The Orange Engine
Changelog: every change to the model, and why.
What changed at each version, which club exposed it, how much each published figure moved, and which automated test now stops it happening again. Almost every entry exists because an audit found something the engine could not yet represent: the tests are the lasting result, not just the corrected numbers.
What counts as an entry: every change that can alter a published estimate, a source, a test or a rule about what may be published. Cosmetic changes to the website are not recorded here.
V2CMcurrent engine version
86recorded changes, across 78 engine versions
20clubs whose figures a change has moved
112automated checks run on every update · updated 3 November 2026
By club
Most readers arrive about one club. Each club's page carries only the changes that moved its figure or its treatment, with where it stands now.
How a change reaches the site
A correction is not finished when the code passes. Several errors below were found at the last step, by reconciling what an article said with what the engine computed, which is why that step exists.
- 1Code changeWith an automated test that fails without it.
- 2Baseline snapshotEngine version and data cutoff recorded separately, so a model change can be told apart from a change in the world.
- 3This changelogWhat changed, which club exposed it, and what moved.
- 4The Engine pageUpdated where the change alters what the engine does or refuses to do.
- 5Affected articlesRe-read against the new baseline by a checker that flags figures from a superseded version.
Every change, newest first
The latest 6 changes are shown in full; open any older one to read it. Open interpretive questions that have not changed the adopted treatment are on the Engine page instead.
- V2CMCorrection
Corrections at twelve clubs, from a check of transfer, loan and contract dates and transfer fees against the clubs' own pages and named reports. Crystal Palace's fee for Anan Khalaili is the reported 21 million pounds, where we had 8 million with no source: Palace's ratio moves from 78.2% to 79.3%, which puts Palace just above Brighton in the ranking. At Tottenham, Tanguy Ndombele's contract ended by agreement in June 2024, a year earlier than we had recorded, which lowers the loss on ending it: Tottenham move from 80.9% to 80.4%. No other club's ratio moves, and no club crosses 85% or 115%.
- V2CLCorrection
Corrections at Leeds. Jean-Matteo Bahoya's loan from Eintracht Frankfurt now carries a reported fee of 2 million euros, where we had allowed a range of nil to about 2 million pounds; whether Leeds hold an option or an obligation to buy him is unresolved, so the case in which they buy him stays published, now on the 23 million euros reported as the price, without the loan fee and add-ons. Leeds's ratio moves from 68.3% to 68.6%. No other club moves, and no club crosses 85% or 115%.
- V2CKCorrection
Four corrections at two clubs. Fulham: Hugo Larsson is now treated as bought, at a reported fee of about 25 million euros, with the loan the clubs describe kept as the other case in the range; and the range now takes in a filed figure that does not fit the reported fee for Kevin. Forest: Gustavo Sa is on loan to Olympiacos, not sold, and Omari Hutchinson's loan to Milan brings a reported fee. Fulham's ratio moves from 102.7% to 105.3%, Forest's from 97.4% to 97.5%. No other club moves, and no club crosses 85% or 115%.
- V2CJCorrection
Three Chelsea sales return to the periods they belong in. Omari Hutchinson's sale to Ipswich was announced on 30 June 2024, and Chelsea's accounts for that year name it; we had read the date as 1 July. Bashir Humphreys and Kepa Arrizabalaga belong in the March 2028 test's period. Chelsea's ratio moves from 92.3% to 91.3%, and 97.5% at the 2028 test, from 98.6%. No other club moves, and no club crosses 85% or 115%.
- V2CICorrection
Sale profit now follows the filed accounts where they print it. Chelsea's accounts show £31.8m of profit on sales after the year end, where we had modelled £82.4m, and two sales we counted in 2025/26 belong to 2024/25: its ratio moves from 88.6% to 92.3%. Liverpool's accounts show £104.7m against our £132.5m: 72.7% to 73.6%. A check of every transfer against published sources corrects 19 entries, which moves Sunderland to 60.5%, Hull to 66.4% and Coventry to 63.5%. No club crosses 85% or 115%.
- V2CHCorrection
A correction to our records, not to any figure: five source references for transfer and loan figures at Crystal Palace, Tottenham and Sunderland were filed under the wrong reference, so the database could not match them to the figures they support. They now point to the right entries. No club's figure, range or position moves.
V2CGCorrectionSixteen transfer entries at eleven clubs corrected after a check of every transfer against published sources, each read on the page cited. Add-ons that have not yet fallen due on a signing now count in the range only, as the rules count a fee when it is payable. Eight clubs' estimates move by less than a percentage point; none crosses a line.
- Arsenal are at 60.7%, from 60.1%: Bruno Guimaraes's contract runs to June 2030, not June 2031, because the option year is not part of it, so his fee is spread over a shorter contract (Rule A.1.22(c)).
- Sunderland are at 61.2%, from 60.6%: Kevin Danso's £2.5m loan fee is a squad cost (Rule A.1.22(e)).
- Chelsea are at 88.6%, from 88.3%: Geovany Quenda's guaranteed fee is £43.4m, not £40.0m, and Pedro Neto's £51.4m, not £54.0m, each with the part not yet due left out (Rules A.1.43 and A.1.22(c)); Kian Best's contract runs to June 2027, not June 2029, so his fee is spread over a shorter contract (Rule A.1.22(c)).
- Coventry City are at 64.0%, from 64.6%: Ethan Pinnock's contract runs to June 2029, not June 2028, so his fee is spread over a longer contract (Rule A.1.22(c)).
- Nottingham Forest are at 97.4%, from 97.6%: Daniel Munoz cost £20m guaranteed, not £22m, with £2m of add-ons not yet due in the range only (Rules A.1.43 and A.1.22(c)); Mason Ayre's scholarship runs to June 2028, not June 2029.
- Crystal Palace are at 78.2%, from 78.1%, and Tottenham at 81.0%, from 80.9%: Munoz's sale brought Palace £20m guaranteed, not £22m, and Djed Spence's sale brought Tottenham £26.9m, not £30.0m, with add-ons counted only when they fall due (Rule A.1.29(g)).
- Bournemouth are at 85.4%, from 85.5%: Alex Toth's contract runs to June 2031, not June 2030, so his fee is spread over a longer contract (Rule A.1.22(c)).
- Brentford (Sam Sene-Richardson to June 2028 and Darwin Guagua to June 2031), Brighton (Michael Dike to June 2027 and Nils Ramming to June 2030) and Fulham (Cesar Palacios from 3 August 2026) take corrections to contract lengths and a date (Rule A.1.22(c)); no published figure moves.
What moved: Arsenal 60.1% → 60.7%; Sunderland 60.6% → 61.2%; Chelsea 88.3% → 88.6%; Crystal Palace 78.1% → 78.2%; Tottenham 80.9% → 81.0%; Coventry City 64.6% → 64.0%; Nottingham Forest 97.6% → 97.4%; Bournemouth 85.5% → 85.4%.
ArsenalSunderlandChelseaCrystal PalaceTottenhamCoventry CityNottingham ForestBournemouthBrentfordBrightonFulham
V2CFCorrectionFees clubs paid for managers, and compensation clubs received for releasing them, are now counted where they have been reported or filed. Six clubs' estimates move by less than a percentage point; none crosses a line.
- Manchester City are at 74.4%, from 73.7%, with Enzo Maresca's reported £17.2m fee spread over his three-year contract.
- Chelsea are at 88.3%, from 89.1%, with the £17m reported from City for releasing Maresca.
- Brentford are at 81.6%, from 82.3%, Ipswich at 97.9%, from 97.1%, Newcastle at 81.7%, from 81.3%, and Crystal Palace at 78.1%, from 77.8%.
- Loan fees received that the accounts print outside player sales widen the ranges at Bournemouth, Forest, Ipswich and Villa; no central figure moves for them.
- Brentford's range now ends just below 85%, and Newcastle's now reaches just above it.
Manchester CityChelseaBrentfordIpswich TownNewcastleCrystal Palace
V2CECorrectionSeven articles are corrected on points a review left open, each with a dated note. No club's estimate for this season moves.
- Tottenham's European places in our count of what Manchester City's results cost other clubs are recomputed on the method of 14 October: £82.0m over the nine seasons, from £122.6m.
- The commercial-revenue method now gives Manchester United's 2025/26 miss in one form, the shortfall below each forecast.
- Statements without a source or a stated basis are sourced from our own data, or shown both ways where no rule decides.
V2CDCorrectionThe items a review of every article found still open after the last release are corrected, each with a dated note, and the Manchester City points-deduction scenarios are rerun on our current figures. No club's estimate for this season moves.
- Figures that had fallen behind are brought into line, among them the gap between Crystal Palace and Everton before player trading, 15.2 points, from 15.5.
- Statements of the rule are corrected where they had drifted: the five-year cap applies to contracts signed or extended on or after 12 December 2023, and a loan fee is set first against the player's wages.
- The Manchester City deduction scenarios are rerun: after a 40-point deduction City would be at 85.4%, from 85.6%, and after 55 points they would lose 17 league points in their second season back, from 20.
- Manchester City's appeal of 2 October is recorded in every piece that still gave a deadline.
Manchester CityManchester UnitedCrystal PalaceEvertonBrightonBrentfordLiverpoolTottenham
V2CCCorrectionEditorial pass, 25 October: wording only, no figure or claim changed. The articles are rewritten for clarity to the house style, and the correction notes of 23 October are reworded so that each gives the figures first and its cause in plain words, with every figure unchanged.
- Openings now state each article's finding, terms are explained where they are first used, and ratio points are written as percentage points throughout.
- The correction notes of 23 October are reworded, with every figure in them unchanged; each article's top line now names the latest change to its figures.
V2CCCorrectionThe figures for later tests now count every sale in the right transfer window at ten more clubs, and every sentence that says when a sale leaves the average is now checked against the rule. No club's estimate for this season moves.
- The March 2028 test counts sales from 4 February 2025. The sales from then to each club's year end are now listed one by one at Arsenal, Brighton, Coventry, Crystal Palace, Fulham, Hull, Ipswich, Leeds, Manchester City and Sunderland.
- Brighton's figure for the March 2029 test, on frozen costs, is about 92%, from about 99%: its sales of February to June 2026 stay in the average for that test.
- Sunderland's sale-profit average at the March 2028 test is about £22.8m, from about £11m: Bellingham's and Watson's sales of June 2025 stay in for that test.
- Articles that said a whole financial year leaves the average are corrected, each with a note.
What moved: No ratio for the 2027 test moves. Brighton, March 2029 test: about 99% → about 92%. Sunderland, average at the March 2028 test: about £11m → about £22.8m.
