The Orange Engine · Method notes →
What the engine does not capture
On 23 October 2026 every step of the engine was checked against Appendix 2 of the Premier League Handbook 2026/27, clause by clause, from the pages themselves. Everything that public information allows was corrected in releases V2CB and V2CF. What remains is below: eleven places where the rule needs information only the League and the club hold, in part or in full. One has a known size: Manchester City’s European average, about 0.1 percentage point on City’s ratio. Two are counted wherever an amount is filed or reported, fees paid and received for managers, and leave out only the cases with no amount, at about 0.04 to 0.05 percentage points for every £1m. Two are shown as ranges: loan fees from earlier seasons, up to about 1.9 percentage points on Bournemouth’s ratio, and impairment, up to about 2.4 on Ipswich’s on last season’s charges. The other four have no known size, so we cannot say that none of them would move a club across 85%. A tenth, added in release V2CG, is fees announced only as a headline, which can include add-ons not yet due: it decides which side of 85% Bournemouth are on, 85.4% or 84.8%. The list covers the clauses of Appendix 2. One limit sits outside them: which companies a club’s figures are taken from. Where a group’s other companies may fall inside the rule’s reporting perimeter and have not been read, as at Manchester City, or a separate retail company may, the article on that club says so.
Each item says what the rule asks, what the engine does instead, which way that moves a club’s figure and by how much where it can be said. The list changes only through a numbered engine release, when the evidence becomes available. The engine in use is V2CM.
Manchester City’s European average Rule A.1.29(e)
The rule. The rule’s European prize money (Rule A.1.29(e)) excludes the UEFA Super Cup.
The engine. City’s three-year average takes its filed 2023/24 UEFA broadcasting line, which may include the August 2023 Super Cup (UEFA’s plan paid each finalist €3.5m and the winner €1m more). UEFA’s final 2023/24 distribution by club was not found, so the line is kept.
Direction. If it does include the Super Cup, City’s revenue is overstated and its ratio slightly low.
Scale. £1.3m to £1.4m of revenue, about 0.1 percentage point on City’s ratio.
Fees paid to hire a manager Rule A.1.22(d)
The rule. Compensation a club pays another club to hire its manager counts as squad cost, spread over the manager’s contract, five years at most.
The engine. Counted since release V2CF wherever the amount is filed or reported, and marked as reported where it is not yet in filed accounts: Manchester City for Enzo Maresca (£17.2m), Newcastle for Matthias Jaissle (€11m), Ipswich for Gary O’Neil (€5m) and Crystal Palace for Pierre Sage (€3m guaranteed, up to €5m in the range). A fee with no reported amount is not counted.
Direction. Understates the squad cost of a club that paid a fee no one has reported.
Scale. About 0.04 percentage points on the ratio for every £1m paid, on a three-year contract.
Compensation received for a manager Rule A.1.29(g)
The rule. Compensation a club receives when another club takes its contracted manager counts with its profit on player sales.
The engine. Counted since release V2CF wherever the amount is filed or reported: Brentford’s £6.7m for Thomas Frank, in their 2024/25 accounts, and Chelsea’s £17.0m for Enzo Maresca, reported. Compensation with no reported amount is not counted.
Direction. Understates the revenue of a club paid an amount no one has reported, so its ratio is slightly high.
Scale. About 0.05 percentage points on the ratio for every £1m received.
Loan fees received before 2026/27 Rule Guidance to A.1.29(g)
The rule. A loan fee received above the player’s wages first offsets his costs, and any surplus counts with profit on player sales, across the whole period the rule averages.
The engine. This season’s loans: the surplus is counted. Earlier seasons: where a club’s filed accounts give loan income as its own amount outside profit on disposals (Bournemouth, Nottingham Forest, Ipswich, Aston Villa and Arsenal), it widens the assumption range, from none of it to all of it, because the accounts do not split loan fees from wages recharged; only income from 2 February 2024 counts (Rule A.1.53). Since release V2CF. Loan income for 2025/26 waits for the 2025/26 accounts, and eleven clubs’ accounts do not state loan income at all.
Direction. The central figures count none of it, so they understate revenue by whatever part is fee.
Scale. At most about 1.9 percentage points on Bournemouth’s ratio, if all of its filed loan income were fees, and 1.3 on Forest’s; less elsewhere. Not known for 2025/26 or for the clubs that do not state it.
Bonuses Rule A.1.22(b), A.1.51
The rule. Bonuses are not this season’s actual payments but a percentage of base pay: the average share of bonuses in pay over the last three seasons, excluding promotion bonuses.
The engine. Bonuses are whatever last season’s filed staff costs held (or an allowance in a player-by-player build).
Direction. Either way. A club coming off a title or European run is overstated here; one coming off a poor season is understated.
Scale. Not known: most accounts do not separate bonuses from base pay.
