Well, in order to do that, you will have to look at the players who actually won. It's not ''playing catch up.'' It's just how you do the thing. About league environments, you do have to assume some stationarity, but there's a pretty strong argument that league conditions are relatively stationary, with a few abrupt changes every decade or so, and it's based entirely on looking at cup winners: cup winners tend to repeat.
Where the slight of hand comes in is the second point: it's not at all about evaluating which players contribute to your winning (beyond ''investing in analytics''), or what team type of team composition would best accomplish the goal of winning - which was the point that old mate was trying to make originally - but how to optimize team operations. But in order to do that, you need to have a goal, and those goals are what players do we need to fill what roles. In so far as your second point concerns ''proper evaluation of what different players contribute to winning'' any analytics team who knew what they were doing would be doing the exact same process that I described, because that's how analytics work.
It seems like you think that inference and prediction are disjoint tasks. Like, we should be proactive, and attempt to predict how to win. But to predict what will be successful in the future, we need to infer with what sensitivity different players and roles affected a team's success. The prediction will be based on this inference. There's never been a prediction algorithm that didn't do basically this.
P.S. I don't know how you can suggest that we can't tank effectively with the new rules, but we can look at how teams like Chicago used back diving contracts to a competitive advantage with a straight face.