V2CBCorrectionWe checked every step of the engine against the squad cost rules, clause by clause, and corrected everything public information allows in one release. Contracts signed since December 2023 for more than five years are now capped at five years even for players already on the books, Villa's Super Cup money counts, and the figures for later tests are worked out window by window and restored. Five estimates move, none by more than 0.6 of a point.
- The five-year cap: a contract signed or extended since 12 December 2023 for more than five years is now spread over five years for players who were already on the club's books, as the rule says, not over its full length (Chelsea, Manchester City, Aston Villa, Manchester United).
- The UEFA Super Cup: Aston Villa's €4m for playing the 2026 Super Cup now counts, as the rule says once the amount is known.
- Sunderland: their FA Cup home tie against Stoke City in January 2025 is now evidenced, so they are credited with two home cup matches, not one.
- Figures for later tests: the figures withdrawn earlier on 23 October are worked out window by window and restored.
- What the engine cannot capture from public information is now listed once, with its direction and size, on its own page.
- Every figure an article prints from the engine is now registered and recomputed at each release. The new check found figures in eight more articles that earlier corrections had not carried through; each is corrected, with a note.
- Regulation part 5, which gives each club's distance from a ratio of 60%, is restated on the corrected figures.
What moved: Chelsea 88.6% → 89.1%; Manchester City 73.1% → 73.7%; Aston Villa 74.4% → 74.2%; Manchester United 56.8% → 56.9%; Sunderland 60.8% → 60.6%
ChelseaManchester CityAston VillaManchester UnitedSunderland
V2CACorrectionWe had averaged each club's profit on player sales over three financial years, but the rule averages the sales made in the six most recent transfer windows, which for the March 2027 test start in February 2024. Sales between February and June 2024 now count, among them several made before the June 2024 deadline of the previous rules, and ten estimates move: Chelsea's to 88.6%, from 92.5%, and Newcastle's to 81.3%, from 84.9%, the largest.
- The period the rule averages: the profit or loss on player sales over the six most recent transfer windows, divided by three, not over three financial years. For the March 2027 test the period starts on 2 February 2024, so sales from then until each club's 2024/25 year began now count.
- Brentford: David Raya's sale to Arsenal, which Brentford booked in June 2024, now counts, and Mark Flekken's sale to Bayer Leverkusen in June 2025 is no longer counted twice.
- Figures for later tests: where an article gives a club's figure for a later year, it is now worked out window by window; where we have not yet done that, the figure is withdrawn while we recheck it.
What moved: Chelsea 92.5% → 88.6%; Newcastle 84.9% → 81.3%; Aston Villa 77.4% → 74.4%; Nottingham Forest 99.3% → 97.6%; Brentford 83.8% → 82.3%; Everton 73.3% → 72.3%; Manchester City 73.8% → 73.1%; Tottenham 80.4% → 80.9%; Leeds 68.6% → 68.3%; Manchester United 56.9% → 56.8%
ChelseaNewcastleAston VillaNottingham ForestBrentfordEvertonManchester CityTottenhamLeedsManchester United
V2BZCorrectionCorrections found while checking our inputs against the clubs' own pages: thirteen signing, sale and contract dates now follow the club's own completion announcement, three fees agreed in euros now use the official exchange rate on the day the deal completed, and one contract end is corrected. Two estimates move: Crystal Palace's to 77.8%, from 78.1%, and Everton's to 73.3%, from 73.4%.
- Dates: where a club's own announcement of a completed signing, sale or new contract gives a different day from the one we held, the club's day now counts. Thirteen dates at Chelsea, Everton, Bournemouth, Brighton and Crystal Palace move, by between one day and about a month, and each contract's length is measured again from its corrected start.
- Fees in euros: three fees agreed in euros are now converted at the European Central Bank's rate on the day each deal completed: Dan Ndoye's and Ousmane Diomande's at Nottingham Forest, and Anthony Gordon's sale by Newcastle.
- Crystal Palace: Quinten Timber's contract runs to 2031, as the club announced, not 2030, so his fee is spread over one more season.
What moved: Crystal Palace 78.1% → 77.8%; Everton 73.4% → 73.3%
Crystal PalaceEvertonChelseaBournemouthBrightonNottingham ForestNewcastle
V2BYModelPlayers a club has loaned out are now recorded at all twenty clubs, with an allowance for the part of their wages the borrowing club pays. Where a named source says how much the borrower pays, the saving counts in our central figure; where nobody says, it only widens the range. Four central figures move, the largest Nottingham Forest's, from 102.3% to 99.3%.
- A club that loans a player out often stops paying some or all of his wages, which lowers the squad cost the rule counts. We had recorded loans out at some clubs and not at others. Every club's squad is now checked against its own website and the League's registration return, and every player out on loan in 2026/27 is recorded with the loan.
- Wages are not published, so each loaned player carries an estimated salary range based on what the club paid for him. How much of it the borrowing club pays is stated for only a few players. Where a named outlet reports the split, we count the saving in the central figure. Where nobody states it, we no longer assume a half share: the saving can be anything from nothing to the whole wage, so it widens the range and leaves the central figure where it was.
- The same treatment applies to last season's loans, which lowered the wages in the accounts we start from. Where a player is back, or was counted out of the wage bill at his full salary, we add his saving back; where the split is not stated, that too goes into the range.
- Nottingham Forest's central figure falls from 102.3% to 99.3%. Omari Hutchinson's wages at AC Milan are reported to be paid in full by Milan, which takes off 1.7 points. Gustavo Sá was sold to Olympiacos on 30 July 2026, and we had still counted him in the squad. No fee has been published, so we record the sale at what Forest paid for him, as an estimate; that takes off 1.2 points. Forest are still above 85%.
- Manchester United's figure moves from 57.3% to 56.9%, Arsenal's from 59.9% to 60.1% and Liverpool's from 72.9% to 72.7%, on reported or relayed wage splits.
- At nine other clubs only the range moves. Because the possible saving on this season's loans is larger than the possible add-back for last season's, the bottom of each range falls further than the top: Chelsea's range is now 83.0% to 97.2%, Brighton's 70.7% to 82.1%.
- Tottenham's saving on this season's loans is held out for now, because the club's own wage figures already include part of it. Newcastle's is held out because last season's loans could not be checked against a source we accept.
What moved: Nottingham Forest 102.3% → 99.3% (range 90.5 to 102.0%); Manchester United 57.3% → 56.9%; Arsenal 59.9% → 60.1%; Liverpool 72.9% → 72.7%; ranges only: Chelsea, Brighton, Brentford, Bournemouth, Aston Villa, Crystal Palace, Manchester City, Everton, Tottenham; league 75.8% → 75.6%
Nottingham ForestManchester UnitedArsenalLiverpoolChelseaBrightonBrentfordBournemouthAston VillaCrystal PalaceManchester CityEvertonTottenham
V2BXModelWhen a club sells a player it once bought, the club that sold him first may be owed a share of the fee: a sell-on. We counted a sell-on as a cost to the club paying it, but had no proper way to count it as income for the club receiving it. This release adds that, applies it to two sell-ons, and corrects the fee Everton paid for Iliman Ndiaye. Newcastle's figure moves from 84.7% to 84.9%.
- The rule averages each club's profit on selling players over three seasons. A sell-on a club receives on another club's later sale is part of that profit, in the club's own financial year, even though it sells nobody. We now record it that way: the full amount, with no selling costs taken off, and without treating it as the sale of one of the club's own players. A payment a club's accounts show for clauses on former players is counted the same way, as a cost.
- Where both clubs are among the twenty we cover, the amount one club records as paid must match, deal by deal, the amount the other records as received. Each club's side is placed in its own financial year, which can differ, as clubs' year ends differ.
- Newcastle owe Everton a share of the profit on Anthony Gordon's sale to Barcelona in May 2026. Reports put it at 15% of Newcastle's profit, about £3.1m, though they disagree on how it is worked out. We now count it at both clubs, as a reported figure rather than a confirmed one: a cost to Newcastle and income to Everton. Across the two clubs it nets to zero.
- Coventry City's accounts say that after their 2024/25 year end the club received £4.0m and paid £0.8m under sell-on and similar clauses on former players. The accounts do not say which players. We had entered the net amount as if it were one sale of Viktor Gyökeres, with selling costs taken off. We now record both amounts as the accounts print them, for no named player. Coventry's figure does not move at the precision we publish.
- Everton paid Marseille €18.5m for Iliman Ndiaye in July 2024, with up to €2.5m more in add-ons reported but not known to have been paid. We had recorded £20m, which was Marseille's own earlier fee for him. At the exchange rate of the day the guaranteed fee is £15.7m; the add-ons are noted but not counted. The lower fee raises Everton's costs for the rest of the squad slightly and raises its profit when he was sold to Manchester City this September. With the Gordon sell-on, Everton's central figure stays at 73.4%.
- Newcastle's central figure is now 84.9%, in an assumption range of 79.1% to 88.8%. It is below 85% on the central estimate, and the range spans the line: the estimate does not reliably tell which side of 85% Newcastle are on.
What moved: Newcastle 84.7% → 84.9% (range 79.1 to 88.8%); Everton 73.4% → 73.4% (range 68.9 to 77.0%; 84.9% before player trading, from 84.4%); Coventry City 64.6% → 64.6%
V2BWPublishing gateA safety check in the engine: it now refuses to run if the file holding the league-wide figures is missing, instead of carrying on without them. No club's figure changes.
- The engine reads one league-wide file: the Premier League's payment parameters, the reference amounts for European competition, and the final tables it checks each club's league positions against. If that file had gone missing, the engine would have carried on with those parts left empty and said nothing. It now stops and names the missing file. No figure changes.
What moved: No club moves.
V2BVModelLast season's commercial revenue is not in any club's filed accounts yet, so we estimate it. Until now each estimate was a judgement of ours. From this release every club's estimate is made by one published method, fitted on 81 filed year-on-year changes, with an assumption range sized from how far such forecasts have missed in the past. Seventeen estimates move. No club changes side of a threshold.
- The rule counts a club's commercial revenue for last season, 2025/26. Those accounts are filed between late 2026 and spring 2027, so for 19 clubs the figure is an estimate. We had built each one by hand, from the last filed figure and our own view of the club's season, and gave each a margin of 7% either way.
- The new method starts from the club's last filed commercial revenue and grows it by a rate that depends only on the size of that revenue: larger commercial businesses have grown more slowly in percentage terms. The rate is 19.3% for a club starting from £20m and 6.8% for one starting from £350m. It is fitted on 81 changes taken from filed accounts, every figure read on the page.
- We chose the method on one test only: how well it would have forecast commercial revenue in past seasons. It was compared with assuming no growth and with a rule that also used each club's own last change. We did not look at what any method did to a club's squad cost ratio, or at which side of 85% it left a club.
