Lines App Question re: Inaccurate Wait Times & Causes

Walt Disney World
121

Ah, yes. The intraday process runs every 10 minutes and adjusts today’s forecasts based on what’s happening in the parks.

The intraday process attempts to do this in two ways:

  • With the posted wait times we get from Disney
  • With actual waits we get from y’all

If we have just posted waits from Disney, we compute the difference between the most recent posted wait and the most up-to-date prediction of that posted wait, to nudge the predictions towards what we’re seeing in the park.

Those predictions are in 15-minute increments, so we predict the posted and actual waits at (for example) 1:00 pm, 1:15 pm, 1:30 pm, and so on. We use linear interpolation (i.e., “draw a straight line”) to get predictions between 1:00 and 1:15. So if we need to compute a wait at 1:08 pm, it’s gonna be pretty close to the middle of whatever the numbers are at 1:00 and 1:15 pm.

When we’re dealing only with posted waits, we use the same actual-to-posted wait ratio for that 15-minute segment that we calculated when we built the models. So if the models say the actual wait is 80% of the posted between 1:00 pm and 1:15 pm, and all we have is posted waits to go on, then we use that 80%.

If we get an actual wait, we re-compute the actual-to-posted ratio and use that instead, for the next 30 to 60 minutes (I don’t recall the exact duration, since I didn’t touch this part of the code).

That 30-60 minutes is our “short delta” - it allows us to respond to sudden changes in wait times, but treat it initially as a temporary thing. It’s primarily useful when rides break down and come back up, or for unusual things like “giant tour group gets in line at Toy Story Mania.”

We also compute a “long delta”, which is the difference from park open until now. We use the long delta for adjustments more than an hour from now.

I toyed with a “very short delta” over the fall, trying to see if using just the most recent wait time for just the next wait time would improve things. It didn’t show a clear advantage so I didn’t do it.

It turns out that the most important things for the intraday adjustments seem to be:

  • Getting the initial actual-to-posted wait ratio correct
  • Running the intraday adjustments as many times per hour as possible

We model around 300 attractions around WDW, DLR, and UOR. Around 200 of them are “core” attractions, like Slinky, Pirates, etc., with posted wait times that vary significantly.

Each of those 200 attractions starts the day with anywhere from ~60 to ~80 wait-time predictions (we include a bit of time before and after official park opening, in case Disney changes park hours during the day itself).

So every 10 minutes we’re looking at around 14,000 - 20,000 pieces of data to figure out what’s going on in the parks right now. Getting that to happen as fast as possible is something I’m going to work on in 2026.

7 Likes
122

yowch thats pretty amazing!

Really interesting thnaks, im going to reread the last few posts as its complicated…

2 Likes
123

Hmm. You seemed to have left out the “thumb in the wind” part of the process. Weird.

:nerd_face:

2 Likes
124

Sometimes I’m so deep into the process and its chain of assumptions that I’m just amazed it works.

2 Likes
125

The new wait-time forecasts are being loaded to production now. They should be available starting tomorrow morning. We’ll still release the new intraday process tomorrow.

I’ll monitor this for the remainder of the year.

:folded_hands: Thank you all for your help and comments over the last couple of weeks. I love this community. :heart_eyes:

7 Likes
126

Thanks for all your work! Making it easier to make the most out of precious park time. Looking forward to testing your new work in mid january.

1 Like
EPIC Low Crowd on 12/28?
127

Amazing data & process. My family will be back in the parks 12/28-1/2 and will input our waits as much as possible. Happy Holidays all!

2 Likes
128

Hello. I’m wondering if there’s any update on this? I have refreshed my touring plan and no go for wait times being accurate. Do I need to make an entirely new plan for these to update?

1 Like
129

Hi Jennfier-

We’ve updated the wait times. I’ve also looked at your Magic Kingdom touring plan and it seems reasonable. Let me know what appears off, please, and I can follow up with specifics.

2 Likes
130

We Have Spreadsheets!

I’ve put into production a report that shows us, for each Liner-submitted wait time, whether our prediction of the actual wait was closer than Disney’s posted wait. My thinking here is that that’s one of the main things people pay us for.

Our Goals For Predicting Actual Wait Times

I think a reasonable goal to start is that 70% of the time, TP’s estimates should be closer to your actual wait than Disney’s posted wait, and eventually getting that to 75%.

  • Another 6% of the time it looks like we tie with Disney on accuracy. For example, if we both say the wait is 5 minutes for Mad Tea Party, whatever the wait time is, we’re going to tie.
  • Disney’s knows ahead of time when rides are about to break down or come back up. It takes us 10-20 minutes to figure that out.
  • Some folks may use slightly different criteria for determining when to start and stop the timer.

How Accurate Are TP’s Wait-Time Predictions?

Going back to May, 2025, we’ve received 45,197 actual wait times from Liners in the parks.

