Using The Crowd Calendar

by 24 Comments

  • Facebook
  • Twitter
  • Google Plus
  • Pinterest

Well, it has been more than a year since the launch of the new Crowd Calendar 2.0 and we at TouringPlans.com have found ourselves taking a little refresher course in how it works and how to use it. It reminded us of a few things that we thought we should share.

What is it?

The Crowd Calendar is a list that ranks from 1 to 10, the wait times that we predict will occur on every day of the year. A ’1′ means that the wait times are predicted to be in the bottom 10% of all days, a ’2′ would be the bottom 20% of all days, etc.

What does it show us?

Since the crowd calendar is a relative rank, we can tell how one day compares to another. A ’6′ is more crowded than a ’5′ which is more crowded than a ’4′, etc. The neat thing about setting the calendar up this way is that it allows us to compare any two days throughout the year on an equal playing field. So a ’2′ in September has roughly the same predicted wait times as a ’2′ in May. The calendar can be used to answer the question “Which day has higher wait times?”.

What does it NOT show us?

The numbers in the calendar do not represent the relative magnitude of the crowd. So a ’4′ is not twice as crowded as a ’2′. The average of a ’7′ day and a ’3′ day is not a ’5′. Similarly, an increase or decrease in the crowd level index may not represent a change in size of the crowd but rather a change in the ranking. It also doesn’t work for comparing two parks against each other because a ’6′ at the Animal Kingdom represents a completely different size of crowd than a ’6′ at the Magic Kingdom.

But the rankings are based on wait times, aren’t they?

Yes but there is an important distinction between a day’s rank and its magnitude. Think of it like the Top 40 music charts. The Top 40 songs are a rank based on sales of records but we can’t say that song number 8 on the charts sold twice as many as song number 16. We can say, however, that it sold enough to finish 8 places higher on the rankings.

How should we use the calendar?

The calendar is set up to answer this question: “Between two days, which day is predicted to have the lower wait times?” We have found through the years that predicting the size of the crowd is not as important as predicting the difference between two crowds. We like to think of ourselves as your personal shopper at the grocery store. We want to save you $20 off a coffee maker just as much as we want to save you 10 cents off a jar of peanut butter. Picking a ’9′ day instead of a ’10′ day may mean a much bigger gain in wait times than picking a ’2′ day instead of a ’3′ day, but it is all relative.

This also means that unless you care about the 10 cents, maybe a ’3′ day isn’t that bad.

For those that are interested, the distribution of wait times throughout the year is bunched up between the rankings of ’3′ and ’8′. The difference in wait times in this range is actually fairly small. The wait times take a big jump when you get to levels ’9′ and ’10′ though, as anyone visiting at Christmas or Easter can attest. We made some adjustments to our methodology this year given this distribution and we have seen some significant gains in our accuracy. On average, the crowd figures on our site are within 1 index point 90% of the time.

  • Facebook
  • Twitter
  • Google Plus
  • Pinterest
Posted on October 18, 2011

24 Responses to “Using The Crowd Calendar”

  • Thanks for posting a good reminder, especially the 90-percent accuracy note. The fairly substantial changes to crowd levels for October 2011 (induced by Disney’s unusual schedule changes) show how the crowd calendar is a useful tool but not a crystal ball, especially for those who might use it to plan dining reservations at “Best Parks” 180 days before a trip.

    Also, the gremlins were at work last night (10/18), yielding some wild data shifts–e.g., Epcot from a high ’8′ to a high ’2′ on 10/22/11, and Magic Kingdom listed as a Best Park on 10/20/11 even though it has morning EMH. Happily, order has been restored today (10/19), and apparently no Crowd Tracker updates were generated by the glitch. But a user looking at the data last night might have been thrown off track by a temporary data problem.

    All of which shows how it’s wise to check crowd levels regularly, sign up for Crowd Tracker updates, and allow some leeway in planning based on the very useful (but not fail-safe) tool offered here.

    • What were “Dinsey’s unusual schedule changes” that significantly affected October’s predictions?

      • As understood from various sources…there was some back-and-forth on Fantasmic! dates/times, with the end result being more shows than was previously the norm. More important, MK opens at 8am thru the coming weekend. Per the TP entry for 10/22: “Magic Kingdom is now open from 8am until 1am, which is considerably longer than we expected, or normal for October.”