- The margin of 7% either way was too narrow. Past forecasts of this kind have missed by more, and by more for small clubs than for large ones. The new assumption range is sized from those misses and is wider on the upper side for smaller clubs. Most clubs' published ranges are wider as a result.
- Four clubs' assumption ranges now span the 85% line: Bournemouth (central figure 85.5%, range 79.5 to 87.8%), Brentford (83.8%, 79.5 to 87.5%), Newcastle (84.7%, 78.9 to 88.6%) and Tottenham (80.4%, 76.4 to 85.6%). For these the estimate does not reliably tell which side of the line the club is on.
- Sunderland and Ipswich Town changed division, where the size rule does not apply. Sunderland's estimate is the typical change of nine promoted clubs and Ipswich's the typical change of six relegated clubs. Both are provisional until their accounts are filed. Coventry City's and Hull City's ratios do not move: the rule counts its £15m minimum for both, and their own commercial revenue is below it. Our estimate of that revenue does change: it is now each club's last filed figure with no growth assumed, provisional until their accounts are filed.
- Manchester United does not move: its 2025/26 figure is filed. Leeds now has the cost of its merchandise taken off, as the rule requires and as at every other club; that cost is estimated, not filed.
- Two figures measured before player trading now sit a fraction of a percentage point from a line: Manchester City 84.9%, from 87.1%, which is 0.1 percentage points below 85%, and Chelsea 114.4%, from 115.5%, which is 0.6 percentage points below 115%. Both margins are smaller than the uncertainty in the commercial estimate, so the estimate does not reliably tell which side of either line the club is on before player trading. No published ratio changes side of a threshold.
- We have withdrawn our estimate of what the five-year amortisation cap would add at Chelsea if it reached the eight registrations signed before it. Two ways of calculating it give different answers, and we will publish neither until they are reconciled.
- From this release we write movements and margins of a ratio in percentage points and sporting deductions in league points, so the two are never confused.
- The forecasts of record are frozen and scored as frozen. The forecasts we published before this release stand in the track record as published and will be scored against the 2025/26 accounts as they were. The estimates of this release are held in its own timestamped record and will be compared with the accounts when they are filed.
What moved: Arsenal 59.5% → 59.9%; Bournemouth 85.5% → 85.5% (range 79.5 to 87.8%); Brentford 84.2% → 83.8%; Brighton 79.5% → 79.1%; Chelsea 93.2% → 92.5%; Everton 73.3% → 73.4%; Nottingham Forest 103.4% → 102.3%; Fulham 104.0% → 102.7%; Ipswich Town 95.8% → 97.1%; Leeds 67.4% → 68.6%; Liverpool 74.0% → 72.9%; Manchester City 75.4% → 73.8%; Newcastle 83.9% → 84.7%; Crystal Palace 78.9% → 78.1%; Sunderland 63.4% → 60.8%; Tottenham 82.3% → 80.4%; Aston Villa 76.0% → 77.4%
ArsenalBournemouthBrentfordBrightonChelseaEvertonNottingham ForestFulhamIpswich TownLeedsLiverpoolManchester CityNewcastleCrystal PalaceSunderlandTottenhamAston Villa
V2BUPublishing gateA new check in the engine: wherever an input is recorded as a range, its low end must not be above its high end, or the engine refuses to publish. No club's figure changes. Our sensitivity testing now moves every input both up and down.
- Some inputs are recorded as a range with a low and a high end, such as the share of a club's wages paid to its squad, or a loan fee that has been reported only as a range. Nothing had checked that the low end was not above the high end. It is not at any club today, and from now on a figure cannot be published from a range that is the wrong way round.
- We test how much each input matters by changing it and seeing whether a published figure moves. Until now every input was changed upward only. Each is now changed both up and down, and we keep the larger effect, because some inputs matter in one direction only: Arsenal's count of European home matches was one such case.
What moved: No club moves.
V2BTCorrectionThe rule counts European home matches at the higher of two figures, and we had applied only one of them. Three estimates move: Arsenal's the most, to 59.5% from 60.7%.
- For a club in Europe the rule adds gate income for European home matches: the higher of the home matches the club is guaranteed this season and the average number it played at home in that same competition over the last three seasons, rounded up. We had counted the guaranteed matches only. The engine now works out both and takes the higher, for every club.
- Arsenal played 7, 7 and 5 Champions League home matches in the last three seasons, an average of 6.33, which rounds up to 7. We had 4.
- Manchester City played 5 in each of the last three seasons. We had 4.
- Brighton's play-off home leg in August 2026 is a guaranteed home match of this season's competition, so its count is 4. We had 3.
- We read 'the relevant competition' as the one the club is in this season: a past season in a different European competition does not count. On the other reading, any European competition, Aston Villa and Liverpool would also move; that case is set out in our method note.
What moved: Arsenal 60.7% → 59.5%; Manchester City 75.7% → 75.4%; Brighton 79.9% → 79.5%
V2BSCorrectionCorrections from a club-by-club check of our inputs against each club's own announcements, the League's published squad lists and filed accounts. Seven estimates move, Fulham's the most: 104.0%, from 106.4%.
- Fulham: we had treated Hugo Larsson as bought from the day he arrived, on a report that an obligation to buy had been triggered by his first appearance. Fulham, Eintracht Frankfurt and the Bundesliga describe a loan with a conditional obligation that completes at the end of the season, and Frankfurt still listed him as its player on loan in October. We could not read the one report that says the condition has been met. Our main estimate now treats him as a loan; the purchase case is inside the range.
- Manchester City: the three-year average of UEFA prize money is now built from its three seasons (two filed, one estimated) and is higher than the round figure we had used.
- Chelsea: the fee for Axel Disasi's loan to Crystal Palace is now counted as received by Chelsea, as it was already counted as paid by Palace, and is set against the cost of his registration, which Chelsea still holds. The amount is an estimate.
- Brighton: Pascal Gross's contract runs to June 2027, as the club announced, not 2028.
- Crystal Palace: Borna Sosa's fee is a reported £3m, where we had none. Coventry: Caleb Yirenkyi's is a reported £26m, where we had £23.1m.
- Signing dates at eleven clubs now follow the club's own dated announcement. Most moves are a few days and change nothing that we print.
- Three of the League's squad lists had been copied into our records with errors. They are corrected against the League's pages; no estimate moves.
What moved: Fulham 106.4% → 104.0%; Manchester City 76.1% → 75.7%; Chelsea 94.3% → 93.2%; Brighton 79.5% → 79.9%; Crystal Palace 78.5% → 78.9%; Coventry 64.1% → 64.6%; Brentford 84.1% → 84.2%
FulhamManchester CityChelseaBrightonCrystal PalaceCoventry CityBrentford
V2BRCorrectionA small correction to our update on Manchester United's 2025/26 results: 57.5%, from 57.6%. The club's main estimate, 57.3%, does not change, and no other club moves.
- To estimate United's player wages from the group's accounts we apply a share to the group's total staff costs. We had worked that share out from one company's table of staff costs and applied it to the listed parent's table, which also includes share-based payments. Both now come from the parent's table. The wages estimated for 2026/27 from the 2025/26 results fall by about £0.25m.
- The same update's range is now 55.0% to 60.3%, from 54.9% to 60.1%. United's 2025/26 accounts show £5.2m of costs for loss of office in the year its head coach left, without saying how much was his. The rule counts a manager's termination payment as squad cost, so the range now allows for anything between none and all of it. The central figure does not include it.
- We also tightened our own checks. Every figure in our evidence files must now equal the figure it supports in the model, or carry a recorded reason why it differs. Two of United's records had fallen out of step when the estimate moved to the group's accounts; no published figure was affected.
What moved: Manchester United, update on the 2025/26 results 57.6% → 57.5%
V2BQPublishing gateA new check in the engine: wherever an estimate is recorded with a lowest and a highest plausible value, the estimate itself must lie between them, or the engine refuses to publish. No club's figure changes.
- Some inputs are recorded as a central value with a documented low and high, and the published range is drawn between them. Nothing had checked that the central value sat inside its own bounds. It does at every club today, and from now on a figure cannot be published if it does not.
V2BPDataManchester United's estimate is now built on the accounts of the group the rule describes, not the club company alone. The group includes the companies that sell the club's merchandise, run its media channel and field its women's team. The ratio falls from 59.0% to 57.3%. No other club moves.
- The rule tests a club together with its football companies. United's shop and licensing business, its media channel and its women's team sit in other companies of the same group, and their revenue was outside the accounts we read. We now read the group's accounts for 2024/25 and the figures the club's listed parent printed for 2025/26.
- Commercial revenue is higher, by about £61m in 2024/25. The rule counts merchandise at its profit, not its sales, so we take off the group's retail costs, about £37m, which the retail company's own accounts show to be the cost of the goods sold.
- Player wages do not change. The squad is employed by the club company, so we count the same wages as before against the larger group total.
- The update we published on United's 2025/26 results used the group's figures converted to the club company. It now uses them as printed. That update moves from 58.6% to 57.6%.
- The September forecast of record, 58.7%, is not rewritten. It was made on the club company's accounts.
- Not counted, and stated as a limitation, is the subscription income of the club's own media channel, about £5.8m in 2024/25. It sits in the group's broadcasting line, outside its commercial figures. If the rule counts it, the ratio would be about 0.4 points lower.
- The update on the 2025/26 results now rests on the audited annual report of the club's listed parent, and is United's 2025/26 accounts checkpoint.
What moved: Manchester United 59.0% → 57.3%
V2BOModelWe now measure every contract from the day a player was registered to the day the contract ends. A player who signs in August on a "five-year deal" has a contract of four years and ten months, because contracts end on 30 June. We had spread his cost over five full years, which made each year's charge a little too small. On that change every club's ratio rises, by 0.1 to 0.9 points. Five clubs also move because we now read the accounts of the group the rule describes, not the club company alone. No club's central figure crosses a threshold.
- The cost of signing a player is spread over the length of his contract. We had used the length as announced, in whole years. We now use the time between the registration date and the contract's end date.
- 504 contract lengths changed. For 190 of them a source states the end date. For the other 314 we have derived it as the 30 June that the announced length points to, which is when contracts in English football end, and we mark those dates as derived until we have a source for each.
- No length changed by as much as half a year. The effect is a few weeks per player, added up across each squad.
- Bournemouth's central figure moves from 85.2% to 85.5%, in an assumption range of 80.1% to 87.7%. The range still spans the 85% line, so the estimate does not tell which side of the line the club is on.
- Brentford (84.1%) and Tottenham (82.3%) move closer to the line. Neither is above it at the central figure.
- The same rule now applies to the contract a loan with an obligation to buy would create. One published figure moves on it, Fulham's, by 0.1 points.
