The Walk-Off Effect

Back on June 21st I stood up a pipeline that’s been quietly polling every ride at Disneyland and California Adventure every five minutes since — wait time, status, land, Lightning Lane state, all of it landing in BigQuery. The plan then was: collect for a few months, then go looking for patterns. That table just crossed 1.29 million rows, and today happens to line up almost exactly with when I’d said I’d come back to it. So I did.

The question that started this whole thing: when a ride breaks down, does the wait actually get shorter once it comes back up? The theory is simple — while a ride is down, some fraction of the people standing in that queue give up and leave. If enough people walk off, the line should be shorter than it “should” be once the ride is running again, at least until it refills.

Finding the breakdowns

For every ride, I found every stretch of at least 10 continuous minutes where its status read DOWN, then looked at the standby wait right before that stretch started and compared it to the wait right when the ride came back, and again 30 minutes later — enough time for the queue to actually respond. That covers every ride with at least 15 qualifying breakdowns since June, which turned out to be most of the park.

The top 10, plus two extras

Ranked by how much the wait dropped in the 30 minutes after recovery. I’ve added Mickey & Minnie’s Runaway Railway below the top 10 since it’s a frequently-asked-about ride that narrowly missed the cut, and Radiator Springs Racers at the very bottom — the one ride in the whole dataset where this doesn’t hold at all.

Space Mountain, the ride that prompted this whole question, tops the list: wait drops from an average of 41.9 minutes right before a breakdown to 17.6 minutes at its lowest point in the following half hour — more than half the line, gone. Rise of the Resistance and Toy Story Midway Mania are right behind it, and the pattern holds across a genuinely mixed set of rides: a dark ride, a steel coaster, a trackless system, a motion simulator. The walk-off effect doesn’t seem to care what kind of ride it is.

Radiator Springs Racers is the opposite story. Its wait doesn’t drop when it reopens — it goes up, from 65.5 minutes before the breakdown to 75.7 right after, and stays elevated for the next half hour. My best guess is that RSR’s Individual Lightning Lane creates a backlog effect that outweighs any walk-off: guests holding an ILL return time queue up the moment the ride comes back, on top of whatever standby crowd was already building. It’s a guess, not a proven mechanism — but it’s consistent with RSR being one of the only rides in the park with paid individual booking on top of standby.

What the curve actually looks like

The table above collapses each breakdown into a few numbers, but the shape of the recovery is its own story. Below, every ride is plotted from 30 minutes before it goes down through 30 minutes after it reopens, averaged across every qualifying breakdown since June. The down period itself is compressed in the middle — its real length varies wildly by ride, from under 30 minutes to well over two hours, so it’s not drawn to scale.

X axis: minutes before/after the breakdown, anchored on the last reading before DOWN and the first reading after OPERATING. Y axis: average standby wait in minutes.

Almost every ride traces the same shape: flat before the breakdown, a sharp drop right at reopening, a further dip for the first five to ten minutes as more walk-offs get counted, then a steady climb back toward baseline as the queue refills. Most rides are most of the way back to normal by the 30-minute mark, but not quite all the way — Grizzly River Run is still nearly 8 minutes below its usual wait half an hour later.

Rise of the Resistance is the most dramatic version of that climb: it doesn’t just recover, it overshoots — from 30.6 minutes right at reopening to 51.6 minutes thirty minutes later, well above its own ~44-minute baseline. My read is that a 96-minute average breakdown builds up a real backlog of guests waiting nearby for it to come back, and they all arrive in the queue within the same few minutes once it does.

And then there’s Radiator Springs Racers, sitting up near the top of the chart the entire time, never dipping below its own baseline once. Every other line on this chart tells the same story in a different accent. That one tells a different story entirely.

What’s next

This is the first real output of the wait-time pipeline, and it’s query code now — scripts/analyze-wait-times.js in the mobile repo, so this is re-runnable as more data comes in (Halloween season starts soon, and I’d bet the shape of these curves shifts once actual crowds show up). The original plan was for this kind of insight to eventually surface inside the app itself — a “this ride just went down, wait 20 minutes and check back” kind of hint. Today was step one: figuring out if the idea was even true. For most rides, it is.