1 · Calibrated on 2024, graded on 2025 — no peeking.
The model is fit on 2024 notable-injury labels and evaluated against the unseen 2025 set with time-correct features. Across every (2025-injured · healthy) player pair, the model ranked the injured player higher 72% of the time. The live backtest reruns on the 2025 Results tab every visit.
2 · Injured players concentrate in the top half of the risk ranking.
78% of 2025 injured players ranked in the top half of the pre-season frailty distribution — the durable half of the league produced just ~1 in 5 missed-game injuries.
3 · Elevated-risk flags hit at ~1.7× the league base rate.
About 1 in 3 players flagged elevated-risk missed time in 2025, against a ~1-in-5 league-wide base rate. That's a real, reproducible edge — the model's strongest single signal.
4 · The riskiest 20% account for a third of all injuries.
The top 20% of the pre-season risk ranking captured 32% of 2025 missed-game injuries — roughly 1.6× what a random 20% slice would catch. That's the tier to hedge.
5 · Full honesty — the misses.
72% ranking accuracy still leaves 28% of (injured · healthy) pairs misranked, and no model can predict a purely random contact injury. We publish the real number, not a cherry-picked one.
6 · 20 measurable features per player, including 5 injury-history channels.
Workload, contact rate, sack exposure, negative-play rate, age curve, wear-and-tear, plus recency-weighted injury count, severity, games missed, structural and soft-tissue history. A player with two ACL tears scores differently than one with zero — before either takes a snap.
7 · Position-aware calibration across all eight groups.
QB, RB, WR, TE, OL, DL, LB, DB — a high-risk RB and a high-risk QB mean the same thing on this scale. Generic "questionable / doubtful / out" labels can't do that.
8 · Nobody else publishes this number.
Consensus rankers and DFS tools rank healthy players — no pre-season durability signal at all (effectively a coin flip). You can pay $500/year for projections that assume every player stays healthy, or check the one variable they can't see.
9 · The price comp.
Low-end aggregators (FantasyPros, Draft Sharks, RotoWire): $72–$130+/yr. Betting data (Action Network PRO, BettingPros): $100–$150/yr. Elite DFS (Establish The Run, RotoGrinders, The Solver): $300–$500+/season. INJSUR: $39/season or $99 lifetime. No upsells — every Pro feature ships free to subscribers.
10 · Everything is computed from public NFL data.
Rosters and stats flow from the official nflverse feed; scores are the model's raw output on a natural 0–100 scale — no display massaging, no curve-fitting to look pretty. If a player grades above 90, that's the number we stand behind.