Our model · Reflections
HOW THE CALLS HELD UP
Side by side — what we predicted, what actually happened
Hit rate
0%
0 of 6 openings
Mean error
335%
n = 6
Closest callerror +44%
PROJECT HAIL MARY
We said $85M – $115M. It came in at $141.0M.
Our smallest miss in the sample. Predicted band sat right on the actual.
Biggest misserror +526%
PROJECT HAIL MARY
We said $85M – $115M. It came in at $613.4M.
Our largest gap between forecast and actual. We surface this not to bury it.
Project Hail Mary
Weekend of March 19-21, 2026
We predicted
$85M – $115M
best guess $98M
Actual
Pending
PENDING72% conf
Wuthering Heights
Weekend of March 19-21, 2026
We predicted
$12M – $20M
best guess $15M
Actual
Pending
PENDING55% conf
Scream 7
Weekend of March 19-21, 2026
We predicted
$8M – $14M
best guess $10M
Actual
Pending
PENDING65% conf
Hoppers
Weekend of March 19-21, 2026
We predicted
$15M – $25M
best guess $18M
Actual
Pending
PENDING60% conf
GOAT
Weekend of March 19-21, 2026
We predicted
$10M – $18M
best guess $13M
Actual
Pending
PENDING58% conf
Project Hail Mary
Weekend of March 19-21, 2026
We predicted
$85M – $115M
best guess $98M
Actual
Pending
PENDING72% conf
Wuthering Heights
Weekend of March 19-21, 2026
We predicted
$12M – $20M
best guess $15M
Actual
Pending
PENDING55% conf
Scream 7
Weekend of March 19-21, 2026
We predicted
$8M – $14M
best guess $10M
Actual
Pending
PENDING65% conf
Hoppers
Weekend of March 19-21, 2026
We predicted
$15M – $25M
best guess $18M
Actual
Pending
PENDING60% conf
GOAT
Weekend of March 19-21, 2026
We predicted
$10M – $18M
best guess $13M
Actual
Pending
PENDING58% conf
Project Hail Mary
Weekend of March 19-21, 2026
We predicted
$85M – $115M
best guess $98M
Actual
Pending
PENDING72% conf
Wuthering Heights
Weekend of March 19-21, 2026
We predicted
$12M – $20M
best guess $15M
Actual
Pending
PENDING55% conf
What we count. The hit-rate above only counts opening-weekend predictions where the actual is in. Tracking entries (cumulative worldwide totals mid-run) sit on a different scale and would inflate or deflate the number depending on the week, so we leave them out. Misses are listed by name on purpose — the best way to keep a forecast honest is to publish where it broke.
Why we publish the misses
CALIBRATION, NOT BRAGGING
Most prediction shops show you their best week. We show you every week. The hit rate above is computed from opening weekends only (tracking entries live on a different scale and would inflate the number). The mean error is the average absolute percentage gap between our best-guess midpoint and the actual three-day take.
If a film blows past our range, we say so. If we missed low, that's on the page too. The point of publishing is not to look smart — it is to keep the model honest.