        • It’s my understanding that WDW adds shows or extends hours in response to increased crowd levels. (For example, more people are booking onsite stays so WDW increases park hours to handle the larger crowd.) Since most WDW visitors plan their trips months in advance, adding shows or changing park hours should have relatively little impact. Stated another way, added shows and longer hours are symtoms of increased crowd levels, not causes of increased crowd levels. (If I was not going to WDW in October, changing the MK’s closing time from 10 PM to 1 AM is not going to cause me to change my mind.) The start of October had the MK’s 40th anniversary. Even though this was not well advertised, did this bring in more people than expected? Is there some other reason why October is seeing an increase in attendence?

          • You make an interesting point Barney. Given your logic, you could say that our crowd estimates are based on the symptoms, not the causes. We do not have access to Disney’s internal booking numbers so we rely on the auxiliary information like park hours and schedules.

  • But by claiming you are within one point, you are actually using a three point spread. In saying it will be a “5″ you have actually hedged your bet through 4, 5 & 6. So 30% of all numbers are actually covered. Thus giving you a one in three ratio of being right. I’ve also noticed that you’ve stopped creating actual vs predicted posts on social media in the last few months. Your crowd calendar adjusts with park hours updates, meaning it’s not actually your prediction that’s correct, but a recognition of Disney’s hours.

    • Johnny, I’m pretty sure you’re describing a 2-point spread. And while that covers 20% of all numbers, it doesn’t automatically give a 1-in-5 chance of being right.

      • Obviously, you’re a statistician and I’m not, but if you are allowing yourself to be within one number (index point) of your choice, either low or high, then you can cover three numbers. If you claim it will be a 7, then your allowable range for correctness will be one number lower and one number higher than actual. So, you are allowing yourself 6, 7 & 8 as potential correct numbers. In math that is 30% of 10 potential, not 20%. You can’t ignore your original number as a point of discussion. Unless you miss dramatically, you’ll be within your target range.

        • You are right Johnny, the 90% confidence interval covers a range of 3 points but that is not how we really measure “correctness”. We want our estimate of the crowd level to be within a couple percentage points of what we end up observing. The confidence interval is a statistic that helps us determine what our error rate is, within a conservative range.

          Also, your statements seem to imply that the chances are 1 in 3 that our estimates are within the range but that is not the case. If we were picking the crowd level randomly out of a hat (which some think we do) then yes, a range of 3 index points would be correct 30% of the time. In fact, as the confidence interval tells us, our chances of being within the range are 90%.

          • Fred, I didn’t nor wouldn’t imply that you are choosing a random number. I realize there is work involved. I’m sure that each of you are quite brilliant in your fields. I was merely pointing out the range of numbers involved for an accurate picture. The plans are great and work well, I just can’t rely on crowd predictions that change one month out or even week or day of. I’ll just choose the days I like and use a good plan.

  • Steve, aren’t the”best parks” generally the ones with no extra magic hours?
    I think your touringplans are effective, but the crowd calendar is sketchy, as it changes often.

    • johnny, it seems that Touring Plan canon does not allow a park with EMH to be a Best Park, which was the tip-off that there was likely a data glitch. As mentioned, the temporary problem was resolved by the TP folks–it was just an example of how it’s worthwhile to be familiar with the crowd-calendar approach and thus occasionally skeptical.

      Interesting hypothetical question, though…if MK was a 1.0 on a day with EMH, and the other parks were at high crowd levels, would TP violate canon and make MK a Best Park? A fair guess is that the other parks would never have sufficiently high-level crowds vs. MK to warrant that call…but it wouldn’t be a surprise to learn it could happen.

      • Actually Steve, this is complicated question. Our algorithm is filled with several rules and exceptions so it is difficult to determine what would happen in certain situations. Our rules about the Extra Magic Hour schedules take a high priority so it would be unlikely that it would be overwritten.

    • Although we might disagree with the qualifier “sketchy” I understand your point and I agree! Put another way, our crowd calendar estimates are subject to error but our touring plans are 100% guaranteed to save you a ton of time in line. Ideally, we hope that our users consult the calendar to either choose a time of year with lower crowds or to prepare for days that might be really busy but then stick to the touring plans whenever possible.

  • If I correctly understand Crowd Calendar 2.0, the average should be approximately 5.5. A couple of months ago, it was close to this. However, the average has been inching upwards and is approaching 6.0. For example, only 19 days are listed as “1″ while 44 days are listed as “10″. Shouldn’t there be 36 or 37 days for each crowd level?