- The rule tests a club together with its football companies. At Brighton and Sunderland the women's team is a separate company whose revenue the rule counts, so their ratios fall, by 0.3 and 0.2 points. At Ipswich the group's accounts remove a fee the club charges its own parent, so its commercial revenue is lower and its ratio rises by a further 0.9 points.
- Newcastle's group includes the women's team and a fan zone beside the stadium that opened in August 2024. We count the women's team's revenue in full and the fan zone at its profit, which lowers the ratio from 84.7% to 83.9%. This rests on our reading of the rule. The rule counts matchday food and drink in full, and if the League treats the fan zone that way the figure is lower, so the assumption range now runs from 80.7% to 86.0%.
- Manchester City's women's team sits in another company of the same group. We add its ticket and broadcasting income, which lowers the ratio by 0.3 points. We do not add its sponsorship income, because the accounts do not show that it comes from outside the group.
- A women's team's company often receives money from its own club. We count only what comes from outside the group, and nothing extra where the team is already inside the accounts we read.
- For seven clubs the group's accounts cannot yet be applied, because the documents do not print what is needed or, at Manchester United, because the change is large enough to need a release of its own. Their figures are unchanged on this point and each club's article says so.
- A check that used to require whole years for a summer signing now requires each length to agree with its dates. It still refuses the error it was written to catch.
- A new check stops a club's European revenue being switched off by mistake. It compares each club's record with the list of clubs in the league phase of their competition.
- We read every figure we take from the clubs' accounts against its page again. Twelve did not match. None of them is used in the ratio, and all are corrected. One, Nottingham Forest's amortisation for 2023/24, had been quoted in our article and was corrected there on 4 October.
What moved: Arsenal 60.4% → 60.7%; Aston Villa 75.7% → 76.0%; Bournemouth 85.2% → 85.5%; Brentford 83.9% → 84.1%; Brighton 79.7% → 79.5%; Chelsea 94.0% → 94.3%; Coventry 63.5% → 64.1%; Crystal Palace 78.0% → 78.5%; Everton 72.8% → 73.3%; Fulham 106.0% → 106.4%; Hull 66.4% → 67.0%; Ipswich 94.1% → 95.8%; Leeds 67.0% → 67.4%; Liverpool 73.9% → 74.0%; Manchester United 58.8% → 59.0%; Newcastle 84.2% → 83.9%; Nottingham Forest 102.5% → 103.4%; Sunderland 63.2% → 63.4%; Tottenham 82.1% → 82.3%
ArsenalAston VillaBournemouthBrentfordBrightonChelseaCoventry CityCrystal PalaceEvertonFulhamHull CityIpswich TownLeedsLiverpoolManchester CityManchester UnitedNewcastleNottingham ForestSunderlandTottenham
V2BNModelThe squad cost rule defines the cost of a player's registration, gate receipts and commercial revenue in its own terms, and in all three we had used the figure from a club's accounts. We now use the rule's definitions. Every club's ratio rises, by between 0.7 and 6.0 points. We also rebuilt one input that had never been calculated, the money a club receives from the cups. One club's central estimate moves across the 85% line: Bournemouth, at 85.2%, in an assumption range of 79.8% to 87.4% that spans the line.
- What a registration costs. The rule counts everything a club pays in connection with a transfer, not the fee alone. Clubs also pay a 4% levy to the League on every fee, and agents' fees. Where clubs' accounts let us compare, the recorded cost of new signings is 11% to 17% above the reported fees. For players signed since a club's last accounts we now count the fee plus 12%, with a range of 6% to 20%.
- What counts as gate receipts. The rule counts tickets, memberships and match hospitality. One club's figure was built on more than that. Bournemouth's accounts print four lines and we had added all four, including shop sales and other income. Its base is now ticket income and hospitality, £11.4m where we had £15.6m.
- What counts as commercial revenue. The rule counts merchandise at gross profit, sales less the cost of the goods, and we had counted sales in full. We now take the cost of the goods off. Four clubs' filings give it, Chelsea's from its merchandising company's own accounts. For eleven we estimate it. For four the shops are run by a partner and nothing comes off. At Leeds, which moved its shops to a partner in July 2025, nothing comes off until we have recorded how our estimate of last season's revenue was built. For a promoted club the rule can take a three-year average of commercial revenue, and at Ipswich it does; the cost of the goods comes off each of the three seasons.
- Cup prize money. The rule counts a three-year average of what a club received from the FA Cup and the EFL Cup. We had entered a figure for each club with no calculation behind it, and every one was too high. We have rebuilt it from each club's cup ties over three seasons and the FA's published prize fund and broadcast fees. The figures are provisional, because the EFL publishes no prize money, some broadcast fees are not published, and the rule does not say whether a share of the gate at away ties counts. Each of those is in the range.
- Bournemouth is on the line. Its central estimate is 85.2%, in an assumption range of 79.8% to 87.4% that spans 85%. The estimate does not reliably tell which side it is on. Three more of Bournemouth's inputs rested on judgement, and their uncertainty is now in its range.
- Still not counted as the rule says. Conferences, concerts and other non-match income should count at net profit. No club publishes that profit, so they are still counted in full. That overstates revenue, so the ratios are if anything still slightly low.
- The ranges beside each figure now include the uncertainty in these estimates.
- Every filed figure this release newly depends on was read against its page by a second reader before release.
What moved: Fulham 102.6% → 106.0%; Nottingham Forest 97.0% → 102.5%; Ipswich 88.8% → 94.1%; Chelsea 88.0% → 94.0%; Bournemouth 81.7% → 85.2%; Newcastle 79.4% → 84.2%; Brentford 81.3% → 83.9%; Tottenham 77.6% → 82.1%; Brighton 77.4% → 79.7%; Crystal Palace 75.2% → 78.0%; Manchester City 74.2% → 76.1%; Aston Villa 72.5% → 75.7%; Liverpool 69.6% → 73.9%; Everton 70.7% → 72.8%; Leeds 64.8% → 67.0%; Hull 64.0% → 66.4%; Coventry 61.0% → 63.5%; Sunderland 61.6% → 63.2%; Arsenal 56.6% → 60.4%; Manchester United 58.1% → 58.8%
V2BMCorrectionTwo errors in the figures we published on 3 October, found the same day. Crystal Palace's ratio is 75.2%, not 74.9%: we had copied the club's filed gate receipts wrongly. Bournemouth's is 81.7%, not 82.8%: we had charged one signing twice. No club crosses a threshold.
- Crystal Palace. Our estimate of last season's matchday receipts starts from the figure in the club's 2024/25 accounts. The accounts print gate receipts of £15.6m. We had £16.6m, a copying error made when the club's record was first built. On the right figure the ratio is 75.2%, from 74.9%, and 97.1% before player trading, from 96.6%.
- Bournemouth. Adrien Truffert signed on 16 June 2025, two weeks before the club's financial year ended, so his cost is already inside the accounts we start from. We had dated him to July and charged him again as a new signing. Charged once, the ratio is 81.7%, from 82.8%, and 100.3% before player trading, from 101.6%.
- Both were found by reading every club's accounts page by page for a separate review. We have since checked all thirteen filed matchday figures used on 3 October against their pages. Eleven match; Palace's was the copying error.
- Still open at Bournemouth. The club's accounts split matchday income into four lines, and we had added all four together. Two of them, shop sales and other income, are not gate receipts as the rule defines them. We are setting one definition for all 20 clubs and it may raise Bournemouth's figure again.
- A filed figure that a release newly depends on is now read against its page before the release.
What moved: Crystal Palace 74.9% → 75.2%; Bournemouth 82.8% → 81.7%
V2BLCorrectionWe rebuilt the matchday income that the rule counts for every club. The rule takes last season's gate receipts per home match and multiplies by this season's home matches. We had not counted the matches properly and had not written down how we estimated the receipts. Both are now done: the matches are counted from archived match reports, and the receipts follow a published method. Fourteen clubs' squad cost ratios move, eight down and six up, and no club crosses a threshold.
- Home cup ties. The rule counts the average number of home FA Cup and League Cup matches over three seasons, rounded up. Twelve clubs had too few. Newcastle and Chelsea each played 13 in three seasons, so the rule counts five a season.
- Last season's home matches. The rule divides last season's receipts by every home match played. Eleven clubs' counts were wrong. Newcastle played 31 home matches, not 24; Crystal Palace 28, not 24; Chelsea 26, not 28.
- Last season's receipts. Most clubs have not filed 2025/26 accounts, so the figure is an estimate. For thirteen clubs it now starts from the last filed year, adjusted for the change in cup and European home matches and for 5% growth. Arsenal's earlier figure, raised for a European run that was no longer than the year before, is withdrawn.
- The largest movements are Tottenham, 79.4% to 77.6%, and Newcastle, 77.7% to 79.4%. At Fulham and Brentford the corrections cancel and the figure does not move.
- Manchester United's records take five small corrections to a signing date, two fees and two contract lengths, worth less than a hundredth of a point together.
- Every figure on the site is now rounded once, from the model's exact value. Three printed ranges change by a tenth of a point as a result.
- The method is set out in "How we estimate matchday revenue". A separate review of agents' fees on signings since the last accounts is under way and may raise these figures; we will report its result.
What moved: Tottenham 79.4% → 77.6%; Newcastle 77.7% → 79.4%; Crystal Palace 73.6% → 74.9%; Liverpool 70.6% → 69.6%; Everton 71.7% → 70.7%; Arsenal 55.8% → 56.6%; Chelsea 88.6% → 88.0%; Manchester United 58.7% → 58.1%; Aston Villa 71.9% → 72.5%; Brighton 77.8% → 77.4%; Nottingham Forest 96.7% → 97.0%; Bournemouth 83.0% → 82.8%; Coventry 61.2% → 61.0%; Manchester City 74.1% → 74.2%
TottenhamNewcastleCrystal PalaceLiverpoolEvertonArsenalChelseaManchester UnitedAston VillaBrightonNottingham ForestBournemouthCoventry CityManchester CityFulhamBrentfordLeeds
V2BKCorrectionA record-keeping correction with no change to any published figure. Brentford's records listed one player twice, once under his full surname and once under a shorter name, and dated his signing three years late.
- Myles Peart-Harris was also carried as "Myles Harris", the name the Premier League's squad list uses. The second entry is removed, so the League's list now matches the player who was sold to Oxford United in January 2026.
- His signing is now dated to July 2021, when Brentford bought him from Chelsea for an undisclosed fee, and his 2025 contract extension is recorded.
- None of this moves Brentford's figure, because neither entry carried a cost. Had his undisclosed fee been counted, the figure would have moved by less than a hundredth of a point.
