Glastonbury Hotspots 2025
One of the biggest criticisms of Glastonbury 2024 was that overcrowding was, at times, verging on dangerous. Whilst it is easy to point fingers at the increased capacity, I believe that the main sets at which this occurred (Avril Lavigne at Other Stage and Bicep at IICON, as well as Example at Glade in 2023) were due to not enough other large acts performing at the same time.
Crowd size and demand throughout this analysis is not scaled to the size of the attendees to prevent needing to model a stage popularity multiplier, and the number of people working shifts rather than attending slots.
Obtainining slot popularity
Set data is scraped from the Glastonbury website using {rvest}, manually cleaned, then matched to artists listed on Spotify via their API. From this the popularity for most artists artists playing at each slot can be determined (some unknown popularity scores are manually set to imrove output quality). If multiple artists are playing a slot, the maximum of those artists’ popularities is used. This metric is unfortunately only available at a global level, so the popularity of artists such as Dochii and SZA is likely skewed by a large American audience.
Stage popularity
From this the popularity of each stage at 5 minute intervals can be determined. If there are gaps of up to 90 minutes between slots (i.e. for band changovers) the popularity is filled using cubic spline interpolation to mimick the volume of crowd increasing/decreasing depending on the difference in popularity between sequential artists. Gaps longer than 90 minutes have their popularity set to 0. If the first set of the day has an unknown popularity, it is assumed to be 5. If another set has an unknown popularity, it is assumed to be that of the previous one. Some modelling improvements could be made here, especially to account for jumps in DJ set popularity, smoothing is later applied that should account for this.
Crowds
Summing together this stage-by-stage popularity then fitting a polynomial surface over the top allows us to model crowd size over the duration of the festival. This broadly makes sense as:
A small portion of punters will go to slots on Wednesday, but most will be arriving and setting up tents
With few bands playing on Thursday, a larger portion of punters will attend DJ sets on Thursday than Wednesday but it is still much quieter overall than the actual festival
Friday-Sunday sees peak times in the evening around 10PM
Some guests do go home on Sunday before the festival finishes so that they can go to work the next day / to avoid traffic
Capacity
The capacity of each stage is the maximum artist popularity that it can sensibly cater for. This is calculated by removing outliers, then taking the 95th percentile of slot popularity. Stages with vastly different IRL capacities may have similar artist popularity distributions (i.e. Pyramid and Other) but more people are likely to attend the larger stages.
You may notice on the box and whisker that some stages have artists playing that are far too large for them (i.e. Tom Walker at Strummerville)
Predictions
Methodology
Demand = (Crowd Size * (Artist Popularity / Total Artist Popularity)) / Stage Capacity
This is then smoothed using a rolling mean across each stage.
2024
To validate that this model works, let’s look at the model’s predictions for 2024. We can’t really look any further back as the popularity of artists may have changed too much.
It probably doesn’t matter too much if small stages have high demand as crowd sizes should still be manageable; filtering these predictions for venues with capacity of at least 10,000 shows Bicep and and Avril Lavigne in the top 10 busiest predicted slots. I imagine that the demand for Kneecap is inflated due to a recent increase in popularity following certain events.
2025
And here are the predictions for 2025:
Filtering this again for the larger stages, we can make the following observations:
Noah Kahan, Olivia Rodrigo and The 1975 will be the biggest acts at The Pyramid stage
Assuming that Dochii’s popularity is over-inflated such as SZA’s was for 2024, Charli xcx will be the busiest headliner of Saturday night
This heatmap reveals long periods of high demand at each stage - the worst of these being Genosys on Saturday (no, I’m not building a model to predict who DJ XXYZ is - seems pretty obvious already…).
Comparison
The following 2 visuals show that, compared to 2024, 2025 has less high-demand slots - which is great to see.
