It's only a prediction made from historical data. That data was wait times at three big time rides at MK, and from there TP is trying to predict crowd levels at all four parks. This can come close, but is by no means perfect.
What is interesting is that the data (sampled from the three rides at MK) still shows MK as a best park, and AK as a park to avoid.
In my opinion, the only way for the prediction to be correct is if there are long wait times for any of those three rides on that day, and that is not at all impossible. Space Mountain will have just reopened from rehab. It's bound to draw a crowd, and as I recall, it was always a long wait before.
We've got that MK planned for that day, and see no reason to change it.
While I respect the opinion of those who endorse TP from positive past experiences, there are a few things to be understood: 1) TP is not going to be 100% accurate. Nothing predictive ever is. 2) If you use any crowd estimating tool, you're likely to perceive a crowd level you expect to see. 3) Just because the crowd levels are at a certain point where you are (or where TP sampled data) doesn't mean the crowd levels are at that point park wide all day.
It's a planning tool. When it jumps this much in the middle of the week on a non-EMH day before school is out for Christmas break, you have to take it with a grain of salt.