Yep, the concept I am trying to get at is really difficult (for me anyway) to define. "We scored 6 runs off their bullpen - they must have all been AAAA callups." "We beat Skubal - we beat Crochet - they must have not been on their game that day." These could be legitimate explanations - or just typical BS after the fact.
Maybe some important factors are just unknowable. Or maybe Big Data techniques have ways of teasing out the meaning from scads of conflicting data. I tend to believe that if an effect is important, then evidence for it can be found if you look at the evidence correctly. (You can also find evidence for your supposition by looking incorrectly.)
It's a bit like, we all believe a .300 hitter is better than a .260 hitter, but when you come down to individual games, anything can happen (and often does). Maybe that .260 hitter has just as many multi-hit games as the .300 guy, but he suffers more o-fers while the .300 guy plugs along with more 1-for-4 days. That kind of variance is completely missed in season-ending averages, but might affect how you value the two players.
I spent my career in a small corner of analytics, but I never had the technical chops to tackle a statistically-oriented question involving variance for sets of data (like in baseball) that are a bit unruly to start with. Uniformity is lacking - some days you get 3 plate appearances in a game, some days you get 5, and how do you do a "variance" study of how many hitless days you have, in that environment?
A bit of spot-checking along these lines a few days ago didn't turn up the pattern I was hoping to be able to demonstrate to others in support of my view, so I didn't post about it. 😀