NFL 2026 Index — Methodology Specification

Version 1.1.0 · Effective 2026-08-12

Version 1.2.0 — announced 2026-08-25 · effective 2026-09-01

1. Index definition

The NFL 2026 Index is composed of 32 constituents. Each constituent is one club's Season Share. The index level is the sum of the listed prices of its 32 constituents.

Contract construction

Each Season Share is a bundle of exactly 100 component contracts, each redeeming $1.00. Thirty-four cover the regular season — two per game, both redeeming on a win, one on a tie, neither on a loss. Sixty-six cover the post-season: 2 for the playoff berth, then 4, 8, 12 and 40 up the bracket. The $100.00 ceiling is the arithmetic result of 100 contracts at $1.00, which requires a 17-0 champion.

League-wide this issues 3,200 contracts across 32 shares, of which exactly 700 redeem each season regardless of results — every regular-season game distributes exactly $2.00, and the post-season distributes $156.00.

2. Constituent selection

The constituents are all 32 National Football League clubs. The set is fixed for the season. There are no additions and no deletions once the season's index is published.

3. Redemption schedule

Each Season Share redeems, cumulatively, as its club records the following outcomes:

  • $2.00 per regular-season win (up to 17 regular-season games).
  • $2.00 for a playoff berth.
  • $4.00 for a Wild Card round win.
  • $8.00 for a Divisional round win.
  • $12.00 for a Conference Championship win.
  • $40.00 for a Super Bowl win.

Redemptions are cumulative. A first-round bye credits the Wild Card rung. A tie credits half a win ($1.00). The maximum redemption is capped at $100.00, which a share reaches only as a 17-0 champion; the cap is never stated without that condition.

4. The fixed aggregate

Total league redemption is a construction constant, shown as arithmetic:

272 × $2.00 + (14×$2 + 8×$4 + 4×$8 + 2×$12 + 1×$40) = $700.00

Total redemption equals total issuance every season, regardless of on-field results: 272 regular-season wins, 14 playoff berths, and the fixed bracket count always redeem to the same $700.00 aggregate.

5. Pricing inputs

Prices are derived from three published markets: the regular-season win-total market, the postseason (playoff-advancement) market, and the Super Bowl futures market. The book of record is DraftKings. The publication date of each input is recorded with every board. Books are never mixed within a single market across the price history.

Naming a single book is necessary but not sufficient. A source may carry that book's name and still report a figure that is stale, or an alternate rung of a market rather than its main posted line. From version 1.1.0, inputs are taken in the following order of precedence:

  1. A page owned by the book, carrying an explicit publication date and an explicit update date.
  2. A third-party table showing an update timestamp within three days of the board's as_of date, where the book's column is identified structurally rather than by position.
  3. The preceding board's value, carried forward and listed in the board's carried_forward field.

Undated sources are never permitted to originate a value at any tier. They may corroborate a value that already agrees. Where two sources attributed to the book quote win totals a full win apart, they are quoting different rungs of the same market rather than disagreeing: the quote nearest even money is taken as the main posted line, the fresher source is preferred, and both values and both dates are recorded.

Where a published table's own date is older than the board'sas_of date, that is disclosed with the board. A dated table is a fact about the date it carries, not about today.

6. Calculation

Each published American-format price quote converts to an implied probability by o < 0 ? -o/(-o+100) : 100/(o+100), where o is the quoted American-format number.

  • Win-total probabilities are scaled so expected wins across the league sum to exactly 272.
  • Berth probabilities are devigged to sum to exactly 14.
  • Title probabilities are devigged to sum to exactly 1.
  • The interior rungs are geometrically interpolated, per team, between P(berth) and P(title) at fixed exponents — f = 0.25 for the Wild Card rung, 0.50 for Divisional, 0.75 for Conference — as p_rung = P(berth)^(1 − f) × P(title)^f. The interpolated values are then iteratively rescaled to sum to 8, 4, and 2 respectively, subject to per-team probabilities that are monotonically non-increasing across successive rounds. The exponents are part of this specification and change only with a version bump.

Fair value = $2.00 × expected wins + Σ(rung probability × rung redemption value). The listed price is the fair value. Major League Markets applies no markup to the published index. A listing venue sets its own commercial terms on top of it.