How commercial revenue is built Rule A.1.19
The rule. Sponsorship, food and drink and direct-to-consumer sales count gross; merchandise at gross profit; stadium hire on non-matchdays, hotels, conference centres and other teams at net profit; non-football and exceptional income not at all.
The engine. The filed commercial line, less the cost of merchandise. Ten clubs file one undivided commercial line.
Direction. Either way: items counted gross here that the rule nets (conference and banqueting, for one) overstate revenue; net income from stadium events outside the commercial line understates it.
Scale. Not known without each club’s revenue note by source. For scale: Sunderland’s 2024/25 base held £7.6m of conference and banqueting at its gross figure.
Outsourced catering Rule A.1.24(a)
The rule. Where catering is outsourced, an independently certified adjustment grosses the food and drink revenue back up to what the club would have taken itself.
The engine. Not modelled: the adjustment is the club’s own and not published.
Direction. Understates revenue at clubs that outsource catering, so their ratio is too high, by an amount not known.
Scale. Not known.
Impairment Rule Guidance to A.1.22(c) and (d)
The rule. Where a club writes down the value of a player’s registration in the season, the impairment counts as squad cost.
The engine. Not in the central figures: an impairment in 2026/27 is known only from the 2026/27 accounts. Shown here as a range, from what each club charged in 2024/25.
Direction. Understates squad cost at a club that impairs a registration.
Scale. If each club charged this season what it charged in 2024/25: Ipswich +2.4 percentage points, Arsenal +2.1, Leeds +2.0, Chelsea +1.9, Nottingham Forest +1.9, Aston Villa +1.3, Newcastle +0.1, Liverpool +0.0. The other clubs’ 2024/25 accounts print none, except Manchester City’s, which combine amortisation and impairment in one line. Liverpool’s accounts expect a charge of £14.4m in 2025/26, including other costs, relating to Diogo Jota’s registration; a charge that size in this season would be +1.9. A range only, never a central value.
Fees announced only as a headline Rule A.1.43 and A.1.22(c)
The rule. A transfer fee counts the amounts payable or which have become due. An add-on that is not yet due counts only once it is recognised.
The engine. Where neither club publishes the fee and the reports give only a headline figure, with no split between the guaranteed fee and the add-ons, the engine holds the headline, so add-ons not yet due may be counted. A fee is changed only when a club or reputable outlet, read on the page, prints the guaranteed sum. Where the guaranteed fee and the add-ons are published apart, the add-ons are held in the range only, since release V2CG.
Direction. Overstates the squad cost of a club whose headline includes add-ons not yet due, so its ratio is high.
Scale. Where it matters most: Bournemouth signed Amine Adli from Leverkusen at a reported €30m. On that headline Bournemouth are at 85.4%. If the reported €21m guaranteed fee is right, the add-ons are not due and Bournemouth are at 84.8%, below 85%. Elsewhere the effect is a few tenths of a point at most.
Deals awaiting a fair-value decision Rule B.11.4
The rule. A transaction under the League’s fair market value assessment that would raise revenue is left out until the League confirms it.
The engine. Every commercial line counts in full: which deals await a decision is not public.
Direction. Overstates revenue at a club with an associated-party deal not yet confirmed, so its ratio is too low.
Scale. Not known; largest where associated-party sponsorship is large.
Sale profit after the year end Rule A.1.13
The rule. Profit on player sales in a period counts as the accounts record it (Rule A.1.13).
The engine. Where a club’s accounts print the profit on sales made after the year end, we use that figure. Where they do not, we estimate it from the fee and our modelled book value. Clubs affected: Arsenal, Brentford, Brighton, Coventry, Everton, Forest, Hull, Ipswich, Leeds, Manchester City, Newcastle, Palace, Tottenham and Aston Villa. Their 2025/26 accounts will replace the estimate.
Direction. In the six cases we can check, the estimate ran above the printed figure in five, by a median of about 14% of the fees involved, so these clubs’ ratios may be slightly low. We have not separated the cause: the guaranteed fee, sell-ons owed, agents’ costs and discounting on instalments may all play a part.
Scale. If the same pattern held, most would move by under 2 points. Brentford’s figure is the one where it could matter: the widest case takes it above 85%.
One fee not confirmed by two independent reports Rule A.1.43
The rule. A signing is carried at its cost (Rule A.1.43).
The engine. Omari Hutchinson’s move from Chelsea to Ipswich in 2024 was undisclosed, and reports differ on the guaranteed part: £18m or £20m. We use £20m, with a range of £18m to £22.5m.
Direction. Either way.
Scale. It changes Ipswich’s figure by 0.01 points at most, and no longer enters Chelsea’s.
The audit, its page images and the tests that hold each item in place are in the engine’s repository (outputs/rule-audit/2026-10-23 and engine/test_scr_clauses.py). Independent estimate, not an official calculation.