Sport-Specific Applications and Limitations
Negative Binomial isn't just for NFL touchdowns. Any count-based prop with overdispersion is a candidate. Let's explore where to look across major sports—and equally important, when not to use this distribution.
NFL: Touchdown Props
Best Candidates for Negative Binomial
Backup Running Backs
- Players who get goal-line carries but inconsistent overall volume
- Their TDs depend heavily on whether the team reaches the red zone and game script
Touchdown-Dependent Receivers (WR2/WR3)
- Players who might see 4-6 targets but have high TD rates when targeted
- Boom when they score, bust when they don't
Red Zone Specialists
- Tight ends or receivers who see concentrated targets inside the 20-yard line
- Limited work between the 20s creates all-or-nothing outcomes
Why NFL TD Props Are Overdispersed
| Factor | How It Creates Overdispersion |
|---|---|
| Game Script | Blowouts create feast-or-famine TD opportunities |
| Red Zone Volatility | Small sample of opportunities, high variance |
| Target Concentration | Some games: 8 targets; others: 3 targets |
| Defensive Game Plans | Opponents may scheme to take them away or ignore them |
Tip
Best NFL Edge: Backup RBs and TD-dependent WR3s with clear overdispersion (VMR > 1.5). Markets often price them using Poisson-like models, underestimating both 0 TD games and multi-TD explosions.
NBA: Three-Pointers and Assists
Best Candidates for Negative Binomial
Streaky Shooters
- Guards who get hot and keep firing
- When they're on, they might hit 6-8 threes; when they're cold, they go 1-for-8
Boom-or-Bust Role Players
- Specialists (e.g., catch-and-shoot wings) whose minutes and shot attempts vary wildly game-to-game
Backup Point Guards
- Players whose assist totals depend heavily on whether the starter sits or plays limited minutes
Why NBA Props Are Overdispersed
| Factor | How It Creates Overdispersion |
|---|---|
| Hot Hand | Shooters get hot and keep shooting, creating clustering |
| Usage Variance | Some games they're featured, others they're not |
| Blowouts | Garbage time can inflate or deflate stats unpredictably |
| Rotation Changes | Coach decisions can dramatically shift opportunity |
Note
NBA Application Note: Three-pointer made props for role players are particularly overdispersed. A shooter who attempts 3-8 threes per game with a 35% make rate will show significant game-to-game variance beyond what Poisson predicts.
MLB: Strikeouts and Home Runs
Best Candidates for Negative Binomial
Power Pitchers with Inconsistent Command
- Starters who might strike out 10+ or walk 4 and exit early
- Their K totals swing based on whether they have their best stuff
Three-True-Outcomes Hitters
- Batters who walk, strike out, or homer—rarely putting the ball in play
- Their HR totals are boom-or-bust by nature
Why MLB Props Are Overdispersed
| Factor | How It Creates Overdispersion |
|---|---|
| Pitch Count Variance | Pitcher might go 5 innings or 7 innings |
| Matchup Extremes | Dominant vs. weak lineup or struggling vs. good lineup |
| Weather/Park Factors | Wind and ballpark dimensions create day-to-day variance |
| Opponent Approach | Some lineups are K-prone, others make contact |
Warning
MLB Caveat: Strikeout props often show higher overdispersion than home run props because Ks depend more on pitcher "stuff" on a given day, while HR props depend on both the hitter's approach and factors like wind and park dimensions.
NHL: Goals and Points
Best Candidates for Negative Binomial
Elite Goal Scorers
- Top-line forwards who shoot frequently
- Goals are rare events, creating natural overdispersion even for stars
Power Play Specialists
- Players whose production is heavily PP-dependent
- If the team gets 3 PP opportunities, they might score 2 points; if they get 0, they're shut out
Why NHL Props Are Overdispersed
| Factor | How It Creates Overdispersion |
|---|---|
| Low Scoring | Small sample (2-4 goals/game) creates high variance |
| Power Play Clustering | Goals and assists come in bunches on PP |
| Goalie Performance | Opponent's goalie having a great/terrible night affects outcomes |
| Line Combinations | Coach changes can dramatically affect ice time |
When NOT to Use Negative Binomial
Like any statistical tool, Negative Binomial has limitations. Here's when to be cautious or avoid it entirely:
Limitation 1: Requires Sufficient Data
Warning
Estimating r accurately requires at least 15-20 games of data. For rookies or players returning from injury, stick with Poisson or empirical distributions (just use historical frequencies). With small samples, the variance estimate is too noisy to trust.
Limitation 2: Assumes Overdispersion Is Stable
Negative Binomial assumes the player's boom-or-bust nature is consistent over time. Role changes or injuries can change this.
Always check if recent VMR matches historical VMR:
| Scenario | Impact on VMR |
|---|---|
| WR3 → WR1 promotion | VMR likely decreases (more consistent targets) |
| Starter returns from injury | Backup's VMR becomes irrelevant |
| New offensive coordinator | Historical patterns may not apply |
| Trade to new team | Reset your data entirely |
Limitation 3: More Complex Than Poisson
Negative Binomial requires estimating two parameters (μ and r) instead of one. For marginal overdispersion (VMR = 1.1-1.3), the added complexity may not be worth it.
Rule of thumb: Stick with Poisson unless VMR is clearly above 1.3.
Limitation 4: Books Are Catching On
Note
As of 2024-2025, sharp books (Pinnacle, Circa) are increasingly using Negative Binomial or similar overdispersed models for boom-or-bust players. Your edge is shrinking.
However:
- Soft books (DraftKings, FanDuel) still often use Poisson-like pricing
- Edges exist in matchup adjustments (adjusting μ for specific games)
- Extreme outcomes (3+ TDs) are still often underpriced
Quick Reference: Distribution Selection by Sport
| Sport | Prop Type | VMR Typical | Model Recommendation |
|---|---|---|---|
| NFL | TD (RB1) | 0.8-1.2 | Poisson |
| NFL | TD (backup RB) | 1.3-2.0 | Negative Binomial |
| NFL | TD (WR3) | 1.2-1.8 | Neg Binomial or Poisson |
| NBA | Points (star) | 0.9-1.1 | Normal (high volume) |
| NBA | 3PM (role player) | 1.3-2.0 | Negative Binomial |
| MLB | Strikeouts | 1.2-1.8 | Neg Binomial or Poisson |
| MLB | Home Runs | 1.5-2.5 | Negative Binomial |
| NHL | Goals | 1.4-2.2 | Negative Binomial |
📝 Exercise
Instructions
For each scenario below, determine whether Negative Binomial is appropriate and why.
A rookie WR has played 8 NFL games with TD counts: 0, 1, 0, 0, 2, 0, 0, 1. Should you use Negative Binomial?
A streaky NBA shooter just got traded to a new team with a completely different offensive system. What should you do with your Negative Binomial model?