What moved: No published figure changes.
V2BJCorrectionRecord-keeping corrections with no change to any published figure. Chelsea's records listed one player twice under two spellings of his name, with a loan that no source supports; both are removed and his actual loans are recorded. A Manchester United source is replaced with a better one.
- Chelsea carried David Datro Fofana twice, once as "David Fofana". The second entry and a 2026/27 loan recorded against it are removed, and his two 2025/26 loans, to Fatih Karagümrük and Strasbourg, are added with their sources.
- None of this moves Chelsea's figure. The duplicate was never counted in the squad cost, because Chelsea's older registrations are costed from the club's filed accounts, and the removed loan recorded no money received.
- Manchester United's record of Habeeb Ogunneye's free transfer to Arsenal, with a 40% sell-on, is now sourced to the Evening Standard rather than to United's announcement, which gives no terms.
What moved: No published figure changes.
V2BICorrectionA further check of the transfer records the previous update could not settle: contract extensions, contract lengths, signing dates and fees, each now taken from the clubs' own announcements or the most reliable reports. Ten published ratios move, by between 0.1 and 0.6 points; no club crosses the 85% or 115% threshold.
- Crystal Palace's ratio rises from 73.0% to 73.6%, mainly because Queens Park Rangers are owed a fifth of Palace's profit on Eberechi Eze, which the figures had counted as Palace's.
- Bournemouth's falls from 83.5% to 83.0%, because Illia Zabarnyi cost more than recorded (27.7m, as Dynamo Kyiv have stated).
- Everton's falls from 72.0% to 71.7%, after Beto's and Iliman Ndiaye's fees and Youssef Chermiti's basic fee were corrected.
- Contract lengths and extensions were corrected where the clubs announced them, moving Leeds to 64.8%, Brighton to 77.8% and Aston Villa to 71.9%; smaller moves follow for Fulham, Arsenal, Manchester City and Sunderland.
- Three records were wrong in kind. A Brentford player recorded as released is still under contract, an Aston Villa player recorded as bought came through the academy, and a Bournemouth signing was dated a year early.
- Where no reliable source exists yet, the figure is kept and the most it could matter is recorded.
What moved: Crystal Palace 73.0% → 73.6%; Bournemouth 83.5% → 83.0%; Everton 72.0% → 71.7%; Leeds 64.5% → 64.8%; Brighton 78.1% → 77.8%; Aston Villa 71.7% → 71.9%; Fulham 102.4% → 102.6%; Arsenal 55.7% → 55.8%; Manchester City 74.2% → 74.1%; Sunderland 61.5% → 61.6%
Crystal PalaceBournemouthEvertonLeedsBrightonAston VillaFulhamArsenalManchester CitySunderland
V2BHCorrectionA further check of the transfer records behind every figure: players recorded as costing nothing, sale fees that included add-ons, contract terms and extensions, and fees a selling club owed to a player's former club. It also corrects the date at which departed players' remaining cost is measured for clubs whose year does not end on 30 June. Thirteen published ratios move, by between 0.1 and 1.0 points; no club crosses the 85% or 115% threshold.
- Aston Villa's ratio rises from 70.8% to 71.7%, because Middlesbrough are owed a fifth of Villa's profit on Morgan Rogers, about £20m, which the figures had counted as Villa's.
- Fees owed to former clubs are now deducted on the largest sales where a reliable report states them, which also raises Newcastle (Bruno Guimarães and Alexander Isak) to 77.7%, Bournemouth (Antoine Semenyo) to 83.5% and, with a sale fee that had been recorded in euros as pounds, Brighton to 78.1%.
- Every player sold for a fee while recorded as costing nothing is now checked; those who were in fact bought are either restored to their cost or listed with the most the gap could matter.
- Sale fees are now the guaranteed fee wherever the clubs or a reliable outlet state it, with add-ons kept separate; Chelsea's ratio rises to 88.6% and Liverpool's to 70.6%, and Crystal Palace's falls to 73.0%.
- Contract terms and extensions were corrected where a reliable source dates them; smaller moves follow for Tottenham, Arsenal, Nottingham Forest, Leeds, Coventry City and Ipswich Town.
- For Arsenal, Liverpool and Coventry, whose accounts end on 31 May, the remaining cost of players who have left is now measured at that date rather than 30 June; Arsenal's ratio falls to 55.7%.
- Where no reliable source exists yet, the figure is kept and the most it could matter is recorded.
What moved: Aston Villa 70.8% → 71.7%; Brighton 77.2% → 78.1%; Newcastle 76.7% → 77.7%; Bournemouth 83.1% → 83.5%; Chelsea 88.4% → 88.6%; Liverpool 70.5% → 70.6%; Crystal Palace 73.1% → 73.0%; Nottingham Forest 96.9% → 96.7%; Tottenham 79.3% → 79.4%; Arsenal 55.8% → 55.7%; Leeds 64.6% → 64.5%; Coventry City 61.3% → 61.2%; Ipswich Town 88.9% → 88.8%
Aston VillaBrightonNewcastleBournemouthChelseaLiverpoolCrystal PalaceNottingham ForestTottenhamArsenalLeedsCoventry CityIpswich Town
V2BGCorrectionSeveral players bought for a fee had been recorded as if they had cost nothing, and some transfers between two clubs in the series were recorded differently by each. Each has been checked against the clubs' announcements and the most reliable reports; six published ratios move, by between 0.1 and 1.3 points.
- Tottenham had recorded eight players it bought, including Brennan Johnson, Cristian Romero and Yves Bissouma, as academy players at no cost, which overstated its profit on their sales; its ratio falls from 80.6% to 79.3%, and from 92.5% to 88.5% before player trading.
- Bournemouth's ratio rises from 82.2% to 83.1%, because Milos Kerkez left before the June 2025 accounts were drawn up and his cost should never have been taken out of the squad's book value.
- Liverpool's ratio falls from 71.2% to 70.5%. Ibrahima Konaté had been recorded at no cost, and the remaining cost of Diogo Jota's registration is now removed from the squad's book value, as the series always intended, without treating it as a sale.
- Aston Villa had recorded Lucas Digne at no cost; its ratio falls from 71.0% to 70.8%.
- Where the selling and buying clubs' records of the same transfer disagreed, both now use the basic fee as stated by the club or the most reliable report, and the completion date the clubs announced; Everton's ratio falls to 72.0% and Crystal Palace's to 73.1%. The Dwight McNeil and Brennan Johnson exchange had no published value, so Crystal Palace's figure could be up to 73.3% and Everton's up to 72.2%.
- Every remaining difference between two clubs' records of a transfer is now listed and checked automatically.
- No club crosses the 85% or 115% threshold.
What moved: Tottenham 80.6% → 79.3%; Bournemouth 82.2% → 83.1%; Liverpool 71.2% → 70.5%; Everton 72.3% → 72.0%; Aston Villa 71.0% → 70.8%; Crystal Palace 73.2% → 73.1%
TottenhamBournemouthLiverpoolEvertonAston VillaCrystal Palace
V2BFPublishing gateThe tables of past seasons in the Newcastle and Aston Villa articles, and every club's table of player-sale profit by season, are now checked against the filed accounts automatically. No current-season figure changes.
- Each past season's row is now rebuilt from the filed accounts and the FA's published payments to agents for the matching period, and a row that cannot be rebuilt is marked unavailable rather than estimated.
- The check found that Newcastle's and Aston Villa's tables of past seasons do not match the filings, including one amortisation figure that no filing contains.
- Every input to Newcastle's figure now records where it comes from, whether a filed page, a regulator's publication, the club, the press or a modelling assumption, as the first club of twenty.
V2BFCorrectionAgents' fees now come from the Football Association's published figures for every club. Thirteen clubs had carried estimates although the FA had published the actual payments in April; with the published figures, thirteen ratios move, by between 0.1 and 2.5 points.
- Nottingham Forest paid agents £12.2m, not the £22m estimated, and its ratio falls by 2.5 points; Bournemouth paid £20.9m, not £15m, and its ratio rises by 1.4 points.
- Manchester City, Tottenham, Arsenal and Everton paid less than estimated and their ratios fall by about a point; Chelsea, Aston Villa, Crystal Palace, Fulham, Brighton and Brentford paid more and theirs rise slightly; Liverpool's falls slightly.
- Newcastle and Manchester United now carry the FA's exact figures rather than rounded ones; their published ratios do not change.
- No club crosses the 85% or 115% threshold.
- Any input still waiting for a source now carries the date that source is expected, and the checks fail once it passes.
What moved: Nottingham Forest 99.4% → 96.9%; Bournemouth 80.8% → 82.2%; Manchester City 75.4% → 74.2%; Tottenham 81.7% → 80.6%; Arsenal 56.8% → 55.8%; Everton 73.3% → 72.3%; Chelsea 87.6% → 88.4%; Aston Villa 70.4% → 71.0%; Crystal Palace 72.8% → 73.2%; Fulham 102.0% → 102.4%; Liverpool 71.5% → 71.2%; Brighton 76.9% → 77.2%; Brentford 81.2% → 81.3%
Nottingham ForestBournemouthManchester CityTottenhamArsenalEvertonChelseaAston VillaCrystal PalaceFulhamLiverpoolBrightonBrentford
V2BECorrectionEvery club's turnover, staff costs, profit on player sales and squad book value has been checked line by line against the filed accounts it was taken from. Five clubs had figures that were rounded, estimated or taken from a third party; they now carry the exact filed figures, and two published ratios move.
- Newcastle's staff costs were taken from an analyst's reading of the accounts, £2.2m above the filed £241.3m; with the filed figure its ratio falls from 77.1% to 76.7%.
- Aston Villa's squad book value had been estimated where the accounts give it exactly; with the filed figure its amortisation falls by about £0.5m and its ratio from 70.5% to 70.4%.
- Arsenal, Brighton and Liverpool carried rounded sale profits or book values; the corrections move their ratios by a thousandth of a point or less.
- Three clubs' company numbers pointed to unrelated companies and are corrected; company numbers and filed figures are now checked automatically against the source records.
What moved: Newcastle 77.1% → 76.7%; Aston Villa 70.5% → 70.4%; other figures unchanged at published precision
V2BDCorrectionAston Villa's wage estimate now starts from the exact staff cost in its filed accounts, £273.4m including social security, instead of a rounded £273.0m. Estimated wages rise by £0.3m and the published ratio is unchanged at 70.5%.
- The filed staff cost is £273.4m (273,376 thousand) and includes £33.8m of employer social security; an earlier note in the model said social security was not included, which was wrong.
- Estimated squad wages rise from £218.8m to £219.1m, and the squad cost before player trading moves from 89.4% to 89.5% of revenue.
