Nobody knows the 2027 lineup yet. We don't either. What follows is a model's best guess, built entirely from public information, and an honest account of where it falls down.
I
What the model looks at
Five things, weighed against each other.
Previous Pyramid headlinersActs who have headlined before are the strongest signal available. Glastonbury invites its own back more often than not.
Touring scaleStadium and arena activity in the UK right now. An act selling out stadiums is already playing at Pyramid scale, whether or not they have ever been to Worthy Farm.
Festival circuit positionWhere an act sits on the usual climb. Smaller stages, then main support, then headline slots at other festivals, before a Pyramid booking makes sense.
UK footprintGlastonbury keeps a British identity even when its headliners are not British. An act with a real following here reads differently to one that is simply globally enormous.
MomentumA new record. An announced tour. A run of festival dates. The things that make a booking plausible now rather than in principle.
II
Two numbers
Every act on this site carries two figures.
The first is built only from touring, festival and Glastonbury history. It never sees a bookmaker's price.
The second takes that number and folds in bookmakers' odds (their standing board prices, plus older quotes with the dates they were reported) together with credible reports.
Both are always shown. The first is never hidden, and never quietly replaced by the second.
The first number is allowed to be wrong in public.
That is the point of publishing it.
III
What it gets wrong
It reads acts with a real UK following better than it catches a global star's one-off moment. It only predicts the three headline slots. The rest of the poster is not modelled at all.
Two things worth knowing about the numbers themselves.
They are per-act estimates
Each figure answers one question: how likely is this act, on its own. They do not add up to three slots, and the board is not a pie being divided.
Rumour adjustments are a judgement call
When a credible report moves a figure, the direction comes from the evidence. The size of the move is ours. We tested it, the sample was too small to give us a number, so we are telling you plainly that it is a guess.
IV
Where the odds come from
Part of the second figure comes from bookmakers. Each price behind it is either a dated quote from a named bookmaker, carried by a named publication, or a standing position read from a bookmaker's own board, which carries no publication date at all. Undated prices are shown as read on the day rather than given a date they do not have. There is no live Glastonbury 2027 headliner market to read, so when a price was reported matters as much as the price itself, and prices past their freshness window are shown with their age rather than dropped.
The source list loads with the latest published board.
When we withhold a market figure
Sometimes the festival's organiser says in public that an act will not be playing, on the record and with a date, while bookmakers carry on pricing them. When that happens we stop publishing the market's figure for that act as a probability.
This is a decision about one number, not a claim that the market is lying and not a deletion of anything. The prices are still collected, the act still appears on the board, and the statement that caused it is shown with its date and its source so you can read it yourself. Nothing about how the model scores changes.
Currently withheld:
Structural rules
Four standing rules sit above the scoring model. None of them is a prediction; each is a rule about what we will publish, with its basis and the condition that kills it stated in advance. Two are editorial: the modeller's judgment, labelled as such, not something new data forced.
Organiser denial. An act the festival's organiser has publicly ruled out, with a date and a citation, publishes no probability while the statement stands. Basis: the organiser's word outranks the market's. Killed by the act appearing on an official lineup, or the edition concluding.
Gap-one promotion exclusion (editorial). An act that headlined the Other Stage at the immediately preceding edition scores zero for the Pyramid slot, in every tier. Basis: a verified record in which that promotion has never happened and was never even priced or tipped, that is the coverage we could find, not everything ever published, plus the modeller's judgment that the statistical residual left on that zero was smoothing, not signal. Killed permanently by any official lineup announcing that promotion.
Bill composition. No projected bill we publish is an all-male trio. Basis: not the record, which says all-male bills are historically common, 18 of 25 including 2023; the basis is the organiser's public 50:50 commitment, the two bills booked since, and editorial judgment. Killed publicly by an all-male official 2027 poster.
Quiet-cycle returner dampener (editorial). A returning act with no announced shows and no studio album in over two years carries a heavy discount on the evidence line, never a zero, because our record contains exactly one quiet-spell booking, in 2011, among twelve measurable returns. The rule reads live data and an act's own next announcement lifts it automatically. Killed by an official 2027 headliner booked from a quiet cycle.
How fresh the streaming data is
Listener data: streaming figures update on a rolling schedule, and probabilities always use the latest figures we hold.
V
Where an LLM fits, and where it does not
Part of the pipeline uses a large language model. Its job is reading, not predicting: it scans coverage for reported bookmaker odds and rumours, and grades what it finds by the quality of the source carrying it.
Those grades are a gate. Only confirmed or strongly reported items can touch the numbers, in bounded, logged adjustments that expire on their own. Speculation is tracked but never scored: a rumour with no named source sits in the record at zero weight, however loud it gets.
The LLM reads and grades. The model does the scoring. Nothing a language model writes becomes a probability on this site.
VI
How current form is measured
Touring form is measured over one window: the two years running up to the festival, ending when the festival does. Everything the model reads about how active an act is right now comes from inside it.
Shows already played are weighted towards the present. A show played today counts in full, a show at the very start of the window counts close to nothing, and the weighting runs in a straight line between the two. Announced dates count at full weight for as long as they are still on sale.
The festival dates are an assumption
The window is anchored to 23–27 June 2027. Those dates have not been announced. They are the late-June pattern every recent edition has followed, and we are using them as a working assumption. That is an editorial choice, the same kind as the weight we give the betting market.
What stops counting, and when
A cancelled announcement drops out on its own at the next data sync, with nobody having to flag it. A date that has already passed never counts as an announcement. Before this changed, thousands of past-dated announcements were being counted twice over: once as a gig that was coming up, and again as the same gig after it had happened.
