I have never used their numbers so I cannot comment but I really wonder what their alogrithim is to predict crowd levels? I would imagine if I were to build something of the sort I would try to factor in the data I could actually get. Things like onsite room availability, the presence of events, holidays, historical seasonal attendance, and as the date approached availability of FP+ in the parks and even Dining reservations. I guess you could even cooralate airfare costs to Orlando to see if there are cheaper than normal incentive rates. Given the factors I would assume that they probably don't start getting good until say the 60 day out window and at that point everyone has already made their plans so at that point the information is antedotal at best.