[cloudwalk9]

Your EDIT. The first thing it suggested is actually very similar to ensembles in meteorology. I actually find myself doing that often if it's something extremely important. Just feels natural to cross-check with other models or with reality. The disclaimer says it may make mistakes after all...

Like you don't predict the weather or a hurricane track with a single model. The NHC uses many.

It's still probablistic, but if multiple models are independently in agreement, then it's at least worth investigating further.