- A new check now requires the wage input and the accounts record to carry the same filed staff cost at every club.
- The wage roll-forward comparison published earlier today showed Villa's base at £219.8m; with this correction it is £220.1m, the same as on 20 September.
What moved: Aston Villa 70.47% → 70.53%; unchanged at 70.5% as published
V2BCRule readingThe wage roll-forward comparison is remeasured on the current baseline. The fifteen clubs that roll their filed payroll forward range from −5.6% to 18.9% of the base it is added to, and the other five are built player by player and have no roll-forward. No published figure changes.
- Fourteen of the fifteen rows are unchanged since 20 September, including Tottenham's corrected 14.0%.
- Aston Villa's base is now £219.8m rather than £220.1m, on the staff cost as reported; its roll-forward and share are unchanged.
- Manchester United remains the only club above the range, at 18.9%, and the explanation (the reversal of a pay reduction linked to the Champions League) still matches the figures.
- Coventry City, Hull City, Ipswich Town, Leeds and Sunderland are listed as built player by player and sit outside the comparison.
- The comparison is still a review step rather than an automated gate.
V2BCRule readingThe disposal-basis analysis is remeasured on the current baseline of 20 clubs. If the five-year cap also set the book value used for profit on a sale, eight registrations at three clubs would carry £12.0m of extra profit (£9.0m on the four whose contract terms are sourced), and the largest move would be Chelsea's, half a point. No published figure changes.
- Nicolas Jackson is added and is now the largest case, because his 2024 extension to 2033 brings a contract signed before the cut-off within the cap.
- João Félix, Liam Delap, Savinho, Renato Veiga and Mathis Amougou are remeasured on corrected signing, sale and extension dates, and each still qualifies.
- Four registrations (Omari Kellyman, Renato Veiga, Morgan Rogers and Mathis Amougou) rest on estimated contract lengths and are now marked as estimated.
- None of the six clubs added since 20 September has a qualifying sale.
- No club changes compliance state. Chelsea's 87.6% would be 87.1%, and Manchester City and Aston Villa each move by 0.03 points.
V2BCCorrectionFour new structural gates, and the corrections they forced. The largest: Chelsea’s figure falls from 101.5% to 87.6%, because ten registrations already inside the filed book value were also being charged individually. No new football or accounting events: the data cutoff is unchanged.
- No double amortisation: a registration inside the filed closing book value that is also charged player by player must now be declared and taken out of the pooled balance, or the run refuses. This is the gate that found Chelsea’s ten double-counted registrations.
- A closed list of registration categories. Invented category labels had hidden two Tottenham loan wages, Marmoush and Mudryk, about £20m, from both the loan ledger and the wage bridge.
- A player who has left leaves the pooled book value by the date he was acquired, whatever his label (Reijnders at Manchester City).
- Signings dated on the first of a month just after a year end must be evidenced. About 36 such rows across 14 clubs were sourced or redated to their real dates.
- A fix found by the new tests: the extension audit looked for one specific category name, and a relabel had silently switched it off at Chelsea. It now covers every grandfathered registration.
- Liverpool: Frimpong and Mamardashvili treated as signed after the 31 May 2025 year end, on the announcement evidence. The figure stays at 71.5%, about 69.7% if both were inside the book value; reopened when the 2025/26 accounts are filed.
What moved: Chelsea 101.5% → 87.6%; Manchester City 79.1% → 75.4%; Tottenham 79.5% → 81.7%; Brentford 84.5% → 81.2%; Brighton 77.8% → 76.9%; Bournemouth 80.6% → 80.8%; Leeds 64.5% → 64.6%. Every other club unchanged.
ChelseaManchester CityTottenhamBrightonBrentfordBournemouthLeedsLiverpool
V2BBModelGroundwork for testing forecasts in public: a club’s forecast can now be updated with new evidence beside the frozen original, never over it; and loan wage savings become sourced data rather than hand-set entries. No published figure moved.
- Dated information updates sit beside a frozen forecast. The first use is Manchester United’s 2025/26 results, on the track record.
- Loan wage relief is now built from a sourced record of every loan spell: who paid what share of which player’s wages, and for how long. For United this shows the September forecast’s hand-set loan entries understated wages: on the corrected method the September information gives 59.6%, not 58.7%. The published 58.7% stays as the forecast of record, and the correction is shown on the track record.
- Found and recorded, not yet applied: departed players’ book value was measured at 30 June 2025 at every club, rather than at each club’s own year end. Correcting it would move Arsenal from 56.8% to 56.6%, and three other clubs by less than the published precision.
V2BBCorrectionTwo Manchester City sale fees taken to their reported base rather than an estimate, on both clubs’ ledgers.
- McAtee to Nottingham Forest: an estimated £25m (Forest’s rejected opening bid) replaced by the reported £22m base, on City’s sale and Forest’s purchase alike.
- Heskey to Köln: £7m, which included add-ons, replaced by the £2.1m base.
- Two duplicate placeholder rows removed from City’s squad census.
What moved: Manchester City 78.9% → 79.1%; Nottingham Forest 99.6% → 99.4%.
V2BACorrectionBournemouth’s sale of Antoine Semenyo recorded at the reported £62.5m base fee, agreeing with the buying club’s ledger.
- The profit on the sale falls slightly, and with it the three-year average of player-sale profit.
What moved: Bournemouth 80.5% → 80.6%.
V2AZPublishing gateEvery club must now state its financial year end, and it must agree with its filed accounts. Players who leave carry their cost out of the books by a single, tested route.
- Liverpool’s files stated no year end, so a check fell back silently to 30 June and misread three June 2025 departures against accounts that close on 31 May. Every club is now checked, and a missing or contradicting year end refuses.
- Departed players’ remaining cost leaves the books consistently, including write-offs, at Brentford and Aston Villa.
What moved: Aston Villa 70.4% → 70.5%; Brentford unchanged at 84.5%.
V2AF–V2AYPublishing gateEvery club’s departures are now declared from sources, window by window, and reconciled against its ledger in both directions. All 20 clubs were migrated to the new check, and it is enforced: a departure missing on either side refuses the run.
- A declared departure with no matching sale in the ledger refuses; a sale in the ledger missing from the declaration refuses. Dates that differ by more than a week, or fall in a different financial year, refuse.
- Free transfers, releases, retirements and contract expiries are exits, not sales, and are never counted as departures.
- Migrating each club from its sources found missing sales, wrong dates and misclassified exits at several clubs. Each was corrected under its own record before the club was declared.
What moved: Chelsea 103.1% → 101.5%; Fulham 101.3% → 102.0%; Nottingham Forest 99.5% → 99.6% (missing sales, sale dates and the Jota Silva sale); Newcastle 76.7% → 77.1%; Brighton 77.6% → 77.8%; Brentford 84.2% → 84.5%; Everton 73.5% → 73.3%; Manchester United 58.6% → 58.7%. Others unchanged at published precision.
ArsenalAston VillaBournemouthBrentfordBrightonChelseaCoventry CityCrystal PalaceEvertonFulhamHull CityIpswich TownLeedsLiverpoolManchester CityManchester UnitedNewcastleNottingham ForestSunderlandTottenham
V2AD–V2AEPublishing gateA new cross-club check: no player may be current at two clubs at once. Its first run across the 20 ledgers found four contradictions, all corrected before the check was switched on.
- Tynan Thompson was still current at Tottenham while Manchester United recorded him as a permanent arrival: a Tottenham sale was missing.
- Dominic Calvert-Lewin (Everton and Leeds), and Frankie Runham and Christopher Atherton (Chelsea and Ipswich) were each on two books.
- Names are matched across spellings and accents (Nørgaard and Norgaard are one player), and genuine namesakes are declared pair by pair, never excused everywhere.
V2Y–V2ACCorrectionA day of frozen-ledger corrections, each signed off separately, most found by checking one club’s ledger against another’s.
- Nottingham Forest: 2025/26 sales corrected.
- Tottenham: the Højbjerg transaction corrected.
- Arsenal: Nørgaard and Vieira recorded as departed, where the ledger still had them at the club.
- Crystal Palace: academy sales corrected, including removal of a sale that had not happened.
- Fulham: the dates of Diop’s and Lukić’s sales to Ipswich corrected from the club’s own announcements. An earlier sale date means more book value at sale and slightly less profit.
What moved: Nottingham Forest 105.1% → 99.5%; Arsenal 58.0% → 56.8%; Tottenham 78.8% → 79.5%; Crystal Palace 73.2% → 72.8%; Fulham unchanged at 101.3% (range 98.2 to 104.7%).
V2XCorrectionHarry Clarke’s undisclosed sale to Preston taken from a £0.5m estimate to the reported £3m.
- Nil book value at sale, so the whole fee is profit, which raises the three-year average.
What moved: Ipswich Town 89.3% → 88.9%.
V2WClub addedIpswich Town added: the third promoted club, and the first case of the rule’s three-year average of commercial income, which applies because Ipswich were a Premier League club within the last three seasons.
- Wages built player by player, as for the other promoted clubs, with the filed Championship-era payroll kept as a check.
- Every single-input test of the uncertainty stays above 85%.
What moved: Ipswich Town enter at 89.3% (range 83.2 to 95.4%).
V2VClub addedHull City added: the second promoted club. The £15m commercial floor binds, worth about 3.7 points.
- Agents’ fees taken from the FA’s published 2025/26 figure, the series standard; the understatement this may carry is disclosed in the article.
What moved: Hull City enter at 64.0% (range 58.9 to 69.2%).
V2UCorrectionFrank Onyeka’s sale corrected: to Coventry City on 29 June 2026, not Ipswich Town in July, from the buying club’s announcement.
- The earlier date puts the sale in 2025/26 rather than 2026/27. The fee is unchanged.
What moved: Brentford unchanged at 84.2%; the lower end of the range moves from 80.8% to 80.7%.
V2TClub addedCoventry City added: the first promoted club, and the first live test of the promoted-club revenue rules.
- The £15m commercial floor binds over the club’s own estimate; wages are built player by player.
What moved: Coventry City enter at 61.3% (range 56.7 to 65.9%).
V2SRule readingThe rule’s special revenue treatment for promoted clubs built in from the Handbook text, before any promoted club was added. No published figure moved.
- A promoted club is credited with the merit payment for 15th place, a gate figure of its Championship average plus 25%, and the higher of its commercial income, its three-year average where it applies, and £15m.
- The £15m commercial floor now applies to every club. None of the existing clubs is near it.
- A new gate refuses a club whose status contradicts its league position last season.
V2SClub addedSunderland added, and a full reconciliation against the 18 September baseline: every movement since then explained and recorded here, and one missing Newcastle sale added.