May 2025 is also when we put the new wait-time models into production. Here’s where we’re at with accuracy:

  • Animal Kingdom: 61% of the time TP’s estimates were more accurate, with 5% ties

  • EPCOT: 59% (5%)

  • Hollywood Studios: 60% (5%)

  • Magic Kingdom: 66% (6%)

  • Disneyland: 52% (7%)

  • California Adventure: 54% (6%)

  • Epic Universe: 48% (3%)

  • Islands of Adventure: 53% (5%)

  • Universal Studios Florida: 53% (5%)

    The report is color-coded like this:

    • 70%+ accuracy is green
    • 50-70% accuracy is yellow
    • Below 50% is red

Accuracy By Ride and Time of Day

The report also tells us, by ride and by time of day, who’s more accurate. Here’s the Magic Kingdom as a whole:

image
image1920×1100 286 KB

So for the first two hours the MK is open, our wait-time estimates are significantly better than Disney’s posted waits.

That makes sense, since we’ve been focusing on “first two hours the park is open” since 2023. So that’s good to see.

I suspect, but I’m not sure, that the middle part of the day is affected by ride breakdowns. And haven’t looked at what’s going on at the end of the day.

Big Rides Seem Better

On a ride-by-ride basis, we seem to do better at popular rides. Here’s Big Thunder at Disneyland:

image
image1920×1100 250 KB

So TP’s estimates are more accurate than Disney’s around 75% of the time throughout the day. But the period from 30 to 60 minutes after the park is open, Disney’s is better.

I like that “timeslice view” because it tells us exactly where the models need adjustment.

Conversely, we don’t do as well with less-popular rides, like Roger Rabbit’s Cartoon Spin and Snow White at Disneyland:

image
image1920×1100 261 KB

We’re Too Pessimistic At Epic Universe

The other thing I got out of the report is that we’re being too pessimistic in predicting UOR waits, especially at Epic U. Part of that is because our models include the period when the park just opened, and the operational issues that came with that. Battle at the Ministry, in particular, needs to come down.

As a side note, I’ve had some great discussions with Universal’s engineers over the past month, and I think we understand more now about where the park is in terms of capacity and reliability.

Universal’s in the middle of putting in some new technology that should help them improve the accuracy of their own posted waits. So that will also help.

Using This Report to Adjust the Models Going Forward

Our last set of model updates was December 17, 2025, for WDW and UOR. I’m going to let those run a full month, until the middle of January, then make adjustments.

New wait-time models went in today (January 1, 2026) for Disneyland Resort, so I’ll probably make those adjustments in February.

Questions and Comments Are Welcome

Let me know if none of this makes sense, please. Here’s a link to the report if you want to drill down on it.

13 Likes
131

What a wildly exciting New Years Eve you had Len.

Only kidding, if I made New Years Resolutions - seeing more fun data in spreadsheet form would be one of them.

I will be making my coffee and then sitting down to look at that report link - thank you for sharing it!

4 Likes
132

I love this level of analysis! Thank you for prioritizing that. I agree that accurate wait time info is likely the biggest draw for customers of TP.

5 Likes
133

oh! You work so hard and this means nothing to me!
I’m just so glad it makes sense to all of you and you direct my days so nicely!

5 Likes
134

There’s the true appreciation!

2 Likes
135

Awesome thank you! If those are accurate, that’s great. I was just bracing for long wait times over March break and the longest one by quite a bit is 42 minutes. That just didn’t seem right to me but is great if it is! :slight_smile:

1 Like
136

I am at a Disney resort and heading to the airport tomorrow. As anyone would predict the crowds and wait times were bonkers this week.

But that leads me to a thought- I think some people are saying “your wait times are off” because when they made their plan the times are different than what they actually see that day.

Background:
This week I tended to use TP in two ways 1) to see what the predicted times were using the “parks” section of the app and 2)looking at my plans to remind myself of what my plan was.

I never reoptimized and I occasionally reevaluated. I am a frequent park goer so my TP is my rough guide, not my holy grail.

Is this new?:
I really liked the fact that when I viewed my plan I could see what TP predicted the wait would be at the time I made the plan, as well as what TP was currently predicting the time would be. That current prediction seemed pretty accurate (basically some bannanas amount of time) bit the “in real time” waits were much higher than the predicted waits when I finalized my plans after making my LL.

My point:
Are people responding to this difference when they say times are inaccurate? It is a more nuanced version of “I printed my TP”. In the app they are comparing predicted predictions to real time predictions and the two don’t line up. To ME this is for obvious reasons - but to someone who staked the success of their vacation on the predicted predictions, this might be the source of their complaint.

7 Likes
137

That’s a great observation!

We tell people to re-optimize once in the park because some rides don’t open on time and others break down. Tour groups show up out of nowhere.

It’s not possible to predict all of that in advance, so the intraday process helps with that.

2 Likes
138

It may be a case of too much information. The person can see right there the difference between the two. Expecting them to be the same would be expecting magic, but who knows.

Is the evaluate different lately? I thought I remebered it asking me if I wanted to evaluate for the current time and then from there out it was all “real time”.

2 Likes
139

I don’t think it’s different, but I haven’t evaluated on the app in a while.

1 Like
140

I have a new phone and suspect my location sertings changed and it didn’t know I was in the park, which makes a difference too.

1 Like