Methodology

How the odds work

Plain-language notes on every step, from team ratings to the last tiebreaker. Every number here is the one the simulator uses.

The short version

Team ratings

A rating is how many points better or worse a team is than an average team on a neutral field. A +3 team would be a 3-point favorite against a 0 team. Three sources go into it:

When major news breaks after the results and preseason numbers were set, those two parts are adjusted by hand. Example: Lamar Jackson's thumb surgery (Nov 1) lowered Baltimore's preseason part by the full gap between him and Joe Fagnano (4.6 pts) and its results part by half that, since he played the earlier games. The market part already reflects the news.

The "Ratings" switch on the Standings tab lets you use any one source on its own. Each team's three values are on the Data page.

From ratings to a single game

The expected margin of a game is the home team's rating minus the away team's, plus 1.5 points for home field. NFL results land within about 13 points of the expected margin in a typical game, so the win chance follows a bell curve with that spread.

home win chance = Φ( (home rating − away rating + 1.5) ÷ 13 )

For the next week's games, the model also blends in each game's win probability from the live game feed, treated like a betting line. A 3-point favorite wins about 59% of the time. A 7-point favorite wins about 70%. Players you mark out on the Rosters tab shift the expected margin for the games they miss. That margin is then blended with the play-by-play matchup sim (see below): 65% ratings, 35% matchup sim.

Simulating the season

Each of the 10,000 simulated seasons plays every remaining game you haven't picked. Your picks are fixed in every run. Before each run, every team's rating is nudged by a random amount, typically within 2.5 points, because no rating is exact. That keeps long shots alive and stops favorites from looking like locks.

Each run then seeds both conferences and plays the playoffs: the #1 seed gets a bye, the field is reseeded after each round, the higher seed hosts, and the Super Bowl is on a neutral field. The same random numbers are reused from one update to the next, so a change in odds after a pick reflects the pick, not luck.

Playoff seeding and tiebreakers

The four division winners take seeds 1 to 4 by record. The three best remaining teams take seeds 5 to 7. Ties are broken in the NFL's order:

• Within a division: head-to-head, division record, record in common games, conference record, strength of victory, strength of schedule, then a coin toss.
• For wild cards: teams from the same division are first narrowed with the division rules. Then head-to-head (for three or more teams, only if one swept or was swept by the others), conference record, common games (at least four), strength of victory, strength of schedule, coin toss.

The NFL's points-based steps are skipped, because most simulated games have no score.

The Playoffs tab shows the field as if the season ended today: only games that have been played or that you picked count. That matches the standings the league publishes. The Standings tab's odds look ahead to the end of the season, so a team can be out of today's field and still have good odds.

Play-by-play matchup sim

On top of the power ratings, every remaining game is also played out snap by snap with a separate simulator. It uses each offense's and defense's real rates, adjusted for the opponents they've faced: completion rate, yards per catch, sacks, interceptions, scrambles, yards per carry, big plays (15+ yard catches and 10+ yard runs), red-zone touchdown rate, pass/run tendency, pace, penalties and 4th-down aggressiveness. Special teams are team-specific too: field goals, punting, extra points, kickoff returns and coverage, and punt returns. On top of that it uses the starting quarterbacks and their own numbers, injuries and trades, rest days, weather and home field. This is where matchups come in: a strong run offense facing a weak run defense gains more yards per carry in this sim than against a strong one.

Each regular-season game is simulated 3,000 times (and every possible playoff pairing 1,200 times, home and away), and its average margin counts for 25% of that game's expected margin, with the power-rating model making up the other 75%. Each team's offense and defense are also nudged for lineup changes: today's depth chart (injuries and trades included) is compared with the non-QB starters who actually played in the games its stats come from, using each player's point value. Chicago's defense gets credit for Jeffery Simmons, for example, and Detroit's loses some for Jack Campbell. It applies everywhere a game is decided: the season odds, playoff games, the 🎲 buttons, Sim week, Sim division, Sim season, Chalk, and re-simulating games already played. Turn it off with the Matchup sim switch.

On the Schedule tab, each week's games show the sim's projected score, win chance, fair spread and total. Tap a game to run 10,000 fresh sims. When betting lines are available, games where the sim and the line differ by 1.5 to 6 points are marked as data leans, and bigger gaps are flagged "check news." These are strictly based on data and are not financial or betting advice.