- Newcastle: Sean Longstaff’s £12m sale to Leeds United in July 2025 was missing from the ledger, found by the Leeds rebuild’s cross-club check.
- Movements since 18 September, each already corrected in its article with a dated note, recorded here: Manchester City 82.1% → 78.9% (players’ share of wages widened after a player-by-player estimate came in lower, and two contract extensions found); Everton 72.4% → 73.5% (fee sweep, and two four-year contracts from the club’s own announcements); Chelsea 104.0% → 103.1% (extension audit).
- A stale register of claims made in the Forest article was cleared, and one remaining superlative removed.
What moved: Sunderland enter at 61.5% (range 57.4 to 65.5%); Newcastle 77.3% → 76.7%.
V2RModelA second way to build wages, for clubs whose last filed accounts describe a Championship season: the 2026/27 squad’s wages are built player by player from sources, with the filed payroll kept as a check. No published figure moved.
- For these clubs, rolling the filed payroll forward would have meant a change of 30% or more, mostly assumption.
- One wage function now serves the central estimate, the stress case and the simulation, so a second method cannot be half-applied.
V2RClub addedManchester United added as the fifteenth club, at 58.6% under the rule and 61.9% before player trading. Engine code unchanged.
- The club company, Manchester United Football Club Limited, is used for both costs and revenue. Its 82 players match the men’s football operation, while Red Football Limited, the consolidating group, reports 133 including the women’s team. The group’s wider commercial income is published as a scope sensitivity, about 54.2%, and not blended into the figure.
- The wage roll-forward is about 19% of base, above the range at every other club, and is explained rather than accepted: the filed year was a season without the Champions League, and about £30m of the £37.5m is the reported reversal of a Champions League-linked pay reduction.
- The population was closed against the September 2025 and September 2026 Premier League squad returns. The two February returns are still to be encoded. The returns found one departure missing from the club-sourced transfer record.
- The FA’s published agents’ figure, £31.77m, replaced a £35m estimate before publication and moved the figure from 58.9% to 58.6%.
- The published figure, 58.6487%, sits close to a rounding boundary. It is published as computed, and the article names the two small inputs that would each print 58.7%: one contract length resting on consistent reporting rather than a club statement, and two undisclosed academy fees.
- Cross-ledger checks against six other clubs found two disagreements: Andrey Santos is carried at £48m on 13 July 2026 here and £47m on 25 August at Chelsea, and two academy sales to Arsenal carry estimated fees here and nil there. The Chelsea row is authorised for correction through the frozen-ledger process.
What moved: Manchester United enter at 58.6% (range 56.0 to 61.4%), 61.9% before player trading.
V2RCorrectionEvery transfer fee in the series checked against a source that publishes the base and the add-ons separately, and the squad-share comparison rebuilt on a like-for-like basis. Engine code unchanged; the data moved.
- Twenty-two fees were stored at the total rather than the base, across eight clubs of fourteen: Bournemouth 5, Manchester City 4, Chelsea 3, Crystal Palace 3, Newcastle 3, Everton 2, Arsenal 1, Aston Villa 1. Six clubs were clean, and all six were built from detailed window records rather than season summaries.
- The largest proportional case was a deal where 70% of the reported fee was contingent: £3m guaranteed against a £10m reported package.
- One earlier correction had run the wrong way, moving a base to a total against a club announcement of a record fee. Both sides of that transaction now move together.
- A loan fee included inside a reported transfer fee was being capitalised. A loan fee is an expense of the season it is paid.
- The independent per-player wage build was being compared against the wrong base. It is a 2026/27 squad snapshot, while the share here is taken of an FY25 filed cost and then rolled forward, so dividing one by the other gives the share plus the roll-forward. Netting it off reverses two of four conclusions: Arsenal needs no recut, and Tottenham is outside its range on the low side rather than inside it.
- Manchester City’s squad-share range widened from 80 to 88% to 70 to 88%, reaching the independent build without moving the centre onto it, because the two figures allocate the same filed total by different methods and the disagreement bounds the range rather than relocating it.
- The range widening stands. Its evidential basis has since been narrowed: a per-head comparison cited in support of it is no longer treated as valid evidence, because City’s filed football headcount combines players with football staff and is not a player-only denominator.
- Two of this series’ own ledgers disagreed on the same transfer, £48m against £47m. Nothing in the engine compares a purchase row with the counterparty’s disposal row, so a cross-ledger gate is the outstanding build.
What moved: The fee sweep moved eight club ratios by between −1.3 and +0.13 points, with no compliance-state change. Manchester City moved separately from 82.1% to 79.3%, mainly because the squad-share range widened from 80 to 88% to 70 to 88%.
ArsenalAston VillaBournemouthChelseaCrystal PalaceEvertonManchester CityNewcastleTottenham
V2O to V2QModelA gap in the version history. The engine is stamped V2R and its own changelog stops at V2N, so three intermediate versions have no entry.
- Tottenham was added as the fourteenth club and its wage roll-forward was corrected from 36% of base to 14%, against a series range of −6% to +13%. That moved the club from 86.4% to 78.8% and reversed the threshold conclusion of a published article. A quantity that is never compared with anything is not validated by the care taken over its inputs.
- The temporary-transfer treatment was requantified after the bridge correction, giving 78.8% at Tottenham and 79.3% at Manchester City on the adopted reading.
- The remaining changes in this range are recorded in the engine repository and have not yet been written up here. They are reconstructed from the repository rather than from memory, which is why this entry names the gap instead of filling it.
V2NDataEvery wage roll-forward in the series now covers every completed window. Six clubs of twelve had been pricing squads as they stood before windows that had already happened.
- The wage input is a share of a filed staff-cost line carried forward through the windows closed since, and six clubs carried an incomplete one.
- The baseline and all twelve articles were brought into agreement in the same pass.
What moved: Arsenal 56.7% → 58.1%, Everton 73.0% → 72.6%, Brentford 84.0% → 84.2%, Fulham 99.6% → 101.3%, Nottingham Forest 106.3% → 105.1%
V2KBaselineChelsea republished with nine recovered senior registrations and a complete wage roll-forward. Twelve clubs published, none withheld.
- Population proved at 143 census identities with none unresolved against a 148-row ledger, after Caicedo, Wesley Fofana, Lavia, Neto, Palmer, Gusto, Slonina, Reece James and Colwill were recovered.
- Completing the roll-forward LOWERED the ratio, which contradicted the expectation stated when the control was written. Both missing windows ran the squad down rather than up.
What moved: Chelsea to 103.3% / 129.5%, Arsenal 56.7% → 60.2%
V2IPublishing gateChelsea’s population proved, and the comparison that proved it found six other clubs carrying an incomplete wage roll-forward.
- Nine senior registrations were missing rather than three, including Wesley Fofana at £70m.
- Fofana was hidden by the strongest false match in the series: the resolver offered David Datro Fofana, a different Chelsea player with the same surname on the same list.
What moved: Chelsea rebuilt to about 107.5%
V2HDataTwo academy registrations recovered. The ledger now carries 147 registrations against 117 census identities with none unresolved.
- Both were at nil acquisition cost, present on three returns and absent from the fourth, one leaving at contract expiry with no fee and no disposal profit.
What moved: no change to the published figure
V2FPublishing gatePopulation closure became a publication gate. Nine clubs were carrying published arithmetic on a population that had not been proved.
- A transaction declaration closes window MOVEMENTS. A registration census closes POPULATION. Both are required and they are separate controls.
- Chelsea settled it: every window declaration could have been correct and the ledger would still have been missing Caicedo, Palmer and Neto, about £211.5m of acquisition cost.
- The gate was advisory for one build, which was the wrong call. Leaving a control in place while publication ignores it repeats the pattern the control exists to stop.
V2EPublishing gateChelsea unpublished. Three major registrations were absent from a 119-registration reconstruction while sitting on the League’s own submitted return.
- Caicedo, Palmer and Neto, about £211.5m of fees and £27.1m a year of unrecognised amortisation.
- Their wages already sat inside the staff-cost share, so the numerator moved on amortisation alone.
What moved: 95.2% → about 99.8% on amortisation alone
V2CRule readingA loan carrying an unconditional obligation to buy was tagged as conditional with the condition unresolved, which contradicted the engine’s own stated rule.
- The Guidance to Rule A.1.22 leaves amortisation with the lending club during a temporary transfer, so City retain the registration and the £11.8m annual charge and the £50m does not enter the 2026/27 disposal pool.
- The treatment is a matter of timing. The registration moves in 2027/28, when City lose the charge and recognise the disposal.
- The article had carried the alternative as a sensitivity rather than as the answer.
What moved: 82.1% → 79.9% at the time, and Tottenham 88.0% → 86.4%
V2BRule readingRule A.1.29(e) was applied inconsistently across the league. Two structural faults, both in the same provision.
- The round-of-16 reference is a property of the competition rather than the club, and had been entered club by club. Arsenal carried £75m for the Champions League while Manchester City and Aston Villa carried zero for the same competition, turning a three-way maximum into a two-way one.
- The first two amounts count only seasons in the competition qualified for, so a club moving between competitions cannot carry the old one’s money into the new one.
What moved: Aston Villa 74.0% → 70.3%, Manchester City republished at 82.1%
V2ADataRebuilt from the filed annual report rather than from secondary reporting. The club had been published on a wage base that was wrong, and the first correction was also wrong.
- The figure carried as total staff costs was employee costs plus the £169.5m amortisation charge, so a squad share was taken of a base that already contained amortisation, and amortisation was then added again.
- The second attempt treated the filed total as a wage line and added employer costs twice.
- The filed total employee cost in Note 7 is £408.4m: wages £355.4m, social security £51.1m, pension £1.4m, share-based payments £0.5m, across 670 employees of whom 252 are football staff.
- Build from the filed document. A summary written by somebody else may define its terms differently and will not say so.
What moved: 88.4% → 83.2% → 82.1% regulatory
v1zPublishing gateManchester City unpublished on a contaminated wage base.
- The tell was in the engine’s own data throughout. Employer social security and pension run at 12.5 to 15.7% of the wage line at every club in the series, and the contaminated base implied 41.7%. Nobody had read across the column.
- That comparison is now the gate’s logic, so it runs on every club on every build rather than depending on somebody thinking to look.
What moved: 88.4% → 83.2%
v1yClub addedTwelfth club published, built from the registrations roll-forward in Note 12, which reconciles on all three identities so the residual is forced rather than estimated.
- The whole summer 2025 senior departure side was missing: De Bruyne, Gündoğan, Ederson, Carson, McAtee. A source that misses a departure leaves no trace.
- Twelve apparent arrivals were new scholarship registrations. Read as signings they invent twelve moves into one window.
