It's the same thing when interpreting polling data from multiple sources. Each has their own methodology. You have to be well versed on how each model is created and how much weight each contributes to the most accurate path. It's not as simple as 9 out of 10 models agree type thing.