How much to trust it. The simulator was tested blind on 329 games (2025 Weeks 3–18 and 2026 Weeks 3–8). For each week, its stats were rebuilt from only the plays before that week. On its own it was a little less accurate than the closing Vegas line (Brier score 0.222 vs. 0.218; better than Vegas in 2026, behind in 2025). Mixed in at about 25% it made the line slightly better, so it counts for 25% here. Its probabilities are reasonably calibrated: games it called 73% on average were won 70% of the time. Against the spread, on games where it differed from the line by 1.5 to 6 points, it went 103–82 (55.7%). That is above the 52.4% needed at −110 but still within luck for 185 bets, about a 1-in-14 chance for a coin flip. Totals showed no edge (93–103). Treat leans as a reason to look closer, not proof of an edge.

The Game sim tab opens the full matchup simulator on its own. Pick any two teams (or a Week 9 game with its betting line and starting quarterbacks filled in), run thousands of sims, and see win chances, projected score, the spread and total it implies, its most common final scores, and how often each side covers or the game goes over. It also shows the full 329-game blind backtest, calibration and against-the-spread record. The offense and defense injury sliders start pre-filled with each team's lineup change.

Power rankings

Power rankings show how good each team is right now, not where it sits in the standings. Five inputs, each measured in points better or worse than an average team on a neutral field:

The score is roughly a point spread: a +4 team would be favored by about 4 over a 0 team on a neutral field. Last week's rankings use the same recipe with only the games played through the week before, so the arrows show real movement. Rankings update with each weekly data refresh and don't change with your picks.

Best bets and sportsbook odds

The odds on this site are not live. They are a snapshot pulled from The Odds API for DraftKings, FanDuel, BetMGM and BetRivers whenever the site is refreshed, and the Bets tab shows exactly when and how long ago. Lines move all day, so always check your sportsbook's current price before betting. If the snapshot is more than two days old, the tab turns into a warning.

For each game, every book's moneyline, spread and total is compared with the play-by-play matchup sim (10,000 sims with the current starters and lineups). The chance used is 25% sim and 75% the books' own no-vig consensus, the mix that tested best, so the sim nudges the market rather than replacing it. A bet is listed only if it comes out ahead at the best available price, the sim on its own also clears the break-even chance, and the sim and the books are within 6 points. Bigger gaps usually mean news the stats don't know about. Totals are shown but never picked, because the backtest found no edge on over/unders. Expected profit is per $100 at the listed price. Strictly based on data, not financial or betting advice.

Clinched and eliminated

A team is marked OUT only when no combination of remaining results can get it in. The site checks this by assuming the team wins every remaining game, then testing whether the other teams' remaining games against each other could leave enough of them below that total. Those games have to have a winner, so a 5-12 team can be out even when only a few teams already have more wins. It is marked IN ✓ when no more than six other teams can still catch it, or DIV ✓ when no division rival can. Ties are never assumed to break a team's way, so these labels are conservative and only appear when certain.

The elimination number (E3, E2…) counts how many more losses by the team, plus wins by the team currently 7th, would end its wild-card chase. It works like baseball's elimination number. The odds never show a flat 0% or 100% unless it's mathematically certain. A long shot that never came up in 10,000 simulated seasons shows as <1%.

The scenarios box below the standings lets you pick one team. It shows its magic numbers for clinching a playoff spot, the division and the #1 seed, and its elimination number. It also checks every possible combination of the next week's open games and lists exactly which results would clinch or eliminate that team that week. It counts your picks, so you can set up a future week and see what is at stake.

A team's magic number is how many more of its own wins, plus losses by the specific team chasing it, lock up the goal. It is worked out against each rival separately, because one rival can still catch a team even when another can't. For the division and the #1 seed, the largest number counts, since the team must finish ahead of every rival. For a playoff spot, it's the number at which no more than six other teams can still catch it.

Tap a team in the standings to open its panel. Win out and Lose out pick all of that team's remaining games, and Clear removes just those picks. The Clinch buttons (Playoff spot, Division, #1 seed) pick only that team's games, with the fewest wins that still guarantee the goal no matter how every other game goes, using the real tiebreakers. The losses are given to its toughest remaining games. If even winning out can't guarantee it, the button says so.

Players and injuries

Every player on each depth chart has a value in rating points. It starts from what his position is worth to a team's margin, then scales by how well he has played.

The quality score comes from per-game production over four seasons, weighted 40% for 2026, 30% for 2025, 20% for 2024 and 10% for 2023, so a star who has missed this season still rates as a star. Players with few games are pulled toward a typical starter or backup. Stats follow the player, not the team, so a free agent keeps his history. Linebackers who mostly rush the passer are counted as edge rushers. Kickers are rated on field-goal percentage. Offensive linemen use the uploaded 2026 line grades, and linemen without a named grade get their team's average.

