The method · kept in step with the index · last updated 15 October 2026
Every number on the site comes from a rule you can read on this page. No panels, no vibes, no adjustments for clubs we like. For each measure we tell you what question it answers, what data feeds it, how the sum works, what it can’t tell you, and how much to trust it. When the method changes, this page changes with it — the log at the bottom says what moved and when.
This page is the plain-English version. If you want the statisticians’ one — the actual regressions, coefficients, shrinkage formulas, censoring rules and validation numbers behind every score — it has its own page: The nerd annex →
Every one of the 92 league clubs gets two marks out of 100, and higher is always worse:
The combined score is simply the average of the two. All three are percentiles: a score of 80 means “worse than roughly 80% of the league on this measure”. Nobody is judged against an ideal club; everybody is judged against everybody else. The window is the five seasons from 2022/23 to the current one.
One honest framing before the detail. The weights that combine the measures are editorial choices, not statistical discoveries — they are the editor’s view of what matters, published in full on this page, so you can see exactly what you’re disagreeing with. The measures themselves are where the statistics live, and each one below carries its own health warning.
| What | Source | How often it updates |
|---|---|---|
| Results, league tables, bookings | football-data.co.uk (Premier League to League Two) | Automatically, every Monday — the same weekly build recomputes every score and regenerates the share cards |
| Attendances | Transfermarkt attendance pages, with source-verified corrections where the feed was simply wrong (see the change log) | By hand, a few times a season |
| Stadium capacities, dated per season | Premier League handbooks, club announcements, stadium records — source recorded per row | By hand, as grounds change |
| Manager changes | Wikipedia’s per-season “managerial changes” tables, curated (every judgment call noted) | By hand, a few times a season |
| Points deductions | Wikipedia / EFL / Premier League announcements | By hand, as they happen |
| Transfer fees (context only — not scored) | Transfermarkt expenditure (gross, in euros) | By hand, after each window |
| Match-by-match crowds (away-support model) | Transfermarkt matchday pages, plus a corrections file with a source URL for every fixed row | By hand, with the attendance refresh |
| Crowd-behaviour verdicts | FA announcements and written reasons | By hand, roughly monthly: a scraper flags candidates, each confirmed manually |
| Football-related arrests and banning orders | Home Office official statistics (per club, annual) | By hand, each autumn publication |
| Manager and club misconduct sanctions | FA disciplinary announcements and written reasons; EFL/PL statements | By hand, roughly weekly: a scraper flags candidates, each confirmed manually |
| Wage bills and pre-tax results | Each club’s own statutory accounts at Companies House | By hand, every April once the filing season ends |
Every wage and profit figure traces back to a named filing on the Companies House website; our source records note which filing, when it was filed, and any judgment calls made while reading it. Before every build, an automated validator checks the lot — crowds against that season’s capacity, fixtures for duplicates and gaps, every club name against every file, sanction statuses for contradictions — and a broken source file stops the build rather than quietly feeding the scores. Nothing is silently capped, corrected or invented: errors are fixed against a named source or the row is excluded, and the published data records which.
Five sins. For each sin a club gets a percentile against the other 91, and the score is their weighted average: the four structural sins (impatience, indiscipline, lawlessness, waste) carry equal weight, and the “your manager is an arsehole” factor carries half their weight — a sanction register is real evidence, but it shouldn’t swing as much as a structural measure. In percentage terms: roughly 22% each for the four, 11% for the fifth. A sin we can’t score for a club (in practice: waste, where accounts don’t disclose wages) drops out and the others reweight to fill the gap — missing is shown as missing, never guessed and never rewarded.
The question: does the boardroom back a manager or burn through them?
We count the permanent managers each club has used since August 2022 and divide their league matches by that number. A club on its fourth manager in three years scores far worse than one still backing its first. Caretakers and interim spells don’t count — unless the caretaker was later made permanent, in which case he counts from the day he took over. Clubs we’ve seen fewer than 46 league matches from (recent arrivals from the National League), or whose spell in our window was broken by relegation out of the league entirely, are held at a neutral 50 rather than judged on a scrap of data.
What it can’t tell you: why the manager went. A sacking, a resignation and a poaching by a bigger club all count the same here — the public record doesn’t reliably separate them for all 92 clubs, and we’d rather count departures honestly than guess at motives. Annotating departure types is on the improvement list. How solid: the manager record is complete for all 92, curated by hand from the dated Wikipedia tables with every caretaker judgment call noted in the source file. We also tested a fancier version that accounts for the current manager’s unfinished tenure; it ordered the clubs almost identically (rank correlation 0.99), so we kept the simple sum you can check on a napkin.