    • Actually Barney, the average by definition is 5.0. But recall that this is a ranking, so technically the idea of an average is not applicable. For example, on a list of the top 100 songs of all time what is the average rank? Also, the ranking is based on all days since 2006 so when we see more days above 5 than below that means that crowd levels are on the rise compared to previous years. Don’t worry though, we are talking about a rise in the order of parts of a minute per attraction.

      • Thanks for the reply. Given that the scale is 1 (not 0) through 10 and, if I correctly understand the definition of the Crowd Calendar 2.0, 10% of the days should be assigned a “1″, another 10% assigned a “2″. etc., the average value should be 5.5. Using the old Crowd Calendar measurement, it would be theoretically possible to have all “10″ or “1″ days because the old Crowd Calendar used an absolute measurement. However, the new Crowd Calendar uses a relative measurement, meaning the 10% most crowded days should be a “10″, the next 10% should be a “9″, etc. Since the Crowd Calendar covers 366 days (367 in a Leap Year), there should be 36 or 37 days in each of these ten “buckets”. However, this is not the distribution of the current calendar. I recognize that crowd levels have risen since 2006 but this should have no bearing on the forward looking Crowd Calendar 2.0, whose intention is to let the reader know which 10% of the days are going to be the most busy all the way through which 10% of the days are going to be the least busy. Am I misunderstanding the definition of the Crowd Calendar 2.0?

  • [The calendar is set up to answer this question: “Between two days, which day is predicted to have the lower wait times?”]

    THIS! This helped me so much last year when I was in Sarasota on spring break and wanted to take a day trip to Epcot and maybe spend a few hours at the Studios. It was during a very busy week in March (resort-wide crowd levels of 9′s, 10′s) and I knew I could either go Tuesday or Wednesday of that week. Wednesday showed Epcot as the recommended park and a park specific crowd level of only 7, while the other parks were 9′s and 10′s. The day was fabulous and I found Epcot to be so much less crowded than I remember from the previous year, same week. I totally attribute this to picking the “right” day out of a very crowded week. THANK YOU :)

  • During the busiest holiday weeks, the parks sometimes approach capacity and stop admitting guests. Historically, how have Touring Plan’s Per-Park-Crowd-Levels correlated with these closures? For example, I suspect that a 9.8 at the Magic Kingdom may indicate a potential closure. But what about a 9.2? Or what if the 9.8 is at Hollywood Studios? Can the crowd calendar be used as a indicator of potential closure phases?

    • This is a great question Karen, one that we get occasionally – might make a good blog discussion on its own. The parks reach capacity only on the busiest of days, and usually only at the Magic Kingdom. Given that, and including our error rate, I would say that at the Magic Kingdom and Epcot when our crowd estimate is 9.9 or 10 we would expect the park to reach one of the four levels of closure due to capacity. At 9.6 to 9.8 the chances would be something like 50/50. Lower than that and the park should remain open all day. At the Studios and Animal Kingdom, even at a crowd level of 10 the chances of reaching capacity would be 50/50 at worst, it doesn’t happen very often.

  • Can you post the actual crowd level distribution? You may have to obscure the detail if it’s too much proprietary information.

    For example, y axis is the actual crowd (numbers in the park) or peak wait times or some concrete measure of the crowd, and x axis is the days of the year ordered by crowd level percentile.

    Basically, you’ve described it in your article, but it would be easier to explain to others with the visual.

    (Other people I talk to want to interpret the crowd calendar level to be an indicator of how crowded it will be not just an indicator of the relative crowd level compared to other days. It’s mainly just because I say things like ooooh that’s a good day. It’s crowd level 3 and not crowd level 6. Then we get there and there really isn’t much difference between a 3 and a 6. She uses low crowd levels as an excuse not to use a touring plan to minimize waiting. Because she really hates too many extra steps.) So, I need to show that even on crowd level 3 days, a good touring plan up to noon is probably very helpful.

  • Thanks! That helps point out which days an afternoon break may or may not be feasible for my group. Perhaps Kristen might want to blog about this further since she has foolishly selected the same over populated week to visit. One more question while I have your attention — does Lines indicate when a closure phase has been implemented? Could it?

  • My husband is an Industrial Engineer and when I showed him the crowd calendar and wait times data before our trip this past May he was really excited. I used both the crowd calendar and the average wait times based on crowd calendar rankings to personalize a touring plan for us and for Memorial Day weekend our longest wait for an attraction was 20 minutes. 4 thumbs up from this family. LOVE IT!

Leave a reply, your thoughts are welcome!

Want a cool avatar next to your comments? Add one at Gravatar