The Orange Engine
Accounts in. Rules applied. Assumptions exposed.
The engine works from filed accounts, the Premier League Handbook and public transfer evidence. It covers three things:
- The squad cost ratio. What each Premier League club spends on its squad against the revenue the rule counts, on the club pages.
- PSR calculations. How clubs stand against the Profitability and Sustainability Rules, the three-year limit on losses.
- The debt ledger. Who each club owes and how much, on the club finance pages and in Who the clubs owe.
It separates what is filed from what is estimated, refuses to publish a ratio where a squad cannot be fully accounted for, and records the assumptions that could still move a club's result.
The data cutoff is recorded separately from the code version, so a figure that moved because the world moved can be told apart from one that moved because the model did.
Current estimates
Club pages →For the 1 March 2027 test. The rule's figure counts a three-year average of profit on player sales; the second figure is the same cost base without it. Generated from the engine baseline, never typed.
| Club | The rule's figure | Before player sales | Range |
|---|---|---|---|
| Manchester United | 56.9% | 59.8% | 54.1 to 59.3% |
| Sunderland | 60.5% | 67.2% | 56.3 to 64.4% |
| Arsenal | 60.7% | 65.0% | 57.5 to 64.7% |
| Coventry City | 63.5% | 65.7% | 58.8 to 68.1% |
| Hull City | 66.4% | 73.6% | 61.3 to 71.7% |
| Leeds | 68.6% | 72.9% | 62.8 to 73.7% |
| Everton | 72.5% | 85.0% | 67.9 to 76.0% |
| Liverpool | 73.6% | 80.0% | 69.4 to 78.3% |
| Aston Villa | 74.2% | 96.9% | 68.8 to 77.0% |
| Manchester City | 74.4% | 86.4% | 68.3 to 78.5% |
| Brighton | 79.1% | 109.5% | 70.8 to 82.2% |
| Crystal Palace | 79.3% | 101.7% | 72.7 to 81.9% |
| Tottenham | 80.4% | 89.5% | 76.6 to 85.8% |
| Brentford | 81.6% | 106.8% | 75.3 to 84.7% |
| Newcastle | 81.7% | 108.8% | 76.4 to 85.5% |
| Bournemouth | 85.4% | 105.1% | 78.2 to 86.8% |
| Chelsea | 91.3% | 115.4% | 82.0 to 95.8% |
| Nottingham Forest | 97.5% | 124.5% | 87.6 to 99.4% |
| Ipswich Town | 97.9% | 109.0% | 91.0 to 104.1% |
| Fulham | 105.3% | 116.2% | 96.8 to 109.8% |
| Total: all 20 clubs together | 75.4% | 87.5% | a total, not ranked: no range |
The second column shows what the first is sitting on. Several clubs are inside the 85% threshold on the rule's figure but above 100% without player-sale profit: their compliance is being bought by sales that will drop out of the three-year average.
Questions that can still move the numbers
Where interpretation, estimation or incomplete public evidence could still move a published figure, each quantified where that can be done.
Rules interpretation
How should carrying value be treated when a capped contract is sold?
The five-year amortisation cap applies to contracts executed or extended on or after 12 December 2023. For those, the alternative reading adds £9.0–12.0m of player-sale profit across 8 registrations and moves Chelsea by 0.39–0.51 points. No club changes compliance state. The effect is small because the cap is young; it peaks about five years after the signing or extension.
Read the note →Who carries the charge on a loan with an unconditional obligation to buy?
The Guidance to Rule A.1.22 leaves amortisation with the lending club during a temporary transfer, and Appendix 2 does not address an unconditional obligation directly. On the adopted reading Manchester City keep Omar Marmoush’s £11.8m annual charge and Tottenham capitalise £50m in 2027/28. Recognising it at inception instead would give Tottenham 83.4% rather than 81.7%.
What does the engine not capture?
Every engine step was checked against Appendix 2 of the 2026/27 Handbook on 23 October 2026, and what public information allows was corrected in release V2CB. Eight items need information only the League and the club hold: fees paid or received for managers, earlier loan fees, bonuses, commercial revenue by source, outsourced catering, impairment and deals awaiting a fair-value decision, plus one line of Manchester City’s European average. Each is listed with its direction and its scale.