The question: how dirty are they on the pitch?
Yellow cards plus three times red cards, divided by matches with card data (which is over 99% of all league matches in the window). A straight red is treated as three bookings’ worth of trouble. That’s it — no adjustment for “mitigating circumstances”, because there aren’t any.
What it can’t tell you: whether the cards are cynicism, chaos or just a pressing style that referees dislike — a card is a card. It measures the players, not the crowd and not the boardroom. How solid: the most direct observation on the site — referee decisions, near-complete coverage, no estimation anywhere. We tested whether divisions card differently enough to need adjusting for: division membership explains about 8% of the variation in card rates and adjusting for it barely reshuffles anybody, so the raw, comparable number stays.
The question: has the club actually been punished for breaking the rules?
Most clubs have never been docked a point, and they all score zero here. A club that has been docked points — for entering administration, breaking spending rules, failing to pay its players, or filming other people’s training sessions — starts at 50 and climbs with the size of its total deduction, judged against the other deducted clubs. The reasons are listed on each club’s report page.
We forgive after five years. The index only looks at the five seasons from 2022/23 onwards, so anything older doesn’t count: a club docked points in 2021 is clean here, however famous the scandal. As the window moves on a season at a time, the oldest season drops off the end — which means every deduction is eventually forgiven. Every sin gets five years, and then it’s history.
One special case, stated as a rule and not a club: the index scores verdicts, not charges. An allegation, however loud, counts for nothing until a commission rules. But a club found guilty whose punishment hasn’t been set yet is not clean either: it sits at exactly the 50 floor that every punished club starts from — the verdict is established fact, the size of the sanction is not, and we refuse to guess it. When the sanction lands, the score climbs to its real place (and the season history records it in the season the ruling attributes it to); if an appeal overturns the verdict outright, the club reverts to clean. As of October 2026 this applies to one club: Manchester City, found guilty by a Premier League independent commission of all charges of serious financial-rule breaches covering 2009/10–2017/18, sanction hearing pending, appeal lodged.
What it can’t tell you: how bad the breach was beyond what the points say — an administration and a paperwork failure both score by their deduction, and the report pages carry the reasons so you can apply your own mercy. It also says nothing about clubs under investigation: no verdict, no score. How solid: deductions are public rulings, so the register itself is as complete as the governing bodies’ announcements. The ranking among punished clubs rests on a small pool (around ten clubs), so positions within it are coarse — and the 50 floor is an editorial statement that punished is punished, not a statistical estimate.
The question: does the club get the results its money should buy?
This is measured with real wages, from each club’s own published accounts: total staff costs per financial year (the three FYs 2023–2025, covering the seasons 2022/23 to 2024/25), each paired with the league season it covers. For every disclosed club-season we ask what points-per-match clubs on similar wages relative to their own division that season actually managed — a single pooled regression of points on division-relative wages, which on the current fit says doubling your division’s typical outlay is worth roughly 20 points over a 46-game season — and score the shortfall: waste is winning less than your money should buy, averaged over your disclosed seasons in the top four divisions. A club that pays double its division’s typical wages and finishes mid-table scores badly; a club that overachieves a humble payroll scores superbly — in any division, because each club is measured against its own division’s money. Seasons spent outside the league never enter the sum. Poverty is not a sin; wasting money is.
Why wages rather than transfer fees? Because fees lie. Lower-league fees are routinely undisclosed (most of League Two shows no recorded spend at all), fee totals ignore the money recouped from sales, and the wage bill is both the biggest cost in football and the best-documented: it sits in the statutory accounts that 65 of the 92 clubs currently file in full at Companies House. Clubs that file small-company accounts without a staff-costs note — or with fewer than two disclosed league seasons — show waste as unavailable, and the other four sins reweight to fill the gap. We don’t guess — and keeping your accounts secret will never earn you a good score.
Each club’s report page also shows its three-year pre-tax profit or loss for context. We deliberately don’t score it: one player sale or one property deal can swing a year’s result by nine figures, and scoring it would punish clubs with honest accounts and reward opaque group structures.
What it can’t tell you: anything about the 27 clubs whose accounts don’t disclose enough — they show “unavailable”, which is the truth. Staff costs also include everyone on the payroll, not just players, and a few group companies bundle in non-football business (each such judgment call is noted against the filing in the data). And an average over two or three seasons is still a short run: one freak season moves it. How solid: middling, honestly — and said so. The expectation model was chosen by out-of-sample testing against three alternatives (the formulas and test results are on the nerd page), wage bills explain over half the variation in points-per-match, and every figure traces to a named filing. But disclosure thins badly in League Two, where the few clubs that file full accounts may not represent the division; each club’s page shows how many disclosed seasons its score rests on.
