I used to have more confidence in the crowd calendars in the past but things have changed so much that I have found nd them to be pretty unreliable over the last few years.
I don't think you can trust crowd calendars anymore. It's always crowded and the holidays and special events are super crowded.
True but even looking into how they do it and what they take into consideration and what their calendar is 'really showing' they still are rather unreliable to me. I don't agree you cannot compare them, but you have to look at what they are saying is a 'crowd' and take that into consideration.They use different methodologies so comparing one to another is like comparing apples to oranges.
@lentesta no doubt there are varying levels of data to pilfer through when devising one's own "crowd predictions" some do it better than others. My original post is driven by the inadequacies found when viewing different websites. I appreciate the complex algorithms used, just as much as someone's veteran experience in the parks. As someone who doesn't pay for the research found on TP's page, I suppose I am missing some of the supplemental information which may give me an edge when it comes down to when and where to visit. Thanks for the post.As TouringPlans is my site, I am naturally biased. Take the following as you will.
Most crowd calendars are bad because the website's owners have made a good business decision.
It's taken us the better part of 10 years and well over $1 million to build our crowd calendar platform. That $1MM includes two professional statisticians/data scientists, programmers, specialized software, and a lot of Amazon cloud computing power.
If you run a Disney website and you want a crowd calendar to attract people to your site, the smartest thing you can do is spend $10K on an SEO consultant to rank highly for the terms "disney crowd calendar", and throw together something in Excel.
The stupidest thing you can do is try to build your own, real, crowd calendar, with actual predictions that people can verify. You'll spend untold time and money doing it, and in the end it probably won't be as good as what we have. You will fail, it will take a long time to fail, and it will hurt the entire time you're failing.
If predictions are not your core business, you should not try to make them.
The only two sites ever to produce a verifiable crowd prediction - an actual crowd calendar whose accuracy people could check - are us and Josh's *******. I don't think Josh is doing those now. I'm reasonably sure that when he was, he had access to some form of Disney's internal crowd projections and/or attendance numbers, along with his own massive time commitment to the thing. So the two sites who've ever produced crowd calendars were backed by literally millions of dollars in investments and teams of people.
That's why most crowd calendars are bad.
I will finish this post with two images. One is a slide from a presentation I gave to UCF's Data Science department earlier this year. It shows all of the alternative prediction techniques I tried in 2015-2016 to beat our stats team's existing crowd calendar models. All of them failed to beat our current models.
View attachment 224482
The second is a screen cap of our spreadsheet tracking our crowd calendar's accuracy. The stats team, our staff in the parks, and I have a standing meeting on Mondays to review the crowd calendar's performance. This is what we start with.
View attachment 224486
There are lots of things one can debate about the crowd calendar: the attractions we pick to measure, the distributions of their 1-to-10 scales, the time of day we include, and so on.
However, the predictions for 2017 are +/- 1 about 70% of the time - roughly 5 days out of 7. When you consider that we can't predict rain, ride breakdowns, Disney staff cuts, or, you know, whether the posted wait times are even real, that seems like a reasonable level of accuracy.
Register on WDWMAGIC. This sidebar will go away, and you'll see fewer ads.