How this works, in plain language.

Every five minutes, we record the posted wait time for every ride at every park we cover, via Queue-Times.com. Posted means the number on the sign and in the park's app — what the park chooses to tell you. It is not always the same as how long you'll actually stand in line, which is exactly why the history is worth keeping.

Alongside waits we record operating hours, weather, and the calendar context (holidays, school breaks) that makes crowd patterns make sense.

A recording day only counts for a park if our recorder verifiably captured at least 97% of that day's polls for it. We call those clean days. A day with an outage — ours or anyone's — doesn't get to shape the statistics, and a day we can't verify is treated as dirty, not innocent.

One definition, used everywhere: the charts on this site, the charts we post, and (soon) the data our forecasts train on all use the same clean-day ledger.

A ride's typical wait for, say, 2pm is the median of every posted wait we've recorded for that ride between 2:00 and 2:59 local time, while the ride was open, across clean days only. Median, not average — one broken-down afternoon shouldn't drag the number around.

"Typical" is a description of recorded history. It is not a prediction about today. When we start making predictions, they'll be labeled forecasts, and they arrive only after the test below.

Small samples lie confidently, so the site declines to render some things until the data has earned them:

• Typical-wait charts need at least 14 clean days of history for that ride.
• The crowd score (0–10) needs at least 14 clean days of park history, and enough rides posting waits right now to make a fair comparison.
• Day-of-week comparisons need at least 4 clean days of each day type before we'll say "Saturdays are worse."

Where a chart would be, you'll see how many clean days we've banked instead. It unlocks when it unlocks.

The score compares right now to this park's own recorded history: for every open ride with enough history, we take the ratio of its current posted wait to its typical wait for this hour, then take the median ratio across rides. Running exactly typical scores 5/10; double the typical waits scores 10; half scores about 3. It's a comparison, not a forecast.

Every number on this site shows its age, because posted waits go stale silently: parks stop updating signs during downtime, at closing, or whenever the intern goes to lunch. If a park hasn't updated a ride's number in over two hours during operating hours, we badge it — the number is still on the sign, but you shouldn't plan your afternoon around it.

We're building wait forecasts. They ship when they beat an honest baseline on data they've never seen — specifically: at least 10% better median error than "assume the wait stays what it is now," tested walk-forward across our top-30 regional parks. Until then, nothing on this site is a forecast, and the Scoreboard stays honestly empty. Every forecast we ever issue gets archived as-issued, so accuracy claims can always be checked against what we actually said at the time.