(11-03-2021, 11:39 AM)mvossman Wrote: Honestly I think there is a limit to what you can get out of raw +/- even for an entire season. It is heavily affected by the quality of players you generally play with, and also the quality of players you play against (starter or bench). I find something like RPM more useful because it at least takes into account those variables, but it still needs a huge sample (potentially multiple years) to get a good picture.
There is of course a limit to what you can learn. I would argue you can learn bird's eye view, big picture, generalizations. It is a big picture stat, not a microscope and if you try to use it as a microscope it will lead you astray. But even in a sample size of seven games (though something like 20 games would be much better) you can get a generalization of what has gone on in that window.
Yes, the quality of the players around you and across from you of course affects the numbers, as does random luck. But on/off +/- is one of the most faithful stats at correlating with stars and superstars and it has the best correlation to winning (outscoring the opponent which also includes random luck). Over large sample sizes it much more so than other numbers will reveal the general, bird's eye view impact of a player on winning and team D and O.
The correlation of on/off +/- to winning is fool-proof, but the causation must be concluded soberly, carefully, and contingently. So for example, we can say without doubt that Luka's presence has NOT correlated to winning for the Mavs. Now how much of that is his fault? Impossible to say, but we can confidently say SOME of it is his fault, but the degree of his fault is impossible to nail down. But we can at least discuss it. And based on what we know about superstars, it is alarming that he is dead last on the team (by a HUGE margin) in on/off +/- (again, something like this never happened to Dirk in 21 seasons). Again, some of that is luck, some is his teammates, and opponents, but some of that is on him.