7. Invariants and rounding tolerance

The fair-value board and the listed board are the same board: $700.00 by construction. That is the same figure the index redeems in aggregate every season, irrespective of results — 272 regular-season wins at $2.00 plus the bracket ladder. Total issuance equals total redemption, and Major League Markets holds no directional position in any outcome.

Prices are published to the cent, so the sum of 32 rounded constituent prices carries rounding of up to ±$0.05 against the exact construction. Published boards have summed between $699.96 and $700.00. Individual prices are never adjusted to make a board sum exactly — doing so would misstate a constituent to flatter an aggregate. The invariants are asserted at ±$0.05 in the build, and a build that fails any of them does not ship.

The four asserted invariants:

  • Fair board $700.00 (±$0.05).
  • Listed board equal to the fair board, and to $700.00 (±$0.05).
  • No listed price above $100.00.
  • All 32 tickers resolve.

8. Publication

The index is published as a machine-readable feed at /api/v1/indexes/nfl-2026, with constituents and history endpoints. The feed is CORS-open and statically generated at build time. The feed's own totals are asserted by the test:api gate at build time: a build whose feed does not tie to the board does not deploy.

9. Repricing cadence

Daily through training camp and the preseason. In-season cadence from Week 1.

10. Edge cases (Proposed, v1.0.0)

The treatments in this section are proposed for version 1.0.0.

Cancelled or forfeited game

If the league records an official outcome, that outcome governs. If no official result is recorded, the game credits half a win ($1.00) to each club — identical to the tie rule. The reason: the $700.00 aggregate depends on exactly 272 wins existing, and this treatment preserves it exactly.

Season structurally shortened or lengthened

Never applied mid-season. Any structural change to the number of games takes effect in the next season's index, with the ladder re-fit to hold the $100.00 cap.

Relocation or rename mid-season

The constituent is the franchise, not the name. The ticker and display name may update; positions and accrued redemption are unaffected.

Schedule change altering the 272-win total

Within a season, absorbed by the tie rule. A structural change is reflected in the next season's index.

A club eliminated with regular-season games remaining

Nothing special happens, and this is intended. The share continues to accrue $2.00 per remaining regular-season win; its bracket rungs are dead. A contract that resolves once is worthless the moment elimination occurs; a Season Share on an eliminated club still holds live win contracts and retains value.

11. Governance

Changes to this methodology are made by Major League Markets through a versioned revision to this specification. The methodology is locked for the duration of a season once the season begins. Any change to the pricing inputs, the calculation, the redemption schedule, or the settlement source bumps methodology_version, and the feed's methodology_version field moves with it.

12. Settlement source

Official National Football League results as published by the league. Where an official result is subsequently corrected, the corrected result governs.

A listing venue may designate its own settlement source of record for contracts it lists. This methodology specifies the settlement events; the venue specifies settlement.

Change log

  • v1.2.0 — announced 2026-08-25, effective 2026-09-01. From the effective date, daily index values are produced by Major League Markets' own repricing engine rather than derived from the three published futures markets. The engine estimates each club's expected wins and rung probabilities by season simulation (N = 4,000,000 paths, date-seeded and deterministic: an unchanged-input day reproduces exactly), anchored to published game-level price quotes, with official league results and schedule as inputs once the season begins. The redemption schedule, the $700.00 issuance-equals-redemption identity, the ±$0.05 tolerance, the freeze rule, and the no-restatement rule are all unchanged. Shadow-run divergence evidence accumulated ahead of this change is retained. The three-market derivation described in v1.1.0 remains the documented fallback if the engine cannot produce a valid board.
  • 2026-08-24 — Publication gap 2026-08-12 → 2026-08-23, disclosed. Daily pricing runs continued throughout this period and priced correctly from the published inputs, but the generator's runbook had not been amended for the 2026-08-12 removal of the issuance spread, so each day's board was emitted with the withdrawn 5% markup and was withheld rather than published. The deploy gates, which require the listed board to tie to $700.00, stand between any such board and the published series. The runbook was corrected on 2026-08-24 and publication resumed the same day with the 2026-08-24 board. No published value has been restated, and the withheld boards will not be backfilled.
  • 1.1.0 — 2026-08-12. Four changes, disclosed together.

    The issuance spread is removed. Through version 1.0.2 the listed price was the fair value multiplied by 1.05, and the listed board was $735.00 against a $700.00 fair board. That 5% markup is withdrawn. The listed price is now the fair value and the listed board is $700.00 — the figure the index redeems. Spread, fees and any other commercial terms belong to the listing venue, not to the index. Every constituent's published price falls by approximately 4.77% as a result. This is a change to the published price definition, not to the instrument: the redemption schedule, the ladder and the $100.00 cap are unchanged.

