REDUCING UNCERTAINTY2026 seasonData · Nov 17, 3:01 PM ET
Every team’s path to the playoffs.
Playoffs. Division. The bye. Every seed. Compare the source consensus with RUPP’s own season simulation.
Source coverage varies by metric. Includes available source snapshots; see timestamps below. Use the checkbox to exclude it.
How many playoff games will each team play?
Expected games combines playoff seeding with market-derived matchup estimates. A bye advances a team without adding a game played. Round columns show the unconditional chance of playing in that round, including the Super Bowl.
Loading expected-games estimates…
Expected games: average playoff games across all scenarios, with zero for missing the playoffs. If qualified: average games conditional on making the playoffs. The four round probabilities sum to expected games. Uses the all-source seed blend, reconciled into valid brackets; the seed view’s source selector does not change these estimates.
Matchup estimates & simulation method
Market-derived estimates, not posted Vegas odds. Hypothetical matchups use the difference in market power ratings plus home field, converted into win probabilities using a fit to no-vig sportsbook moneylines. The Super Bowl is neutral. These estimates do not forecast future injuries or quarterback changes.
Each simulated conference has four division champions in seeds 1–4, three wild cards in seeds 5–7, and a bye for No. 1. Wild Card matchups are 2–7, 3–6, and 4–5. The top seed then hosts the lowest remaining seed. Original source averages are reconciled to the bracket constraints, then a fitted maximum-entropy model samples valid brackets. This estimates missing relationships between teams’ seeds; it does not reproduce a full remaining-season simulation.
Explore a team’s path
Potential opponents
Round columns show the unconditional chance of meeting that opponent. Win estimate averages the simulated venues when the teams meet; it is a model estimate, not a quoted moneyline.
| Opponent | Wild Card | Divisional | Conference title | Super Bowl | Win estimate if they meet |
|---|
RUPP model · Research release
One model. Every remaining game.
Our own model combines team performance, EPA, rest and ratings fitted from past sportsbook spreads. It simulates the remaining schedule and playoff bracket without averaging other sources’ playoff forecasts.
Loading RUPP forecasts…
Click any column to sort. All probabilities are unconditional unless marked “If qualified.” Expected games includes games already played in the postseason. A bye advances a team without adding a game. Simulation 0% or 100% does not establish official elimination or clinching.
Explore a team’s playoff path
Potential opponents
Each round shows the chance of meeting that opponent. Win estimates average the venues across simulated meetings. Completed games reflect known results.
| Opponent | Wild Card | Divisional | Conference title | Super Bowl | Win if they meet |
|---|
How accurate is it?
Models are selected on earlier seasons and evaluated on later seasons withheld from selection. Lower Brier score and log loss are better. Historical archives can contain later corrections; EPA models may have been fitted retrospectively. These tests are useful evidence, not a guarantee of future accuracy.
Model, data & limitations
Game results, schedule, EPA and archived sportsbook lines: nflverse / Lee Sharpe. NFLverse’s release data is offered under CC BY 4.0, subject to dataset-specific terms. No FTN data is used. Download forecast and provenance · Validation report.
DraftKings playoff draft guide
Compare the cost of a player in the screenshot draft market with the playoff opportunity of their team. The all-source blend supplies team forecasts, not player fantasy-point projections.
ADP last updated: November 17, 2026.
Less games at a similar ADP
Opportunity signals, not best/worst picks. Games edge is team expected games minus the median for other screenshot players at the same position within ±12 ADP slots, shown only with at least three peers. This ignores talent, usage, injuries, scoring rates and roster construction. Do not treat it as projected points, fair ADP or tournament expected value.
All-source blend games: uses the existing all-source seed-probability blend, reconciled into valid brackets and simulated with market-derived matchup estimates. It is not a direct average of sources’ expected-games forecasts. Games, games edge and round columns all use this all-source blend.
All round probabilities are unconditional. The four round columns sum to expected games, including the risk of missing the playoffs. A first-round bye contributes no Wild Card game. Strong later-round odds do not solve the need to field a viable first-round lineup in an advancement contest.
Use this alongside player projections and stack/roster planning, not in place of them. DraftKings playoff best-ball rules · Download transcribed ADP and provenance
Sources & how to read the chart
The default blend gives equal weight to all available sources for each metric: five statistical models plus Kalshi and Polymarket market prices. Choose “Models only” or “Markets only” to compare the two groups. Missing or rejected quotes never count as zero. Coverage varies by team and metric and appears above the chart; seed columns can therefore have a different denominator from the playoff and division columns.
Kalshi contributes playoff and division probabilities from its playoff-picture page, with its official public market-data API as a fallback if that page cannot be retrieved. The API uses two-sided Yes bid–ask midpoints for the same contracts; inactive or missing quotes are excluded. Polymarket contributes public two-sided Yes bid–ask midpoints from the 2026–27 postseason, division-champion and No. 1 seed events. A quote is excluded if the market is inactive or closed, either side is missing, the bid is zero / ask is 100%, or the spread exceeds 10 percentage points. Market coverage varies as quotes change. Kalshi has no bye forecast on the linked page; neither market source supplies seeds 2–7. Market prices are not normalized to sum to 100%, and do not guarantee calibrated probabilities.
The source cards show the last successful market retrieval. Market rows update with the scheduled refresh, just like the models. Download the initial market quote audit (November 16 snapshot).
Each source’s timestamp and refresh status appear above. Failed sources retain their previous snapshot and are flagged as older. The “Exclude older / failed snapshots” checkbox removes flagged inputs from the interactive average. The Datawrapper table uses the all-source blend and links here for freshness details. Records describe the latest successfully retrieved standings.
Playoffs = seeds 1–7; division = seeds 1–4; first-round bye = seed 1. Seed odds are unconditional. PlayoffStatus playoff chances come from the complement of its “No Playoffs” column; its division chance is the sum of seeds 1–4. ESPN FPI contributes playoff and division odds from its projections page; it does not supply bye or individual seed probabilities. ESPN’s Divisional probability means reaching that round, not earning a bye. The original three models use seed sums. Source rounding can affect totals.
Published bounds such as <1% and >99% are retained in individual model views. For averages and bars only, each bound contributes its midpoint (0.5% and 99.5%). The ≈ symbol identifies an average that includes bounded input; its tooltip gives the resulting input range. These are rounding bounds, not statistical confidence intervals. No missing odds are filled in. Simulation 0% and 100% do not establish official elimination or clinching.
Posting on Substack
For the custom interactive chart, add this page's link to your post. Use “Download chart PNG” for a newsletter image, and link that image to this page. For an interactive table directly inside Substack, import “Download embed data” into a Datawrapper table, enable search, sorting and heatmap columns, publish it, then paste its shareable link into your Substack editor. The CSV contains all 32 teams, the all-source averages including both prediction markets, and the source count for each metric. It refreshes automatically with the predictor. Custom HTML cannot be pasted directly into Substack.
The playoff picture, by division
Choose a division or conference. Sort any column within each division. These tables follow the model and snapshot settings above.