Each player also gets a grade from 1 to 99: the share of players at his position, starters and backups, with a lower quality score. A grade of 90 means he rates higher than about 90% of players at his spot. Grades are for comparing players. The odds use the point values, because a backup quarterback can grade low and still be worth more points than a good receiver, simply because quarterback play moves the spread so much.

A team's lineup strength is the sum of the first healthy player at each spot. Marking a starter out means his backup takes the spot, and the team's rating drops by the difference. You choose how many games he misses, and only those games change.

Injury and starter news. Statuses (Questionable, Out, IR, PUP, Suspended) come from the depth charts and are updated from the week's injury reports and team news for notable players. A 📰 line under a player on the Rosters tab says what happened and when, and teams with news show a 📰 count. When a starter is out, the next healthy player takes his spot in the lineup and in the matchup sim (for example Joe Fagnano for Baltimore and Gardner Minshew for Arizona in Week 9). Statuses alone don't move a team's odds, because the betting market already prices known injuries; marking a player yourself does.

Simulated scores and drive logs

The winner of a simulated game (🎲, Sim week, Sim season, the bracket) comes from the full model's odds, so favorites win as often as they should. The score and drive log then come from the play-by-play matchup sim: it plays the game snap by snap with each team's own pass/run mix, pace, sack, interception and fumble rates, run game, kicker and punter, the current starting quarterback (a backup with little data plays at replacement level), and the lineup changes described above. It keeps the first simulated game whose winner matches. So a slow, run-heavy matchup produces fewer drives and lower scores, a leaky run defense gives up longer drives, and a turnover-prone quarterback throws more interceptions. Scores and totals land close to the sim's projection. Drives end in a touchdown, field goal (made or missed), punt, interception, fumble, turnover on downs, safety, return touchdown, or the end of a half. Fourth downs are team-specific: the sim starts from a league-wide table of how often teams go for it at each field position and distance, then scales it by how often that team has actually gone for it in the same spots this season (2025 counts about a third, and small samples are pulled toward average). The Raiders, Bears and Lions go for it more; the Seahawks, Broncos and Texans much less. The 4th-down play itself succeeds based on that team's offense against that defense. Backup quarterbacks: a team's stats were built mostly with its usual starter, so swapping in a backup's own passing numbers understated the drop (the rushing, protection and play calling built around the starter stay in the team's numbers). When the starter is not the team's usual QB, the difference between the two QBs is scaled up (exponent 2.5, tuned on the backtest); in the 75 test games with a different starter, this made the sim more accurate than the closing line (Brier 0.214 vs. 0.220). Fumbles by team were tested and dropped too: they made the backtest slightly worse, which fits the long-standing finding that fumbles are mostly luck from week to week, so every team uses the league rate. A separate short-yardage conversion rate by team was tested and dropped, because it made the 329-game backtest slightly worse. Overtime follows the rule that both teams get a possession. Re-simulated past games use each team's season-long quarterback mix instead of today's starter. For games already played, the Schedule tab shows the real box score instead.

What the model doesn't know

• Weather, travel, rest days and coaching changes, except as far as betting odds already reflect them.
• Injury news after the last update (November 3, before the midweek practice reports) unless you mark it yourself. Lesser-known players' statuses come from the November 2 depth charts.
• Points-based tiebreakers, which are rarely needed.
• Two games are missing real box scores: Broncos at Chiefs (Week 2) and Eagles at Bears (Week 4). Their final scores are included.

Full source details, dates and known gaps are on the Data page.

Terms used on this site

Playoff odds

Share of simulated seasons in which a team finished in the top seven of its conference.

Projected wins

Average final win total across all simulated seasons.

Power rating

Points better or worse than an average team on a neutral field.

Chalk

Picking the favorite in every game. "Chalk" is old betting slang for the expected result.

Player grade

1 to 99 rank of a player among everyone at his position. Separate from his point value.

Seed

A team's playoff position, 1 to 7. Seeds 1 to 4 go to division winners and #1 gets a first-round bye.

Magic number

Wins by a team plus losses by its closest rival that clinch a goal. At 0 it's clinched.

Elimination number

Losses by a team plus wins by the team ahead of it that would knock it out. Shown as E3, E2 and so on.

Matchup sim

A play-by-play simulation of two specific teams, using their offense and defense rates, starting quarterbacks and lineup changes. 25% of each game's odds.

Data lean

A game where the matchup sim and the betting line differ by 1.5 to 6 points. Strictly based on data; not financial or betting advice.

Strength of victory

Combined record of the teams a club has beaten.

Strength of schedule

Combined record of every team a club has played.

What-if

A finished game whose result you changed. It shows with a dashed outline and ↩ restores the real result.