What matters
Seventeen games a season means every sample is small. The most repeatable inputs are opportunity — snaps, routes, targets, carries — and the team context that creates it: expected game script, pace and the quarterback. Efficiency (yards per target, touchdowns per touch) is far noisier week to week.
Important data
| Input | Why it matters |
|---|---|
| Snaps, routes, targets, carries | Opportunity — the most stable inputs |
| Game script (spread, total) | Trailing teams pass more; leading teams run more |
| Availability | Injuries redistribute usage across a roster |
| Weather | Wind in particular affects passing |
| Market prices | For comparison after a model scores, not as an input |
Common metrics
| Metric | What it measures | How to read it |
|---|---|---|
| Target share | Targets / team pass attempts | Receiving opportunity |
| Air yards | Depth of targets | Role and ceiling |
| Snap share | Share of offensive snaps | Role stability |
| Red-zone opportunity | Touches near the goal line | Main driver of touchdown chances |
| Down and distance | The live situation | Core of game state |
Model concepts
Yardage is continuous and skewed, so normal or skewed distributions around a projection fit better than Poisson. Touchdowns are rare counts with heavy variance. NFL model products on the network run in a tracking lane until they pass a pre-set sample and week gate, and tracking records stay separate from official ones.
Live intelligence
The NFL PBEcast Game Center leads with score, clock, possession, down and distance, then box score, leaders and win probability. Late-game plays carry the most leverage.
PropBetEdge NFL tools
PBE Picks ↗
Locked pregame model decisions with publication gates.
NFLTouchdown Targets ↗
Anytime-touchdown probability research.
NFLPBEcast Game Center ↗
Live game center: score, clock, possession, down and distance, box score.
NFLTrack Record ↗
Official and tracking records, kept separate.
NFLUsage Research ↗
Snap, route and target usage.
NFLPlayer DNA ↗
Position-by-position player profiles.
Worked example
- A receiver with a true 35% chance to score each week, over a 17-game season.
- Expected touchdown games ≈ 0.35 × 17 = 6.0, but the binomial spread (±3.9 games at 95%) makes anywhere from about 2 to 10 perfectly normal.
- Two scores in one week tell you almost nothing about next week's probability.
Glossary
Related lessons
Touchdown Targets
What does a PBE TD Target tell me, and why are some games empty?
PBE Picks & track records
How do I know a published record is real, and what should I look at in it?
Context changes
A player is reported out. How much should that change a number — and when?
Pregame vs live
The pregame model said 60%. Why does the live number say 35% in the second quarter?
Distributions
A player is projected for 6.2 strikeouts and the line is 6.5. Is the over likely?