SupaPark

SupaPark Data — Prediction Accuracy

How accurate are SupaPark's predictions?

Over the last 30 days (2026-08-10 to 2026-09-09), SupaPark's served ride-wait predictions missed the actual posted wait by an average of 8.1 minutes across 131,612 graded predictions, and park-level forecasts missed by an average of 9.1 across 13,789 graded predictions — graded against what actually happened, misses included.

Of the 131,612 graded ride predictions that carried a confidence band, the actual value fell inside the served band 79% of the time.

By park (ride-wait predictions)

ParkMean absolute error (min)Graded predictions
Magic Kingdom9.135,929
EPCOT7.833,552
Disney's Hollywood Studios9.931,719
Disney's Animal Kingdom5.330,412

Biggest misses, last 30 days

The largest ride-wait prediction errors on record in this window — included on purpose, not filtered out.

DateParkRidePredicted (min)Actual (min)Error (min)Lead (days)
2026-08-06 16:00 Magic Kingdom Space Mountain 39 171.3 132.3 4
2026-08-06 16:00 Magic Kingdom Space Mountain 39 171.3 132.3 7
2026-08-06 16:00 Magic Kingdom Space Mountain 41 171.3 130.3 3
2026-08-06 16:00 Magic Kingdom Space Mountain 42 171.3 129.3 2
2026-08-06 16:00 Magic Kingdom Space Mountain 43 171.3 128.3 6
2026-08-06 16:00 Magic Kingdom Space Mountain 43 171.3 128.3 1
2026-08-07 12:00 Disney's Hollywood Studios Star Wars: Rise of the Resistance 57 184.2 127.2 4
2026-08-06 14:00 Magic Kingdom Space Mountain 40 163.3 123.3 4
2026-08-06 17:00 Magic Kingdom Space Mountain 41 164 123 3
2026-08-06 14:00 Magic Kingdom Space Mountain 41 163.3 122.3 7
2026-08-06 14:00 Magic Kingdom Space Mountain 41 163.3 122.3 6
2026-08-06 17:00 Magic Kingdom Space Mountain 42 164 122 7
2026-08-06 17:00 Magic Kingdom Space Mountain 42 164 122 6
2026-08-06 17:00 Magic Kingdom Space Mountain 42 164 122 4
2026-08-06 14:00 Magic Kingdom Space Mountain 42 163.3 121.3 3

How this is measured

These numbers come straight from SupaPark's forward-grading log (forward_accuracy): every prediction SupaPark made is later compared to what actually happened and scored, model_version v1 only — the model version actually SERVED to users. (SupaPark also runs an internal pre-blend research lane, v1_pure, used for model development; it is not included on this page.) "Predicted" and "actual" are the ride's predicted vs. observed average posted wait in minutes at that hour; "error" is the absolute difference. Park-level rows use the same predicted-vs-actual grading applied to park-wide forecasts.

Download this data

data.csv (biggest misses)data.json (full summary)

Free to reuse with attribution — see the license.