The question: does the dugout keep getting itself sanctioned?
Cards are the players’ sins; this one belongs to the dugout. The FA and the leagues sanction managers for things that never show up as cards or points: charges over comments about referees, touchline bans for confronting officials, fines for fracas in the tunnel — and they sanction clubs for collective misbehaviour too, from losing control of the technical area to spying on other clubs’ training. We keep a register of every such sanction since 2022/23, each row traceable to the governing body’s own announcement.
Scoring: every sanction is worth its touchline-ban matches plus £10,000s of fine (a suspended ban counts half; the fine component recognises that commissions fine what they can’t ban). The crucial rule: a manager’s sanctions belong to the manager, not the scene of the crime. The factor scores each club on its current manager’s full record since 2022/23, wherever it was earned, plus the club’s own club-level sanctions, which never move. Hire a serial offender and his rap sheet arrives with him; sack him and it leaves. A clean dugout and a clean club score zero; offenders start at 50 and climb by rank among offenders, exactly like the deductions pillar. We publish the rule and each club’s score, not a league table of named individuals — the register exists to score institutions, not to pillory people (but yes, Ian Holloway scores badly on it).
What it can’t tell you: whether anyone is actually an arsehole — it counts governing-body findings, not character. Because the sheet follows the manager, a club’s score can jump on a hire before a ball is kicked, and the measure deliberately ignores a club’s history of employing offenders once they’ve left. Both are owned choices. How solid: every row is a published ruling with a source link, so the facts are firm; the editorial constants (fines at £10k a point, suspended bans at half) are choices, which is exactly why the whole factor carries half weight.
The fans score asks a crowd three questions: do they stay, do they travel, and do they behave? Each is a percentile, higher = worse. Staying carries 70% of the score — it rests on five years of revealed behaviour and is the hardest signal to fake — with travel at 20% and behaviour at 10%. A question we can’t yet answer for a club drops out and the rest renormalise, rather than counting against it.
The question: when the football turns bad, does the crowd?
This one needs a paragraph of honesty about method, because it’s the cleverest and most misunderstood number on the site.
You can’t compare attendances directly — of course Manchester United outdraw Accrington. What you can compare is how a crowd responds to change. Every summer, every club carries a result from the season just gone: promoted, relegated, up five places, down ten. We line up every one of those season-to-season transitions for all 92 clubs — currently 355 of them — and ask: when a club rises or falls through the 92-place pyramid, what happens to its gate?
The answer, averaged across the whole league: attendance moves by about +0.28% per place climbed within a division (the exact figure updates weekly and is printed on the front page), with a separate one-off surge for getting promoted (crowds turn up for a new division, currently about +7 points of gate) and a slump for going down. Those jumps are a crowd responding to the occasion, not loyalty — so they’re stripped out rather than credited to promoted clubs’ fans. What remains is the national norm: what an average crowd does. Each club’s fans are then judged by their deviation from it: fans who turned up more than the football deserved sit above the line (loyal), fans who melted away faster than the results justify sit below it (fair-weather). Each club’s loyalty score is its average deviation across its transitions, damped toward zero when it rests on few transitions — we tested it, and a club’s one or two most recent summers simply don’t predict its next one well enough to take them at face value, so a single spectacular season (hello, Barnet) is read with appropriate suspicion rather than crowned as destiny. The Shit Fans score is that damped number flipped into a percentile — so 90 means fans more mercenary than 90% of the country.
Three honest footnotes:
A club needs at least one observed transition for the loyalty question (the damping above supplies the caution, so a single transition counts for little); brand-new arrivals are scored on the other two questions until they’ve been observed rising or falling once.
What it can’t tell you: season-ticket inertia versus bodies through the turnstile (the attendance figures are the clubs’ reported gates), why a crowd moved (ticket prices, a takeover and a cup run all land in the same number), or anything upward about a sold-out ground. And a “transition” is one summer — this is revealed behaviour at the moments results changed, not a census of devotion. How solid: the national response is estimated from 355 transitions and is firm; any single club’s deviation rests on one to four transitions and is the weakest signal on the site — which is exactly why it’s damped. The damping is estimated from the data, not hand-tuned: currently a one-transition club keeps about a seventh of its raw deviation and a four-transition club about two-fifths. Each club’s page shows the transition count, the raw deviation and the damping weight, so you can see precisely how much evidence sits under the number.
The question: who actually turns up at the away end?
