Over the past six years, we have collected over 7 million wait times. Thank you to all the Lines users that have contributed to that monumental statistic.
Besides using a day’s wait time data to predict future wait times and crowd levels, we also use a day’s data to validate our predictions for that day. And we’re the only crowd calendar to show you how our predictions worked for each day. In this blog post, I will explain how we calculate the “predicted” and “observed” wait times and crowd levels for our calendar.
First, let’s cover the methodology we use to convert wait times to crowd levels. Our crowd levels are based on Disney’s posted wait times at specific attractions, between 10 a.m. and 5 p.m. Crowds at Disney parks are usually largest between 10 a.m. and 5 p.m., so that’s the best time to get a consistent view of their size. Also, every park is open between 10 a.m. and 5 p.m. every day of the year, so it’s a consistent measure.
The attractions we choose are among the parks’ most popular, and most are able to handle large crowds. Because they’re popular, it’s a good bet that high wait times mean high crowds, and low wait times mean low crowds. For example, if we see a 15-minute wait at Space Mountain at noon, it probably means that crowds are low, not that Space Mountain suddenly became less popular.
Just like the crowd calendar’s 1-to-10 scale, we score each attraction’s daily crowds on a scale of 1 to 10. After the attractions are scored, those values are combined to make the park and resort crowd levels.
Now let’s look at an example:
Before the day starts, we predict the posted wait times at each attraction. Below is what we predicted for Space Mountain on July 9, 2015.
We predicted that the average posted wait time between 10 am and 5 pm for Space Mountain on July 9, 2015, would be 65 minutes. Next, we look up the number 65 in our distribution of average wait times for Space Mountain on the table below. There we see that a 65-minute average wait falls into the range of 59 and 67 minutes representing an ‘8’.
|Attraction||Level 1||Level 2||Level 3||Level 4||Level 5||Level 6||Level 7||Level 8||Level 9||Level 10|
|Big Thunder Mtn||<23||23-26||26-28||28-32||32-35||35-38||38-41||41-45||45-49||>49|
|Great Movie Ride||<15||15-17||17-18||18-20||20-21||21-23||23-25||25-27||27-31||>31|
|Kali River Rapids||<12||12-14||14-18||18-24||24-30||30-36||36-42||42-50||50-60||>60|
|Living w/ Land||<12||12-14||14-15||15-17||17-18||18-19||19-21||21-23||23-27||>27|
|Msn: SPACE Orange||<16||16-18||18-20||20-22||22-24||24-27||27-29||29-35||35-43||>43|
|Peter Pan’s Flight||<42||42-44||44-47||47-52||52-56||56-61||61-66||66-71||71-77||>77|
|Pirates of Caribbean||<11||11-13||13-14||14-16||16-19||19-21||21-23||23-28||28-35||>35|
|Tower of Terror||<34||34-37||37-42||42-47||47-53||53-59||59-65||65-72||72-80||>80|
|Toy Story Mania!||<57||57-62||62-65||65-69||69-73||73-77||77-81||81-87||87-95||>95|
|Winnie the Pooh||<24||24-26||26-28||28-29||29-31||31-32||32-34||34-35||35-38||>38|
The above chart shows all the posted wait times collected for the July 9. The average observed wait time between 10 a.m. and 5 p.m. is 75 minutes. Wait times between 67 and 77 are a ‘9’. If you look at the forecast table at http://touringplans.com/magic-kingdom/wait-times/date/2015-07-09 you will see all the predicted and observed wait times for all the attractions at the Magic Kingdom on July 9. The attractions that are used for the Crowd Calendar will have a crowd level score next to the average wait times.
|ATTRACTION||AVERAGE WAIT TIME||CROWD LEVEL|
|Big Thunder Mountain Railroad||48||43||9||8|
|Buzz Lightyear’s Space Ranger Spin||40||38||8||8|
|Peter Pan’s Flight||71||72||9||9|
|The Haunted Mansion||31||36||8||9|
|The Many Adventures of Winnie the Pooh||31||32||5||7|
Based on the predicted and observed data, the predicted crowd level for the Magic Kingdom was an 8, and the observed crowd level is a 9. The top of Crowd Calendar page will show how good our predictions were on the prior day. To see the details on any day, use the “Jump to Date” on the left-hand side of the Crowd Calendar page. To see all attractions’ predicted and actual wait times click on a specific park. To see only the attraction that feed into the crowd level choose “Click here to see individual attraction details used in determining crowd levels.”
We would like to say that we are 100% accurate with our predictions. From a statistical standpoint, this is an imposable goal. As you can see from the above chart, there is a lot of variation in submitted wait times. We do identify and exclude submissions that appear to be inaccurate. Some of the variation could be explained by effects that we cannot be aware of when the predictions are made. Some examples of these effects are weather, breakdowns, and schedule changes. For custom touring plans, we adjust the attraction predictions for day-of effects, but it is too late adjust the Crowd Calendar.
People may ask why we are focusing so much on posted wait times, when what is important is the actual wait time. The answer is that we are leveraging the information we have. Out of the millions of wait times we have, less than 1% of the data are actual wait times. Fortunately, the 1% gives us a way to predict the actual wait time based on the posted wait time. Another problem with that actual wait times we do have is that they are usually for times of day when it is optimal to ride an attraction. We do not have much data for actual wait times for the longest wait times of the day. For example, over the past two years, we only have two actual wait times for Soarin’ where the wait time is over two hours.
Our goal is to have our crowd level index within one point 90% of the time. Over the past year, we are within one index point 78% of the time, and within two index point 90% of the days. The higher than expected crowds has hurt our accuracy rate. Fred Hazelton will have a blog post coming up shortly, looking at possible explanations for why we’re seeing higher crowds in September.
We strive to have the most accurate predictions, and we are the only Disney World Crowd Calendar that publishes its accuracy.