How to Determine What Is Predictive and What Is Not?

I do a bunch of machine learning so it'll feature select neural networks perceptron algorithms k means clustering that will probably go over people's heads unless they come from a data science background. There's scientific mathematical ways of finding features that are strongly correlated and then having watched sports for over a decade um you can see if those are things that are just by happenstancer if there's actually reason there would be causation there. Yeah I mean there are a ton of players I disagree with the market with a lot in basketball a few in football one good example is Kobe Bryant who many thought was more valuable without him than he really was. So i like to find players who are overrated or underrated