Elo-Implied 2026 Playoff Odds: 20,000 Seasons from the Frozen Ratings

From Expected Wins to Actual Odds

This is the third rung of one ladder. The prediction ledger turns the site's frozen opening ratings into a win probability for every 2026 game and grades itself weekly. The win-totals page sums those probabilities into each team's expected wins — Seattle 12.2, Las Vegas 4.9. But expected wins dodge the question fans actually ask, because the playoffs don't admit teams for win totals; they admit four division winners and three wild cards per conference, and which division you're in changes what ten wins buys. So this page plays the whole 2026 season out 20,000 times and counts what happens.

The Machine, Stated Fully

Each simulated season draws every one of the 272 scheduled games as an independent coin weighted at the model's probability — the identical logistic-in-the-rating-gap arithmetic the ledger publishes, with its +48 Elo home edge waived for the schedule's eight neutral-site games. Records are tallied; each division's best record wins it; the three best non-winners per conference take the wild cards; the best division winner takes the 1 seed. Repeat 20,000 times from a fixed seed (2026, for reproducibility) and the frequencies are the odds.

One simplification is big enough to disclose in bold: ties are broken uniformly at random. The real NFL tiebreaker ladder starts with head-to-head and walks through division and common-games records; simulating it faithfully would change which specific team survives some knife-edge seasons. Averaged over 20,000 seasons the distortion is small — but a team whose playoff hopes rest on winning coin-flip tiebreakers should treat its number here as having a point or two of fuzz. Everything else inherits the win-totals page's honesty budget: the ratings are frozen at their opening values, know nothing about injuries or August news, and carry a ±2-win binomial band on any single season even if they're exactly right.

The Exhibit: Every Team's Odds, and Why Wins Aren't Fate

What the Division Does to a Team's Fate

The right-hand scatter is the page's real lesson: expected wins buy different amounts of October depending on the neighborhood. Three worked pairs:

• The Rams' tax. L.A. carries the sixth-best expected wins in football (10.6) and just a 26% division shot — the lowest division share of any 70%-plus playoff team — because Seattle shares the NFC West. Three-quarters of the Rams' playoff probability arrives through the wild-card door, the highest wild-card dependence of any contender. Good team, wrong address.
• Tampa Bay's discount. Green Bay projects to more wins than Tampa Bay (8.95 to 8.40), yet their playoff odds are identical (44.7% and 45.1%) — and the Bucs are more than twice as likely to win their division (37.9% to 17.0%). A winnable NFC South is worth about half a win of expected-wins deficit, converted at the steepest part of the curve.
• Philadelphia's compounding. The Eagles' 73.6% division share is the league's highest — higher than Seattle's — on two fewer expected wins, because the NFC East's next-best team projects to 7.3. Weak divisions concentrate odds on their one good team.

The curve's slope is the other number worth carrying: through the middle of the table, one expected win moves playoff odds by roughly 13 percentage points — which is why the schedule bends of a half-win, invisible in a record, are entirely visible in October math.

Reading the Odds Like the Ledger Will

A few numbers on this page are bets the model is making in public. Kansas City at 38% is the coldest: the frozen ratings carry the Chiefs' 2025 into 2026 with a third of it regressed away and see a coin-flip team in Denver's division. New England at 72% is the warmest surprise on the other side. The mean last-team-in wins 9.3 games in the AFC and 9.5 in the NFC — so ten wins is probably safe and nine is a sweat, exactly the historical shape — and the average 1 seed lands at 13.2. None of these are opinions to argue with in the comments; all of them are checks the ledger will cash weekly from September 10 onward, the same way it grades every game probability it publishes. When the season disagrees with this page, the disagreement gets measured, not explained away.

Method notes: 20,000 simulated seasons, fixed seed, from the frozen opening ratings and the published 272-game schedule (eight neutral-site games carry no home edge). Ties broken uniformly at random — a stated simplification of the NFL's head-to-head-first tiebreaker ladder. Odds are conditional on the ratings; the ratings know nothing after the 2025 season ended. Every number regenerates deterministically from explainer_src/make_playoff_odds_chart.py.