How this site works
Editorial method
Every capacity on this site is attributed to whoever published it and marked according to whether a person has opened that page and read the figure off it. This explains what the marks mean, and why a page here will print two numbers rather than pick one.
Most stadium sites answer a capacity question with one number and no provenance. The number is usually copied, often stale, and occasionally traceable to nobody at all. This site prints where each figure came from, how well checked it is, and — where publishers disagree — all of them.
Verified and accepted
Every figure carries a source record, and every source record carries one of two dates. The difference between them is the whole of what this site claims about its own reliability.
- Verified — a person opened the cited page and read the figure off it. The date is when they did that.
- Accepted — the figure came from research and nobody has opened the cited page to confirm it. The date is when it was recorded.
A page reports the weaker of the two. If a single source on a page is accepted, the page says Accepted, and it shows the oldest date among its sources rather than the newest — the label describes the page as a whole, so it should say how stale the weakest part might be, not how fresh the strongest is. Where a page mixes the two it also says how many of its sources are verified.
Most pages here still say accepted. That is not a defect being hidden behind a nicer word; it is the current state, printed at the top of the page rather than in a footnote, and the only thing that improves it is somebody opening pages.
Nothing automated may promote accepted to verified
Not a link checker returning 200, not a figure that looks plausible, not a bulk pass over the dataset. A script that upgrades an assurance date is not a maintenance task, it is a bug — because the only thing "verified" asserts is that a human looked, and no program can supply that. The build has a link checker and it deliberately cannot write to these dates.
Printing both figures
When two permitted sources give different capacities for the same ground, this site shows both with their publishers attached. It does not average them, pick the rounder one, or quietly prefer whichever is more widely repeated. Two examples, both from pages here.
Croke Park
The GAA contradict themselves inside one document. Their 2025 annual congress report gives 82,300 in its text and 82,006 under the heading Overall Capacity — 294 apart, in the same publication, reconciled nowhere. No reason is published, so none is recorded here: terraces, media positions and accessible platforms would all explain the gap and every one of those would be this site's invention rather than the GAA's statement. Croke Park seating plan
Stamford Bridge
Chelsea state 40,022, plainly and without qualification, on their own general information page. Almost every other stadium site prints 40,341, which appears nowhere on Chelsea's pages. This is the case that decides how the rest of the site treats consensus: the widely repeated figure is not the sourced one. Stamford Bridge capacity
Two figures are not automatically a conflict
Before recording a disagreement, the question is whether the two sources are measuring the same thing at the same moment. An outdated club page and a published projection are a timeline, not a dispute; a seat count and a licensed matchday capacity are two different measurements that happen to describe one ground. Where the capacity genuinely changed between two statements, the row says it changed rather than that the sources differ — presenting a timeline as a disagreement is its own kind of wrong answer.
This is why each record stores the capacity concept alongside the number. "Matchday capacity", "seats", and "capacity during redevelopment" are not interchangeable, and a correct figure still misleads without the words that say what was counted.
Which sources count
Official sources only: the stadium's own site, the resident club's, or the body that owns or governs it. Not Wikipedia, not stadium databases, not aggregators — and the reason is not snobbery about the format.
An aggregator's figure is evidence that other aggregators print the same figure. Stamford Bridge's 40,341 is agreed on across most of the web and traces to no publisher at all, while the club's own 40,022 sits on a page anyone can open. Consensus among sites that copy each other is not corroboration; it is one unsourced number with a large audience.
Murrayfield is the same lesson from the other direction. Its record once led with 67,130, cited to a Scottish Rugby page that had started returning 404. The figure could not be found on any live Scottish Rugby page — it survived only on the dead URL and in search-engine summaries paraphrasing it. Meanwhile 67,144 appears on two live Scottish Rugby pages, one of which explains it: the capacity reached with the corners filled in after the 1994 West Stand. So the fix was not a new URL for the old number. The two changed places, and a dead link is now treated as a reason to re-check the fact rather than to go looking for where the fact moved.
The club is not the only first-party source
"Official" is broader than the club whose name the ground carries. A co-tenant is a resident club, and a venue operator is first-party. Two Premier League grounds are on this site only because of that: the MKM Stadium's 25,586 is stated by Hull FC, the rugby league club that has shared the ground since 2003, and the Coventry Building Society Arena's 32,609 by the arena operator, a separate company from the football club. Both are seat counts in their publisher's own words rather than licensed capacities, and both records say so.
A documented absence
Where a club publishes nothing, that is often the most useful thing a page can report — provided it is reported as a finding rather than as a gap in the writing.
The Emirates
Arsenal publish neither a seating plan nor an away section on any page a reader can open, and both were available once: one document answers 404, the other 410 — the status a server sends for something withdrawn on purpose. This is not a ground that never documented itself. It is one that has taken down what it had, which means every site still naming an away block at the Emirates is repeating a figure the club has removed. Arsenal stadium tour
An explained absence only beats a blank when the explanation is sourced, and the site distinguishes the two. Spanish grounds have no separately marketed away allocation because Real Madrid state on their own ticket pages that legislation requires away tickets to be sold through the visiting club — quoted, first-party, settled. American grounds appear not to publish away sections either, but the reason usually given for that is general knowledge supplied by a writer rather than read off any page, so it is labelled an observation and not a finding.
The same rule decides what gets a page at all. A ground earns one only where its club or operator publishes enough for the page to say something a reader cannot get from the club's own site. Where it does not, the ground stays a row in a ranking, which is an honest answer rather than a missing one — sixty thin pages would have been easier to produce and worse to read.
What a date means
A date attached to a figure on this site is when a person opened the page that publishes it. It is not when this site's page was edited, not when the file changed, and not when the build ran. Those are all facts about us; the date a reader needs is a fact about the source.
That is also why the sitemap carries no last-modified date. The only date available when the site is built is the build itself, and stamping every page with today's date whenever any one page changes tells a crawler that everything changed. Silence is more accurate.
This page is one of four that cite no figures — the others are about, contact and privacy. They carry an ordinary "last updated" date in the byline, meaning when the page was last written. That is a different kind of date from the assurance line on a stadium page, and it is labelled differently on purpose.