How our NFL prop model works

Everything on Backed By Stats comes from real game data and a published method. Here is exactly how we find NFL player prop trends, how Our Rating works, how the Touchdown Model prices scorers, and how we grade ourselves — in plain language.

• The data
• How a trend qualifies
• Who we show
• Our Rating
• Matchups & weather
• Line value & odds
• Touchdown Model
• Track Record
• Limits

The data

Play-by-play, box scores, snap counts, rosters and schedules come from nflverse, the open NFL data project. Live injury designations come from Sleeper, kickoff-window weather from Open-Meteo, and Vegas lines from the official schedule feed. We never type in a stat by hand. Everything refreshes automatically many times a week, and hourly on Sundays.

How a trend qualifies

A trend is an alternate line a player has cleared in at least 80% of his games across last season and this season combined — for example "3+ receptions" or "40+ receiving yards". Yardage lines end in 0 or 5, the way sportsbooks post alternate lines. A player needs at least six games to qualify, and every hit rate shows its sample so small samples are obvious.

We cover the prop types sportsbooks post as alternate lines — passing yards, passing touchdowns, passing + rushing yards, receiving yards, receptions, rushing yards and rushing + receiving yards — so every trend on the board can be added to a bet slip. The Track Record grades exactly those.

One detail most sites miss: the public box-score feed drops games where a player was on the field but recorded zero stats. We rebuild those games from snap counts, so a game where a receiver played and caught nothing counts as the miss it was.

Who we show

Only players on an active roster, on their current team. Anyone ruled Out, Doubtful, on injured reserve or suspended is removed automatically; Questionable players are flagged. For quarterbacks we show only the projected starter; if he is ruled out, the QB expected to replace him. Games leave the board about two hours after kickoff.

Our Rating

Our Rating (0–100) is our estimate of the probability that this exact line hits. It starts from the player's own hit rate, pulled toward the league average according to sample size — a 100% trend over six games is treated far more cautiously than a 94% trend over eighteen. That adjustment is calibrated on our own graded results: historically, a 100% trend on fewer than ten games hit only about 55% of the time, while one on fifteen or more games hit about 91%.

The rating then adjusts for this week's opponent defense, the team's Vegas implied total and the player's home/away split, and subtracts a penalty for wind or rain on passing and receiving props. In the Test Any Line tool, the Adjusted Rating recalculates the same model at whatever line you choose.

Matchups & weather

Each opponent gets one defensive rank per category — run defense, pass defense, and pass defense against wide receivers, tight ends and running backs — based on yards allowed, with this season weighted about three times as heavily as last. #1 is the toughest defense, #32 the easiest. Weather is the forecast for the game window itself — kickoff through the final whistle — not the daily high, and domes and closed roofs are marked as controlled.

Line value & odds

Where shown, odds are the best available price across major regulated US sportsbooks — DraftKings, FanDuel, BetMGM, Caesars, Fanatics, theScore Bet, BetRivers and Hard Rock Bet. The Sportsbooks filter on each board narrows prices, line value and the props shown to just the books you use. Line value is our probability minus the probability implied by that price. Because the price includes the sportsbook's margin, the comparison is deliberately conservative. Prices refresh up to seven times on Sundays and move constantly, so always confirm the live price before acting.

Touchdown Model

The Touchdown Model estimates each player's chance to score at least one touchdown. It works in three steps:

1. Opportunity value. Every carry and target is valued by where it happened — a carry at the 2-yard line is worth far more than a target at midfield — using league-wide touchdown rates for each part of the field.
2. Share of the team's touchdowns. A player's recency-weighted share of his team's opportunity value, times the touchdowns the team is expected to score this game from its Vegas implied total.
3. Adjustments. Finishing skill (pulled strongly toward average, because touchdown luck is noisy), recent snap-share trend, and a blend with the player's touchdown history, then a final calibration.

We also tested opponent position defense, red-zone defense and each team's own red-zone trend. None of them improved accuracy beyond the Vegas total, which already prices the defense — so we left them out rather than add noise.

Graded the same way on 2,444 player-games this season it had never seen, the Touchdown Model ranks scorers better than a simple touchdown hit rate (AUC 0.746 vs 0.713) with more accurate probabilities (Brier 0.127 vs 0.132).

Track Record

We grade every trend we would have shown, week by week, using only the data available before that week — no hindsight. Hits and misses are published together on the Track Record page.

So far our surfaced trends have hit 1,908 of 2,452 (77.8%), graded point-in-time.

Limits

These are historical base rates and probability estimates, not predictions or guarantees. Football is noisy, samples are small early in a season, and injuries and game scripts change quickly. Backed By Stats is not betting advice. 21+ and only where legal. If gambling is a problem, call 1-800-GAMBLER.