The Signal Labs NFL model is built on EPA/play (Expected Points Added), the single most predictive metric in football analytics. Every play in every game generates a measurable impact — and the market consistently underprices teams whose EPA profiles don't match their win-loss record. That's where we find edges.
We're combining EPA with DVOA (Defense-adjusted Value Over Average) from Football Outsiders, success rate data, and our own proprietary adjustments for rest, travel, weather, and divisional familiarity. The model targets spread and total markets — moneylines in the NFL are less efficient than MLB or NBA because every team plays only 17 games and the sample sizes are tiny.
| Metric | Weight | Why It Matters |
|---|---|---|
| EPA/Play (Offense) | 30% | Most predictive single metric for offensive quality |
| EPA/Play (Defense) | 25% | Defensive EPA correlates to scoring prevention better than yards allowed |
| Success Rate | 20% | Consistency measure — how often a team gains positive EPA on a play |
| DVOA | 15% | Schedule-adjusted performance — accounts for opponent strength |
| Situational Factors | 10% | Rest days, travel distance, weather, divisional familiarity, primetime |
The 2026 NFL season kicks off in September with a loaded slate. The offseason has already reshaped several contenders — major free agent signings, the draft, and coaching changes will all factor into our preseason model. We'll publish our initial power rankings and early-season spreads analysis the week before Week 1.
Our backtesting on 2023-2025 NFL data shows the model performs strongest in Weeks 5-17 when sample sizes stabilize and the market begins to overreact to early-season narratives. Early season (Weeks 1-4) is where the public overvalues preseason hype — and where we find the biggest fades.
September 2026: Full weekly coverage begins — spreads, totals, props, best bets
Weekly: Every game analyzed, 50-pass evaluation, only the strongest edges published