How the forecast works

Crowd levels are derived from the structure of the calendar alone. We hold no historical attendance data, so this is not a model trained on past visitor numbers. Here is exactly what goes in and what does not.

What we use

The list of Japanese public holidays published by the Cabinet Office, through 2027.
Measured crowd levels from queue-times.com (Sept 2024 to Aug 2026, both parks). The order of the weekday, season and school-break effects is set to match it.
The day of the week.
The shape of holiday runs: how long, which day of it, and whether a workday follows.
Japanese school holiday periods (summer, winter, spring).

How it is calculated

A starting score: 8 for Saturday, 7 for a public holiday, 6 for Monday, 5 for Sunday, 5 for Friday, 3 for Tuesday to Thursday. The order comes from the measured data; the spacing is our own.
Then adjustments: +1 inside a run of three or more days off, a further +1 in the middle of such a run, −1 on its final day, −1 when the next day is a workday.
The month moves it too: +1 in February, March, November and December; −1 in January, May, June, September and October. Only months where two separate years agreed are moved. July swings by more than 40 points between years, so it stays at 0.
The three school breaks are treated differently. Spring break is +1; summer break is 0, because the measurement shows almost no effect. Over the New Year period and Obon week the weekday effect disappears, so the score is replaced with 8 rather than added to.
The result is clamped to 1–10 and shown as one of seven words. Every date page lists the individual adjustments, so any number can be traced back.

Room rates are calculated separately

Park crowds and hotel prices move for different reasons. The park fills up when the day itself is a day off; room rates rise when the *next* day is.
That produces days when the park is packed and the room is still cheap — the following day is a workday, so demand for the night before is weak. Because we model the two separately, we can point those days out by name.

What we do not use

Weather. Forecasts only reach about two weeks out, which is no use across six months.
Events and show schedules. We have no reliable machine-readable source for them.
Ticket prices, for the same reason.
Historical attendance figures. They are not published.

About wait times

Wait times come from Queue-Times.com and are live readings. Only the current moment is available, so we show them only on the page for today.
On future dates we show an estimate: that measured snapshot stretched by the crowd level for the day. It is labelled an estimate, not a measurement. The order of the levels is calibrated against measured data; the spacing between them is not, so read the minutes as a guide.

About accuracy

We publish the result of scoring our forecast against 669 days of measured data across two years. The rank correlation is 0.565; simply avoiding weekends and public holidays scores 0.114. So the ordering is better than that, but it is not enough to lock in a single date.
We do not compute a hit rate — the share of days landing on exactly the right level — so we do not publish one.
Crowd levels are our own forecast, not attendance figures.

Who runs this and how to cite it

This site is operated as “Japan Park Crowd.” It is not affiliated with Tokyo Disney Resort or Oriental Land Co. Park and attraction names are used for description only; all rights belong to their respective owners.
The crowd levels here are forecasts, not official announcements. They can disagree with what actually happens. Always check the official park calendar before committing to a trip.
You are free to quote the numbers and charts. Please credit this site by name and link to the page you took them from. You do not need to ask permission.
Wait times come from queue-times.com, opening hours and show times from ThemeParks.wiki, weather from Open-Meteo, and hotel availability from Rakuten Travel. We follow each provider’s attribution requirements.