expected-point baselines for every drive.
Raw efficiency numbers are contaminated by schedule strength. A team that looks elite after beating weak defenses will be overrated by any system that ignores context. Two sequential rounds of opponent normalization strip the schedule signal from the efficiency signal, leaving clean team quality estimates.
Not all games have the same number of scoring opportunities. Average drive duration in seconds determines how many possessions each team expects. Faster-paced matchups produce more variance in totals. The model adjusts projected scores and total confidence explicitly for pace, not as an afterthought.
From raw play-by-play data to a betting edge. Five stages, no black boxes.
Starting field position is the most underappreciated input in NFL modeling. Every yard closer to the end zone is worth exponentially more, not linearly more. This curve, fit on historical play-by-play data, assigns expected-point values to each starting position on the field.
A drive from your own 20 carries roughly 1.3 expected points. The same drive from the opponent's 30 carries roughly 3.5 expected points. That 50-yard difference is worth more than twice as much.
Once field position baselines are established, every drive in the current season is evaluated on how many points it produced above or below expectation. This "points vs expected" metric separates elite offenses from ones that simply benefited from favorable starting position.
A team that scores 7 points starting from its own 20 outperforms expectation by far more than a team scoring the same 7 from the opponent's 10. The same result carries very different meaning depending on where it started. Context changes everything.
A team's raw efficiency tells you how well they played against who they faced, not how good they are. If the best games came against the league's worst defenses, the raw rating is inflated.
The model runs two sequential rounds of opponent-context removal. Round 1 strips first-order schedule bias. Round 2 removes residual bias left after Round 1, since those adjusted ratings still partially reflect opponent quality. After two rounds, ratings converge to a stable, schedule-independent estimate.
Both teams' adjusted efficiency ratings are blended with opponent defensive weakness to generate matchup-specific expected points per drive. Multiplied by estimated possessions (derived from both teams' pace tendencies), this yields projected final scores.
Win probability uses a calibrated non-linear formula to translate projected score ratios into clean win percentages, tuned to match empirical historical outcomes across multiple seasons.
Model spreads, totals, and win probabilities are compared against opening market lines. Plays are flagged only when the divergence clears a minimum threshold. Edges that fall below it are discarded regardless of perceived conviction.
Smaller edges are discarded regardless of perceived conviction. The goal is not volume — it's precision. A 60-play week at 52% CLV is worse than a 15-play week at 62%.
Every play logged against closing line value, not just win/loss. Lines moving toward our number after release is the real confirmation of edge.
NFL picks across all three major markets, released on a consistent schedule before lines move.
Picks for the full upcoming week's slate are released every Monday morning, ahead of typical line movement. Early access gives you the best opportunity to get on the opening number.
Every release includes model outputs across all three markets. Each play shows the model's projected line, the edge versus opening, and the market price at time of release.
The model operates on team-level drive efficiency and does not have visibility into individual player availability. Significant injuries, particularly at quarterback, are manually reviewed and accounted for before picks are finalized and released.
Subscribers are encouraged to cross-reference injury reports before placing any plays, especially for mid-week updates that occur after Monday's release.
All picks land directly in the Whizard subscriber Discord. Each post includes the model output card, edge breakdown, and unit recommendation.
Available individually or as part of the All Access bundle covering every sport and model we offer.