Sellout Model

Experimental. Probability a performance sells out, scored against 2025 outcomes.

← Back to Show Radar
This page is a staging surface, not a booking signal. The probabilities come from a frequency table fit on 2025 outcomes: hours-to-showtime × remaining availability × recent movement, tilted by venue and time-slot reputation. It has never been shown to users and has not been validated against 2026 outcomes. The “Radar pace” column is the estimate the live site already uses, shown for contrast — where the two disagree is the interesting part.
—Loading

Where the answers come from

The prior is hierarchical: it tries the finest cell first — lead time × availability × movement × room size — then drops room size, then movement, then answers on lead time alone, then falls back to the 13.9% base rate. Under the 2026 fit every room-size cell exists, so a row below that level means the state itself was never scored, not that the sample was thin: thin cells are shrunk toward their parent rather than discarded, and report a low confidence instead of disappearing.

Going fast

Next 7 days, one row per show — its most at-risk upcoming performance.
Show Performance Venue Avail Model Radar pace Δ Confidence Answered by
Loading board…

Recently gone

From the sellout ledger — observed crossings, not predictions.
Show Performance Sold out at Lead time
Loading ledger…

Known limits

    Ed Fringe Guide is an unofficial, fan-made guide. Not affiliated with the Edinburgh Festival Fringe Society or edfringe.com.