Nobody official publishes away-end counts, so this one is an estimate, and labelled as one: ~8,500 match crowds (source-verified against per-season ground capacities — see the change log) feed a model that learns each ground’s normal gate and reads each visiting club’s crowd lift — the extra bodies that arrive when they come to town. Three fairness rules, all editorial: the lift is expressed as a percentage of the visiting club’s own average home gate (3,000 travelling Mackems are a smaller share of Sunderland than 1,500 Grecians are of Exeter); it is adjusted for the miles the trip demanded, so Plymouth’s 400-mile round trips are judged as the schleps they are; and away-allocation sell-outs are treated as floors — a club that fills every away end it’s given can only be at least that good, and is marked ≥ on its report page. Weak travel = high score.
What it can’t tell you: the difference between travelling support and box-office pull — a crowd lift is everyone who came because that club came to town, home fans included. It’s a proxy for away following, not a turnstile count. At consistently sold-out Premier League grounds it can barely move at all, which is why those clubs sit on floors. How solid: strongest in the EFL, where we checked it against the only independent ground truth that exists — fan-compiled away-attendance tables for all 72 EFL clubs in 2023/24 — and the two agree at a rank correlation of about 0.94 overall and roughly 0.9 within each division. The Premier League publishes nothing to check against, so PL away scores are allocation-capped floors with honest error bars of unknown size. The estimate is also fitted once over the whole window, so in the season history it stands still while everything else moves.
The question: is following this club a public-order problem?
Two halves, averaged. First, FA verdicts on crowd behaviour — pitch invasions, thrown objects, pyro, discriminatory chanting — counted as verdicts (not pounds: the FA fines by club resources, so pricing in money would just rank turnover), with discriminatory cases counting double and fines breaking ties; clean crowds score 0, sanctioned ones start at 50, verdicts-not-charges as everywhere else on this site. Second, the Home Office’s official per-club football arrest statistics, domestic competitions only, per 10,000 of average home gate over the last three published seasons, gently shrunk toward the league average so a small ground’s noisy numbers don’t swing it.
What it can’t tell you: an arrest is not a conviction, and policing intensity varies by club and fixture — a heavily policed derby produces arrests a quiet afternoon doesn’t. The statistics also lag about a season behind, and a club with no published arrest seasons yet (a new arrival) is held at a neutral 50 on that half rather than presumed clean or guilty. How solid: the FA half is published rulings — firm facts, small counts. The arrests half is official statistics with the confounds above, which is why it is shrunk toward the average and carries just half of the 10% behaviour share — about 5% of the fans score.
Raw numbers (cards per game, pounds of wages, percentage points of gate) are never mixed directly — each is first turned into a percentile: the share of clubs you’re worse than, with ties split down the middle. That keeps one wild number (Chelsea’s wage bill, say) from drowning out everything else, and it makes every component mean the same thing: your place in the national queue of shame.
The front-page scatter plots every scored club with the institution on one axis and the crowd on the other. Top-right is the full package — a shit club and shit fans, failing each other simultaneously. Bottom-left is the blessed corner. Clicking any dot shows that club’s path across the chart season by season.
The site lets you do more than read the verdicts — and each feature follows the same rule as the scores: generated from the data by a published rule, never written to order.
Each club’s path on the quadrant is not interpolation. For each past season we recompute the whole index as it would have stood at the time: only the matches, cards, manager changes, deductions and completed transitions known by the end of that season, wage bills from the accounts covering the seasons played by then, and percentiles taken against the clubs that were in the league at that point. One honest wrinkle: accounts are matched to the seasons they cover, even though they’re filed months later — the history asks “how wasteful were they being?”, not “what could we have known that May?”. A second: the away-travel estimate is fitted once over the whole window, so in the history it is held constant while the other components move season by season.
Two different questions, two different series. The paths above are recomputed
under the current rules (the calculation version printed in the data), so a
club’s drift is real drift, not a methodology change. Separately, every build
appends the day’s published scores to an as-published ledger
(data/score_ledger.csv, also shipped with the site data) that is never
rewritten: what we said at the time stays what we said at the time. If we ever change the
method, the gap between the ledger and the recomputed history is the method changing, not
the club — and the change log below will say exactly what moved.
Two more pieces of honesty in the history views. A season where a component couldn’t be scored yet is named as such on the club’s season breakdown — so a score jump caused by a component arriving (loyalty becoming scoreable, say) is visible as exactly that, not mistaken for behaviour. And the “why have we got shitter?” feature, which decomposes a club’s move between any two seasons into per-component contributions, separately states any composition effect from availability changing — with the standing caveats that percentiles are relative (the comparison pool changes), everything is recomputed under today’s method, and a contribution says which dial moved, not why.
The index is alive. Every change to the method will land here, newest first.
Launch — First published version.