    The published series is restated. Removing a markup applied on top of the fair value would otherwise render, on 2026-08-12, as every constituent losing 4.77% of its value on a day when no input market moved. Rather than publish that, the entire series from the 2026-02-19 open forward has been restated to fair value. This is the first restatement of a published price by Major League Markets. The pre-restatement series is preserved in full and published alongside the current one, so the two can be compared directly and every difference attributed. For the five nodes dated 2026-08-03 onward, the restated price is the fair value recorded at publication, used verbatim. For the five earlier nodes no per-constituent fair value was retained, and the restated price is derived as the published price divided by 1.05 and rounded to the cent. Because the published price and the fair value were each independently rounded to the cent, a derived value may differ by one cent from the fair value originally computed. The derived nodes are identified as such in the published data. Every restated price is accurate to ±$0.01 per constituent and ±$0.05 at the board level, and every percentage change in the series is unaffected. From this version, the fair value of every constituent is stored at publication, so no future restatement can be an approximation.

    The 1.0.2 disclosure below is corrected. That entry attributed day-over-day price artifacts to an iterative rescale with more than one fixed point. On measurement that attribution is wrong. Holding the algorithm fixed and varying the iteration count from 5 to 4,000, the column order across three permutations, and the placement of the monotonicity clamp produces no price change at all — the rescale converges to a single fixed point in fewer than five iterations. The actual cause is that this specification did not state the interpolation exponents, and each daily repricing chose its own; varying them within a plausible range moves a constituent by up to $0.31. The exponents are now pinned in section 6. The 1.0.2 entry also disclosed that the 2026-08-11 board was identical to the 2026-07-23 board across all 32 constituents. That was true of the pre-restatement series and is no longer true of the restated one, because the two nodes were restated by different routes. Both series are published; the coincidence is visible in the pre-restatement file. The 1.0.2 entry describes the deterministic repricing engine as taking effect at version 1.1.0. That is no longer accurate: version 1.1.0 is the spread removal and restatement described above, and the engine cutover has been renumbered to version 1.2.0, which will carry its own advance notice.

    A source-of-record defect is corrected. Cincinnati's regular-season win total was taken from a table dated 2026-07-23 that carried an alternate rung, 10.5, rather than the main posted line of 9.5 published by DraftKings' own outlet and updated 2026-08-10. Cincinnati was listed approximately $1.96 too high — 6.8% — on the boards of 2026-08-09 through 2026-08-11. Six playoff-berth inputs had also moved and were carried unchanged for 23 days. Correcting all seven inputs moves approximately $4.04 of absolute listed value across the board; every constituent other than Cincinnati, Detroit and Atlanta moves by $0.07 or less, and that movement is the win-total rescale redistributing rather than news. Consistent with the restatement above, these prices are corrected forward and not restated. This version introduces a ranked source hierarchy in section 5 to prevent recurrence: a stale source that reports a confident, unchanged, incorrect value was not something the previous date-lineage rule could detect.
  • 1.0.2 — 2026-08-11. Interior-rung advancement probabilities were recomputed from the input markets on every run, and that rescale is iterative with more than one fixed point, so two runs over identical inputs could settle on different interior values. Day-over-day price changes of up to $0.14 were published as a result — artifacts of recomputation rather than market moves, and above the ±$0.05 rounding tolerance stated in section 7. One consequence is disclosed here: the 2026-08-11 board is identical to the 2026-07-23 board across all 32 constituents, which is that same artifact and not a restatement — the 2026-07-23 values have never been altered. From 2026-08-12, a board whose three input markets are all unchanged republishes the preceding board's prices verbatim instead of recomputing them. No published price has been restated, and none will be. The deterministic repricing engine effective as v1.1.0 removes the underlying nondeterminism entirely.
  • 2026-08-03 — display erratum corrected: the 2026-08-03 node briefly rendered the advancement probabilities in the wrong units (fractions instead of percent). Corrected in place. Prices, fair values, and expected values were unaffected.
  • 1.0.1 — clarification of contract construction. No change to pricing inputs, calculation, redemption schedule, or settlement source.
  • 1.0.0 — initial publication, 2026-07-30.