PropSims › Methodology › How the simulations work

How the PropSims simulations work

PropSims simulates every NFL game of the week 10,000 times as a whole game, both offenses together, and reads each player's projection, floor, ceiling and prop probabilities off those simulated games.

Why simulate instead of projecting one number?

A single projection tells you the middle of the road and nothing about the risk. Football outcomes are lumpy: one long touchdown or one extra red-zone trip can double a player's score. Simulating a game thousands of times shows the whole spread of what can happen, so we can publish a median, a bad-day floor and a good-day ceiling, and answer questions such as "how often does this player clear 59.5 receiving yards?" directly.

Step 1: set the game environment

Each game starts from the betting market's spread and total, which set how many points each team is expected to score. Injury reports and inactive lists then update who plays and how big each player's role is. All games of the week are simulated, Thursday through Monday; games already played are simulated from their pre-game inputs.

Step 2: simulate the whole game, both teams at once

Each simulation draws how many plays each team runs, its pass/run mix, its efficiency and its touchdowns, for both offenses together. The two teams aren't simulated in isolation, so the two offenses' results are linked in every simulated game.

Step 3: hand the team volume to the players

Each player's targets, carries, yards and touchdowns come from his share of that simulated team volume. Usage shares, efficiency and how much they vary from week to week are fitted on past NFL play-by-play data from nflverse (CC-BY 4.0); the current fit uses seasons [2024, 2025] weeks 1-17 + 2026 weeks 1-3. Nobody hand-edits a player's projection. If a number looks wrong, we fix the model or its inputs, and the fix applies to everyone.

Step 4: read the results off the simulations

After 10,000 simulated games, every number on the site is a simple count or percentile of the results:

• Median: the middle outcome; half the simulated games finish above it. See floor and ceiling explained.
• Floor (p10): only 1 simulated game in 10 is worse. Ceiling (p85): only about 1 in 7 is better. Some tables also show p90.
• Prop probability: the share of simulations on each side of the line. If a player tops the line in 5,800 of 10,000 simulated games, the raw over probability is 58%.
• Head-to-head % on the Start/Sit assistant: the share of the same simulated games in which one player outscores another (ties count half).
• Fantasy points: each simulated stat line is scored with PPR, half-PPR and standard rules for the season-long rankings, and with FanDuel's and DraftKings' own rules for the DFS projections and optimizer. (Scoring formats explained.)

Why whole-game simulation matters: correlation

Because every player in a game comes from the same simulated game, real-world relationships carry through. When a quarterback has a big day, his receivers usually do too. When one offense scores, the other often has to answer. That correlation matters for DFS stacks, for same-game prop combinations and for start/sit decisions between teammates or opponents. A model that projects each player on his own can't capture it.

Keeping the ranges honest

Raw simulations tend to be too confident. We recalibrate FanDuel point ranges and yardage distributions on past weeks the model did not see, so the ranges come out closer to how wide reality is. They are still slightly too narrow; the current calibration numbers and backtest results, including the parts that don't flatter us, are on the methodology page.

Updates during the week

When the injury feeds we track show a player ruled out, downgraded or cleared before his game, we re-run that game's 10,000 simulations and post what changed on the news feed: who absorbs the work and by how many points. On game days we re-check injury and inactive reports about every 7 minutes and re-run only the affected games.

Every prop forecast is logged before kickoff

Prop forecasts are written to an append-only, hash-chained log before kickoff and graded after the game against the official box score. The track record shows every graded pick. We make no win-rate claims: so far no single-prop strategy has beaten the market in our backtests.

What isn't simulated yet

• Kickers, and full distributions for team defenses (D/ST rows show a mean only).
• Fumbles lost and 2-point conversions in the published distributions.

NBA projections come from a separate model (PropSims NBA v3) and live under /nba/; this page describes the NFL simulations.

Related: Full methodology & backtests · Glossary · FAQ · About PropSims. Model opinion, not advice. For entertainment and informational purposes only. 21+. 1-800-GAMBLER.