Going quiet is now gradual
An act that stopped touring used to fall off a cliff. Crossing two years without a show cost it close to half its standing as a returning act, overnight. That happened to The Rolling Stones in August 2026, while this site's own touring tag, reading from a different window, still showed seventeen recent shows beside it. Two windows on one screen. Moving every surface onto the window above closed that gap, and going quiet now fades in month by month instead.
Where an act carries a touring tag on this site, its counts mean shows inside this window, live announcements included.
Why a figure moves
A number on this site can change for two quite different reasons, and they are worth telling apart.
Sometimes the evidence changes: an act announces a tour, releases a record, plays a bigger room than it used to. That is the model working.
Sometimes our data changes and the act has done nothing at all. We find a show missing from a history. We correct a venue we had filed at the wrong size. We discover a figure was being read from a months-old copy of a record instead of the current one. Those moves can be just as large, and they are corrections to us, not reassessments of anyone.
We do not smooth either kind out, and we do not present the second as though it were the first. When a figure moves because we were wrong, that is what we will say.
VII
Bills, not just names
Alongside the per-act figures, the model scores whole bills: three names taken together as one prediction, rather than three separate numbers stacked up. That layer sits downstream of the per-act scoring and cannot reach back into it.
Nothing about an act's gender, age or background goes anywhere near its own figure. The patterns below describe a trio as a whole, and a test in the model's own suite fails if any of them ever reaches a single act's score.
How many first-timers a bill usually has
From our own record of the twenty-five headline bills between 1994 and 2025:
0 of 25 bills had no first-time headliner at all. It has never happened, so the record weights it at zero. That is the record talking, not a judgement we made.
3 of 25 had one first-timer.
13 of 25 had two.
9 of 25 had all three.
Gender balance
A trio with no women in it is weighted down. The size of that weighting comes from how often it has actually happened lately: 3 of the 8 bills from 2015 onwards. Across the full record it is 18 of 25, and across the two bills since 2023 it is 0 of 2.
The weighting is anchored to the festival's own stated aim rather than to ours. Emily Eavis, speaking to BBC Radio 1 Newsbeat in February 2020:
Our future has to be 50/50.
It weights down. It never zeroes an act or a bill out.
Genre balance is not modelled
A real Glastonbury bill spreads across genres, and we do not model that at all. We hold no genre data for these acts to model it with. It is a gap, and we would rather name it than let its absence read as a decision.
Two stadium acts at once
A bill occasionally lands two acts who are both in the middle of a stadium run going into the festival. On the verified record that is rare: it has happened once, in 2023, across the seven editions where we hold enough touring data to check it. Only seven of the twenty-five editions can be checked at all, because our show data begins in 2012.
That is context worth knowing beside a bill. It is not one of the factors above, and nothing in the scoring reads it.
Preview: showing , not today's date. Set or remove the otd query parameter to change it.
A record of announcements that have already happened, not a prediction of what comes next.
On this day in Glastonbury announcement historyGlastonbury announcement history
Every announcement in the record, in calendar order. Open the cycle below to see when the festival usually says what.
Licence
Glastonbury 2003
Michael Eavis lodged an appeal after the council's regulatory board voted to refuse a licence for the following summer's festival.
Organisers lodged the entertainments licence application for the following summer, to be considered by the regulatory board before the end of the year.
Oasis and Muse were reported among the headline acts, days before tickets went on sale. The same report named a third act who did not in the end top the bill, so it is recorded here only for the two that did.
Tickets went on sale at 8pm by telephone and through the festival website, priced at 112 pounds, with credit cards barred and a limit of two per person after the previous year's problems.
The festival's own history records that final confirmation of the band's first Glastonbury booking came through on this day, during a phone call witnessed by a BBC Today programme presenter.
Mendip District Council granted a new ten year licence running to 2024, holding the public ticket allocation at its existing level while allowing more staff on site.
Foo Fighters withdrew from the Friday headline slot after Dave Grohl broke his leg falling from a stage in Gothenburg, forcing the band to call off their remaining European dates.
Coldplay were announced as the first confirmed headliner, taking the Sunday night and becoming the first act to top the bill at the festival four times.
Muse were confirmed to headline the Friday, their third Pyramid Stage headline booking and the one that made them the first act to top the bill on all three nights.
Michael and Emily Eavis announced that the fiftieth anniversary edition was cancelled because of the coronavirus pandemic, making it an enforced fallow year, with deposits rolling over to the following year.
Michael and Emily Eavis announced that the edition would not take place, a second consecutive enforced fallow year, with deposits again rolling forward.
The licensing sub-committee granted a premises licence for a scaled down two day event at Worthy Farm that September, capped at just under 50,000 people and with no camping.
A batch of resale tickets went at 9am and was exhausted in a little over twenty minutes, the edition having already been fully allocated through deposits rolled over from the cancelled years.
Elton John was confirmed as the Sunday night Pyramid Stage headliner, months ahead of the rest of the lineup, for what was billed as the final UK show of his farewell tour.
General admission tickets were exhausted that morning. Reports of how long it took differ between thirty two and thirty five minutes, and we have not been able to settle which is right.
Neil Young said publicly that he would not play the festival, objecting to its broadcast arrangements. He had not at that point been announced on the lineup.
The festival named the days it will run in 2027, from the Wednesday to the Sunday of the last week of June, and offered the first pairs of tickets through a charity prize draw.