Many have probably seen TP's blogs showing 'what they expected' vs 'what they saw' visa vi crowd levels. Sometimes, these numbers are off. For example, TP predicted a 6 when I was at DHS last week and actually saw a 9. I thought something was off at the time b/c the predicted waits were not what I was seeing that day. Now, I know that thousands of us each day submit wait times and this is how TP figures out the 'what we saw' bit. Does anyone know if those submitted wait times are used real-time to improve personalized TP's? In other words, do we get the benefit of those submitted wait times on the same day? Does this happen when we re-optimize??
That sounds like a question for @Lentesta.
I know that they use wait times of certain "anchor" rides to gauge crowd levels.
Here's an example: http://touringplans.com/hollywood-studios/wait-times/date/2014-07-24
"Attractions with an * and bolded title are a better indicator of park crowd level."
I bet they look at these retroactively to see what the crowd level was/is.
I could be wrong, but I think WDW publishes something about crowd levels too.
@ejj. Fascinating indeed! LOVE this stuff. TP has revolutionized the way I tour. Great company!
Since the models are updated up to a couple times a week, I'm guessing that user submitted wait times aren't applied real-time. So, occasionally, TP's can be off. I guess. The upside seems to be that predictions usually get within minutes of actual, so- on average- TP's work very well.
The data are used 2 ways:
1) When you submit a wait time in a park, we update the forecasts for that park for that day, within 5 minutes. So if you tell us that Big Thunder is 50 minutes right now, we'll use that bit of data to re-forecast Big Thunder for the rest of the day, along with everything else in the park.
The advantage to that is, of course, for the touring plans. Even if we're off by a wide margin on the daily prediction, the in-park forecast usually adjusts within 30 minutes of the park being open. It's actually pretty fascinating to watch.
We have a set of graphs that show how the in-park, daily forecasts get adjusted. I'll see if I can find them.
2) Every night we collect the entire day's wait times and re-run our models for the next 365 days. All parks, all attractions, all days.
@Len I have been sold on TP but now that I know you guys update predictions real-time, I am mega-sold. Very cool and thanks for taking the time to respond (also impressed by that).
@sam2071 don't know if you used TP while in the park, but if you call up the TP on the phone, and if you mark each ride as you do it as "done", you can re-optimize mid day and it'll re-optimize what is remaining based on current park conditions. However, I would copy the TP to another first b4 doing that so you always have the original to go back to.
@ejj That's exactly what I do too. Love the flexibility of being able to re-optimize ESP if something comes up: like a longer than planned lunch or whatever. Cool stuff.