The OldPubs dataset holds 340 old pubs in 31 countries, and gives each one two dates rather than one: the year the house claims, and the earliest year a document actually supports. 155 of the 340 have that second date; the rest carry a claim nobody has yet been able to trace to a record. It is free to reuse with credit, and the earliest documented year in it is 1259 at Taverne zur Krone.
Most published lists of old pubs repeat the first number and stop. Keeping the two apart is the whole point of the file, and it is why a row can say “claims 1189” and “first recorded 1751” without either being wrong.
The files
- oldest-pubs-graded.csv — one row per published pub, ranked pubs first. Two comment lines at the top explain the grades and the listing licence; drop any line beginning
#before parsing. - oldest-pubs-graded.json — the same rows, with the generation metadata in an object at the top.
- pins.geojson — map positions. This one is not under the same licence: the coordinates come from OpenStreetMap and carry ODbL, which is why they are kept out of the other two files.
What the columns mean
slug- The identifier used in the page URL. Stable: a pub keeps its slug when its data changes.
name- The name over the door today.
city, country, nation- Where it stands. `nation` separates England, Scotland, Wales and Northern Ireland inside the United Kingdom.
type- pub, inn, tavern, bar, beer hall, brown café and so on — what the house is, in its own country’s vocabulary.
claimed_founded- The year the pub claims, as a number. This is what the sign, the website or the local tradition says. It is not a finding.
earliest_documented- The earliest year a document actually supports, and empty where no document has been traced. This is the column to trust.
gap_years- earliest_documented minus claimed_founded. Negative where the record is older than the claim, which is the case in 7 rows.
building_year, building_year_source- When the fabric was built, where that is known separately from the drinking, and who says so.
evidence_grade- A, B or C. What kind of evidence exists — never how good the pub is.
trading_status, status_confidence, status_checked- Whether it is still pouring, whether that was confirmed from a dated source, and when we last looked.
listed_status- Statutory listing, where there is one. Contains public sector information licensed under the Open Government Licence v3.0.
top100_rank, not_in_top100_because- Its place in the ranked hundred, and — where the cell is empty — the reason it is not eligible.
trade_today- One sentence on what the house is now, where the type label alone would mislead.
evidence_notes- The reasoning: what was read, what it said, and what remains unresolved.
sources- Every source cited for that row, separated by a pipe.
page- The pub’s page on this site, where the same evidence is set out for a reader.
The three grades are set out in full on the methodology page, with the tests each one has to pass. In short: 56 rows are grade A, 63 are B and 221 are C, and a C means no document has been traced — not that the claim is false.
What it covers, and what it does not
Every row is a house somebody has read the evidence for, one at a time. There is no scraped directory behind it and no automatic import: 340 pubs across 31 countries, with 155 carrying a documented date. Pubs whose data is incomplete are researched but not published, so the file holds fewer rows than the research does.
It is not a guide to pubs open near you, and it is not complete. Old pubs that nobody has written about are missing, and whole countries are thin. A correction with a source is the fastest way to change that — see corrections, where every change made so far is listed with its date and reason.
Licence and attribution
The dataset is free to reuse under an open licence. Use it commercially, change it, build on it — the one condition is credit. This line is enough:
Data: OldPubs (oldpubs.com).
Two things travel with their own licences and are not ours to relicense:listed_status contains public sector information under the Open Government Licence v3.0, and the GeoJSON coordinates are © OpenStreetMap contributors under ODbL.
How it is updated
The files are rebuilt whenever the underlying rows change, and the date stamped on them is the newest date the data itself asserts — never the moment the site was last deployed. A file whose date advances on every rebuild tells a reader the research is fresher than it is. This release was last changed on 18 September 2026.
Found something wrong?Suggest a correction