Loading markets...
Loading markets...
Model receipt · fandaq-multi-season-audit-v2-candidate-20260913
All 3,800 matches are replayed in order. Six seasons choose a candidate rule; four later seasons remain unseen until the holdout. The active model and candidate appear together, including rank error, volatility and market-wide drift.
The 2021/22–2024/25 results did not choose the coefficients. Price-rank ties use club name, never final points.
openfootball/football.json · CC0 public domain dedicationCandidate decision
The active rule produced +35.2% average equal-weight growth across ten seasons. The candidate re-centres every settled round to ƒ1,000 of total quoted value, so owning one share of every club cannot create FanBucks from market drift. Relative winners can still rise and losers can still fall.
Activation gate
Existing replay rooms still use v1. Before v2 can price a live room, Fandaq needs independent economics review, migration rules for open positions, abuse testing, a published effective date and a rollback receipt. This page is evidence for that decision, not a silent switch.
Train / holdout receipt
Rank correlation compares price order with the final table. Mean absolute error is the average number of table places missed. Drift is the equal-weight return from holding one share of every club.
| Season | Split | Active rank corr. | Candidate rank corr. | Active error | Candidate error | Active drift | Candidate drift | Champion / candidate rank |
|---|---|---|---|---|---|---|---|---|
| 2015/16 | training | 0.992 | 0.998 | 0.40 | 0.10 | +29.6% | 0.0% | Leicester City / #1 |
| 2016/17 | training | 0.994 | 0.992 | 0.30 | 0.50 | +38.6% | 0.0% | Chelsea FC / #1 |
| 2017/18 | training | 0.988 | 0.992 | 0.70 | 0.30 | +35.7% | 0.0% | Manchester City / #1 |
| 2018/19 | training | 0.995 | 1.000 | 0.30 | 0.00 | +40.7% | 0.0% | Manchester City / #1 |
| 2019/20 | training | 0.992 | 0.997 | 0.40 | 0.20 | +33.5% | 0.0% | Liverpool FC / #1 |
| 2020/21 | training | 0.986 | 0.998 | 0.50 | 0.10 | +30.6% | 0.0% | Manchester City FC / #1 |
| 2021/22 | holdout | 0.986 | 1.000 | 0.60 | 0.00 | +37.8% | 0.0% | Manchester City FC / #1 |
| 2022/23 | holdout | 0.989 | 1.000 | 0.50 | 0.00 | +34.3% | 0.0% | Manchester City FC / #1 |
| 2023/24 | holdout | 1.000 | 0.998 | 0.00 | 0.10 | +37.9% | 0.0% | Manchester City FC / #1 |
| 2024/25 | holdout | 0.998 | 0.998 | 0.10 | 0.10 | +33.4% | 0.0% | Liverpool FC / #1 |
A win starts at +3.5%, a draw at 0%, and a loss at −2%. The round then re-centres all 20 quotes to the original ƒ1,000 market total. Prices remain between ƒ5 and ƒ200. Goal difference is 0% in this candidate.
The training search evaluated 100 disclosed combinations: wins 2–4%, losses −2% to −4%, draws fixed at 0%, and goal difference 0–0.6% per goal. The published candidate balanced rank agreement, error, return separation, drift and one-round movement.
Across the four holdout seasons, candidate rank correlation was 0.999, average rank error was 0.05, every champion ranked first by price, and passive drift was 0.0%.
2015/16 active-model champion
Finished first in the table and #1 by final active-model price.
Cleanest rank reads
Each finished within 0 market-rank places of the final table.
Deepest drawdown
58.6% from its prior replay peak. Drawdown is path risk, even when the final return recovers.
2015/16 active model · every club
This is the exact v1 quote used by the playable replay. Price ties break by club name so final points never decide price order.
| Club | Table | Price rank | Rank error | Points / GD | Final quote | Return | Max drawdown | Up / down rounds |
|---|---|---|---|---|---|---|---|---|
| Leicester City | #1 | #1 | 0 | 81 / +32 | ƒ134.30 | +168.6% | 4.2% | 35 / 3 |
| Arsenal FC | #2 | #3 | 1 | 71 / +29 | ƒ103.75 | +107.5% | 6.7% | 31 / 7 |
| Tottenham Hotspur | #3 | #2 | 1 | 70 / +34 | ƒ105.95 | +111.9% | 7.5% | 32 / 6 |
| Manchester City | #4 | #4 | 0 | 66 / +30 | ƒ88.17 | +76.3% | 10.5% | 28 / 10 |
| Manchester United | #5 | #5 | 0 | 66 / +14 | ƒ85.35 | +70.7% | 10.2% | 28 / 10 |
| Southampton FC | #6 | #7 | 1 | 63 / +18 | ƒ80.47 | +60.9% | 14.8% | 27 / 11 |
| West Ham United | #7 | #6 | 1 | 62 / +14 | ƒ82.70 | +65.4% | 7.5% | 30 / 8 |
| Liverpool FC | #8 | #8 | 0 | 60 / +13 | ƒ75.45 | +50.9% | 7.8% | 28 / 10 |
| Stoke City | #9 | #11 | 2 | 51 / -14 | ƒ54.24 | +8.5% | 14.7% | 23 / 15 |
| Chelsea FC | #10 | #9 | 1 | 50 / +6 | ƒ60.13 | +20.3% | 13.3% | 26 / 12 |
| Everton FC | #11 | #10 | 1 | 47 / +4 | ƒ55.79 | +11.6% | 16.6% | 25 / 13 |
| Swansea City | #12 | #12 | 0 | 47 / -10 | ƒ51.36 | +2.7% | 22.9% | 23 / 15 |
| Watford FC | #13 | #13 | 0 | 45 / -10 | ƒ47.81 | -4.4% | 23.4% | 21 / 17 |
| West Bromwich Albion | #14 | #14 | 0 | 43 / -14 | ƒ47.05 | -5.9% | 16.1% | 23 / 15 |
| Crystal Palace | #15 | #15 | 0 | 42 / -12 | ƒ44.11 | -11.8% | 30.3% | 20 / 18 |
| AFC Bournemouth | #16 | #16 | 0 | 42 / -22 | ƒ42.80 | -14.4% | 18.3% | 20 / 18 |
| Sunderland AFC | #17 | #17 | 0 | 39 / -14 | ƒ42.43 | -15.1% | 28.8% | 21 / 17 |
| Newcastle United | #18 | #18 | 0 | 37 / -21 | ƒ38.57 | -22.9% | 34.2% | 19 / 19 |
| Norwich City | #19 | #19 | 0 | 34 / -28 | ƒ33.58 | -32.8% | 34.6% | 16 / 22 |
| Aston Villa | #20 | #20 | 0 | 17 / -49 | ƒ21.59 | -56.8% | 58.6% | 11 / 27 |
A win applies +4%, a draw +0.5%, and a loss −3%. Goal difference adds or subtracts 0.4% per goal, capped at three goals. Prices stay between ƒ5 and ƒ200 and round to cents after each match.
The same deterministic process now measures 3,800 matches, multiple champions and relegation groups, while a dated split reveals whether the candidate survives later seasons.
EPL history cannot establish cross-sport coefficients, opponent-strength or expectation effects, injury and transfer handling, crowd-pressure calibration, collusion controls, or live-feed fairness.