- Three registrations appeared on the senior 25 for the first time without being signed, having aged out of Under-21 eligibility.
What moved: published at 88.4% / 103.4%
v1xPublishing gateA departure dated before a squad list the player appears on is wrong, and it had happened at three clubs. Prompted by asking whether an error found at one club also sat in the eleven already published.
- Ninety-two departures carry no disposal row. Most are legitimate releases and expiries; nine had no stated reason at all.
- Six departures mention a loan in their note. Five were genuine loan-then-sale. One was not.
v1wPublishing gatePublished after the missing September 2025 squad list was obtained, which dissolved six apparent recall-and-re-loan chains.
- All three windows close by registration-census delta, each bracketed by its own endpoints rather than by a twelve-month span.
- Swinkels, Özcan, Pavey, Sam Lewis, Edwards and McWilliams had each looked like a recall and re-loan across the February-to-February gap. With the September list between them, every one holds at most one movement per window. They were an artefact of the missing endpoint.
- The source remained marked insufficient for closure and Villa passed anyway, which is the v1v design working: a degraded source is barred from being the closure authority, not from being evidence.
- One registration is still unexplained and recorded as such: signed in January 2026, absent from the February list, present in September.
What moved: 74.1% → 73.9%, and 145 registrations carried against 89 when the census work began
v1vPublishing gateScope and source authority became separate conditions. A declaration can be genuine, reconciled and cross-checked, and still rest on a class of source already caught omitting a deadline-day disposal.
- The argument for publishing failed on one question: how many agreeing rows is enough? There is no principled threshold.
- A cross-check answers whether a NAMED transaction is right. Completeness is a claim about the POPULATION, and no quantity of the first establishes the second.
- The risk sat in 58 academy and loan movements, the category a season record drops while getting every headline transfer right.
- A dataset that passed under v1u can refuse under v1v on identical inputs, which is precisely what a version string exists to distinguish.
What moved: no published figure moved; Aston Villa withheld pending census closure
v1uCorrectionA disposal corrected to the club’s own base figure, against which £17.5m of add-ons sit. No engine behaviour changed, and the version moved anyway.
- The snapshot tool refused to write a new baseline under an unchanged version, and it was right to refuse. If the baseline changes while the version does not, that string names two different baselines.
- The data cutoff field distinguishes the REASON a figure moved. It does not licence two baselines under one version.
What moved: 95.3% → 95.2%
v1tPublishing gateTenth club published, and the first closed by registration-census delta rather than by a transaction declaration. Two conventional complete transfer sources each failed.
- Chelsea’s own summer 2025 census omits a deadline-day arrival, and the 2026/27 season record omits a £125m deadline-day disposal. A source caught twice is failed, not incomplete.
- Fifty-four of 108 endpoint identities were not in the ledger, most consequentially the entire summer 2025 departure side.
- A disposal was attributed to the wrong registration: £20m recorded against a player who was never sold. Every aggregate reconciled and only the identity was wrong.
- The direction inverts the programme’s pattern. Every previous completeness failure understated a club by omitting a cost; this one overstated it by carrying amortisation for departed players and missing their disposal profit.
What moved: 101.7% → 95.3% regulatory, 123.2% → 119.3% underlying
v1sPublishing gateNinth club published. The whole 2025 cohort was missing from the ledger, so every signing from that window was absorbed into the residual rather than amortised on its own fee.
- Amortisation rises from £141.7m to £201.0m once the 2025 cohort exists.
- The base-versus-inclusive fee pattern at its largest scale: Gyökeres spans £14m across public sources, £55.0m on the club’s own table against £64m, £69m and €63.5m elsewhere. Konsa settles what the pattern is, reported at £55m against the club’s £51.0m plus £4.0m in add-ons. 55 is exactly 51 plus 4.
- Conversions initiated by the OTHER side: Hein and Kiwior both left permanently because their loan clubs exercised options. Every conversion found before was a club buying a player it had borrowed.
- Two transactions no window census can hold: Martinelli’s £60m sale three days after the English deadline, because the Saudi window runs later, and Dudziak’s youth loan agreed in December and ended inside January.
- Third cross-club validation: Zinchenko’s loan out and recall match Forest’s ledger on both dates.
- Both windows closed at full_club under Gate 18, the gate Arsenal itself caused. Test 1l was rewritten to construct the condition rather than depend on Arsenal’s state, because closing Arsenal broke a fixture built around it.
What moved: 49.4% → 56.7% regulatory, 52.8% → 60.4% underlying
v1rPublishing gateA declaration can be genuine, source-backed and internally reconciled, and still certify only part of the club.
- Arsenal’s summer 2025 declaration came from the club’s own published census, post-window, and reconciled exactly, against the MEN’S TEAM only. The same census lists four academy arrivals and sixteen academy departures that were not in the ledger.
- The counts matched because they were compared against the rows that were there. A ledger can be internally perfect while the universe it describes is incomplete.
- Gate 18: a declaration now carries a SCOPE as well as counts, being full_club, senior_only, academy_only or category_specific. Absent, scope is full_club, which is what every frozen club declared in substance, so none is affected. Anything narrower blocks publication however well it reconciles.
- Test 1l verifies both halves: a narrowed scope refuses, and full-census clubs are unaffected.
What moved: no published estimate changed. Arsenal moves from the table to the withheld list
v1qRule readingA component error that could not reach an output, and therefore could not be caught by any test on a published ratio.
- A loan-out row used the SIX-year accounting amortisation charge inside the temporary-transfer waterfall instead of the FIVE-year charge A.1.22(c) requires for the ratio.
- Output-invisible: the loan receipt was fully absorbed against the player’s amortisation under either value, so the surplus to the transfer amount was nil either way and no headline moved.
- Test 1k asserts the INTERMEDIATE component directly, being the rule amortisation, the amount the waterfall applies to it, and the nil surplus, rather than the final ratio. Verified to fail on the original value.
- The general lesson: a suite that tests only published outputs is blind to an error inside a downstream-absorbed component. Where the rule and the accounts prescribe different treatments of the same figure, the component itself has to be asserted.
What moved: no published estimate changed. The version bump records a behaviour change rather than a figure change
v1pPublishing gateNinth and final club of the completeness audit, and the largest correction in the programme, not from one bad fee but because the transfer census was structurally incomplete.
- Test 1j: a conditional obligation that FAILS must capitalise nothing. The third and last of the three obligation states to get a fixture, and the mirror of the two that fire.
- Six missing arrivals, a miscohorted player, two unrecorded loans in, and five fees corrected against the club’s own squad table.
- A registration bought and immediately loaned to a sister club, absent from every squad-based source because the player never appeared.
- A loan out that had already converted to a permanent sale while the season record still described a loan.
- Source hierarchy extended: a season article is a presence source, never a completeness source, and at a multi-club group it fails in both directions.
What moved: 98.6% → 106.3% regulatory, 121.4% → 130.4% underlying
v1oPublishing gateEvery window rebuilt from the club’s own site with no aggregator in the chain, after no retrievable season record was found. The headline moved a point and almost every input beneath it changed.
- The five-role source hierarchy, being declaration, post-window category closure, backwards oracle, presence and transaction, was derived here and is now applied to every club.
- `loan_ended` made a first-class cohort: referenced by the conversion gate but previously never excluded from the amortisation sweep or the remuneration check.
- Academy and B-team reconciliation made a requirement, after the club’s B team defeated an all-clubs transfer record three separate times in three different ways.
- The base-versus-inclusive fee pattern found running in the opposite direction on a disposal, where correcting it raises rather than lowers the ratio.
- Methodology correction: Academy Player status is a registration category, and a professional contract does not by itself end it.
What moved: 83.1% → 84.0% regulatory, 106.4% → 107.0% underlying
v1nCorrectionA contract term that was internally consistent, passed every check, and contradicted its own source.
- Gate 17: a summer signing’s nominal contract term must not be fractional where the source reports a whole-year deal.
- Test 1i as the regression fixture.
- The January convention was deliberately left unasserted. Whether a January five-year deal is carried as 5 or 5.5 is a definition this engine has not settled, and asserting it would have failed four clubs on a matter of definition rather than fact.
What moved: 77.6% → 77.2% regulatory, 98.6% → 98.1% underlying
v1mPublishing gateThe first club whose position improved on audit, and the first loan OUT to matter more than any loan in.
- Loan-out wage relief modelled through a named bridge entry where the player’s wages were never in the filed base.
- The A.1.29(g) waterfall exercised on a player whose loan fee offsets against his own amortisation with nothing reaching the transfer amount.
- A widely reported fee found to be the club’s own figure plus contingent add-ons, the fourth such case, and one quoted inside the reporting of the player’s own loan.
What moved: 80.6% → 77.6% regulatory, 102.4% → 98.6% underlying
v1lPublishing gateThe first observable conditional-obligation trigger in the series: a loan the club called an option, reported elsewhere as an obligation fired on a debut appearance.
- Conditional obligations modelled as a branch, with the triggered treatment published and the loan-only counterfactual stated alongside it.
- A third loan-to-permanent conversion found, confirming that permanent registrations appearing between windows are a recurring class rather than an anomaly.
- Test 4 decoupled from live club data after correcting one club’s transfer broke an unrelated selection test.
v1kPublishing gateArrival fees found to be overstated, the first club where the errors ran in the opposite direction to every previous audit.
- A capitalised-obligation bug fixed: moving a player out of the loan cohort to capitalise his obligation silently deleted his remuneration. Test 1h.
- A related-party loan with a conditional obligation from a club part-owned by the same person, with the salary not sourceable and bounded across a five-fold range by declared judgement.
- Year-end straddle confirmed as an expected condition rather than an exception: arrivals completing before 30 June belong in the prior cohort even though the window is the same.
v1jBaselineThe first club with a filed disposal-profit figure to anchor against, and the first demonstration that the anchor changes the answer.
- Filed-accounts override: where a reconstructed ledger and an applicable filed disclosure disagree, the divergence is recorded and the filed figure governs the overlapping period.
- Components made deterministic so the published table reconciles to the published ratio rather than to a simulation mean.
- The concentration metric corrected to use a signed rather than absolute pool.
earlierPublishing gateThe gates that make the later audits possible, added as each failure exposed the need for one.
- Window completeness: a club is not publishable until every completed window is declared and its counts reconcile to the ledger.
- Source dating: a declaration source must have been published after the window it declares actually closed. One club nearly passed on a round-up last updated five days before deadline day, and the counts matched because the source and the ledger shared the same omission.
- Duplicate and conversion detection, for players appearing in more than one cohort as a loan becomes permanent.
- Variance attribution corrected, and the engine version stamped on every published figure.
How each forecast has held up against later evidence is on the track record. The method as it stands today is on the Engine page.