The Orange Engine · Technical notes →

The wage roll-forward

A carefully built estimate can still be wrong if its result is never compared with the same quantity elsewhere. This is the case that established that at Tottenham, what it cost, and the control that followed.

What a roll-forward is

No club in the series discloses its player wage bill separately. The accounts give one staff-cost line covering everyone the club employs, so the playing payroll is estimated in two independent steps:

1. The share: what proportion of that line is players, rather than stadium, retail, administrative and coaching staff.
2. The roll-forward: how that figure moves between the filed year and the season being tested, as players arrive, leave and re-sign.

At most clubs the roll-forward is small: arrivals and departures largely offset, and pay rises for players who stay are single-digit percentages. 5 clubs (Coventry City, Hull City, Ipswich Town, Leeds, Sunderland) are built player by player instead, because their filed payroll describes a squad that has since changed too much to roll forward: for most, a Championship season. They have no roll-forward and are not in the comparison.

What went wrong

Tottenham's reconstruction added £75.5m to a share-derived base of £207.2m: a net increase of 36.4%. The club had bought eleven senior registrations in one summer, so a large figure was not obviously absurd, and the arithmetic beneath it was internally consistent.

But a 36.4% net increase means the movements since the accounts added more than a third of the playing payroll, after departures had already been allowed for. Rebuilt player by player from the actual movements, the figure is £29m, 14.0% of base. The difference is £46.5m of annual wages and 7.6 points of ratio.

The comparison nobody had run

The same quantity exists at every club that rolls its filed payroll forward, and it had never been put side by side. As a share of the base it is added to:

Leaving aside the explained case, the range runs from −5.6% at Crystal Palace, where departures removed more payroll than arrivals added, to 14% at Tottenham, with Liverpool next at 13.4%. Tottenham's September figure sat at 36.4%. One comparison, a single query, would have found it at any point in the three months it existed.

Manchester United sit above the range, at 18.9%, and the comparison is what required the figure to be explained rather than accepted. The filed year, 2024/25, was a season without the Champions League, and United's player contracts are widely reported to reduce pay by around a quarter in such seasons. About £30m of the £37.5m is the reversal of that reduction and Champions League bonuses on qualification for 2026/27; the two transfer windows since the accounts date roughly cancel out on wages.

Tottenham shown at its corrected figure; before correction it sat at 36.4%. Figures are the mid-point of each club's stated range.

Why nothing caught it

Every gate in force at the time passed. One required the roll-forward to cover every transfer window since the accounts date, and Tottenham's did. Others covered squad completeness, window declarations, contract terms and amortisation. They all tested completeness, consistency and reconciliation. None asked whether the size of the roll-forward was plausible beside the same quantity at other clubs.

The attention had gone to the other half of the estimate. The players' share of staff costs was obviously uncertain and obviously important, so it was built from the club's own employment disclosures and defended at length. The roll-forward was a downstream arithmetic step applied to a number already checked, and it inherited a confidence it had not earned.

What it cost

The correction reversed the conclusion of a published piece, and is recorded as such on the article itself. Two things were not the cause, though a reader could reasonably suspect either: the treatment of Omar Marmoush's loan and its obligation to buy was already reflected in the published figure, and so was the players' share of staff costs, which was unchanged. (A separate error found later, on 27 September, concerned Marmoush's and Mykhailo Mudryk's loan wages, which were missing from Tottenham's figure; it is recorded in the changelog.)

The control that followed

The comparison above is now part of the review whenever a club is built or rebuilt. A figure outside the range is not necessarily wrong, but it has to be explained rather than accepted, as Manchester United's was.

It is not yet an automated gate. The existing gate checks that the roll-forward covers every closed window, a completeness test rather than a magnitude one, so this control still depends on somebody running it. Making it automatic is outstanding work.

The general lesson

An input can be built carefully, documented fully and still be wrong by a factor of two and a half if nothing asks whether its result is plausible beside the same quantity elsewhere. Care over the inputs does not by itself validate the output. Here, the missing control was comparison.

The roll-forward is the net wage effect of arrivals, departures and renewals between the filed accounts date and the season under test, added to a base of the filed staff-cost line multiplied by the estimated share attributable to players. Both steps are estimates. Independent estimate, not an official calculation, and not advice.

← Back to the Orange Engine · Changelog · Technical note: the disposal basis