Then for the predictibility why would it compare all the models instead of choosing the models that mirror each other better for a better percentage. They all will just show the same percentage.

You are referring to the spread of forecasts among the Ensemble members of a given model.
While in windy it is the predictability calculated by Meteoblue within its Multimodel system.
For instance, predictability is higher at the center of an anticyclone than within a low-pressure area for wind and rain forecasts or for thunderstorm predictions generally. This applies for each model.