Read the note →What does the European reference amount pay when a club changes competition?
Rule A.1.29(e) pays the highest of three figures rather than what a club earned. Two of the three count only seasons in the competition qualified for, so a club moving between competitions cannot carry the old one’s money across, and the third is a property of the competition rather than the club.
Input construction
What share of a filed staff-cost line is players?
No club discloses it. The share is estimated from employment disclosures and is the largest single uncertainty at every club. A player-by-player salary build now runs alongside it; where the two disagree, the range is widened rather than re-centred.
How is next season’s payroll carried forward from a filed year?
Arrivals, departures and renewals are netted against the filed base. Across the series that movement runs from −5.6% to +13.4% of base, with one explained exception: Manchester United at about 19%, mostly the reported reversal of a Champions League-linked pay cut. One club once carried 36% unchecked, which cost a published article its conclusion. The comparison is now made by hand on every build, not yet as an automated gate.
Read the note →How is amortisation rebuilt without a forward schedule?
Registration by registration from the filed net book value, capped at five years from the execution date. Registrations already on the books at the accounts date are reconciled to the filed note, which makes the total right even where the distribution between named players is not.
Evidence and completeness
How is the playing population proved complete?
Against the Premier League’s own submitted senior and Under-21 squad lists rather than a club’s transfer record. At one club the lists found 92 identities missing, including a whole window’s senior departures. Publication is refused where the population cannot be closed.
Why is comparing two squad lists not enough?
A registration can appear on both lists and still leave and return between them. One January window carried five recalls and re-loans that no comparison of endpoints could see.
How are both sides of a transfer kept in agreement?
A cross-club transaction is one event with two mirrored ledgers and a dated recognition state, agreeing on amount, date and direction. The check is currently manual: one fee stood at £48m on one club’s disposal row and £47m on the other club’s purchase row for a week, and no gate could see it.
What the engine refuses to do
Publish a club whose playing population cannot be closed.
Every registration after the last filed accounts has to be accounted for against a complete source of population evidence, with the Premier League’s submitted senior and Under-21 lists as the primary check. A club that cannot be closed is withheld rather than estimated.
Accept a declaration that covers only part of a club.
A club census can be genuine, source-backed and internally consistent while describing the men’s first team alone. Scope is recorded as well as counts.
Compute a feedback-loop deduction from a March estimate.
Rule A.1.39 requires the season outturn, so no figure is produced before there is one.
Build a club from a secondary source’s summary of its accounts.
Two errors in one fortnight came from doing so, including a staff-cost figure that combined wages with transfer-fee amortisation and overstated a club by about six points.
Average two readings of a rule.
Where the Handbook is genuinely unclear, one reading is adopted and stated, and the size of the alternative is published.
Corrections and lessons
The model has been wrong in public, and the record stays up. The cases worth knowing:
A wage roll-forward that was never compared with anything
It ran at 36% of base against a series range of −5.6% to +13.4%, and cost a published article its threshold conclusion when corrected. A quantity that is never compared with anything is not validated by the care taken over its inputs.
Read the note →Transfer fees stored at the total rather than the base
A sweep of all fourteen clubs found 22 such fees across eight of them, moving eight ratios by −1.3 to +0.13 points. No club changed compliance state. The six clean clubs were all built from detailed window records rather than season summaries.
A movement type supplied where the evidence established only a population change
A squad list showed a registration leaving and the reporting said an offer had been accepted. It was a loan with an option. In the same window at the same club, one loan carried an obligation and another an option.
Every error found in two full audits had passed every gate
The gates tested completeness, consistency and reconciliation, and each failure was found instead by comparing a figure with its equivalent somewhere else.
Follow the record
Two public records keep the engine honest: one for the model, one for its forecasts.
V2CMOpen the changelog →The track recordEvery forecast, scored in publicEach club's forecast is frozen when published and scored against later evidence, such as results and filed accounts, with the error shown, including our own method errors.1 of 20 forecasts tested so farOpen the track record →How to cite a figure
Orange Engine estimate: Nottingham Forest 99.4% squad cost ratio for the 1 March 2027 test, version V2CM.
Independent reconstructions, not official calculations, and not advice.