Dutchball97 wrote:f4p wrote:which stats would make the case for Tatum? in addition to all the stuff posted above, we're talking 31.7 vs 26.2 PER, 0.273 vs 0.217 WS48, and 11.1 vs 6.7 BPM. those are just massive differences. and i'm not sure how much the defensive difference could mean if the celtics are 17th in defense and the mavs are 8th. and this is with tatum's FTr and TS% way above his career numbers, so we'll have to see if it is sustainable.
and as also mentioned, they just aren't the same in the playoffs. luka has thrown down 3 massive playoffs already and, for all the talk of how well tatum played last year, a 17.6 PER / 4.5 BPM playoff run is not blowing anyone's skirt up.
Like LukaTheGOAT showed, more advanced metrics show a very similar picture as the boxscore metrics(as it honestly does quite often) but Tatum does have a pretty clear and big advantage in terms of raw +-. Tatum is currently 5th in the league at +129, while Luka is 61st at +51.
Some people shun the boxscore metrics and even the boxscore/+- composite metrics like EPM, RAPTOR and LEBRON, while pretty much regarding +- (whether raw +- or RAPM) as the "true impact" of a player so I guess those people could make a statistical case for Tatum at the moment?
In the context of 1 year composite metrics are gonna beat out raw RAPM. The fact that they’re more predictive than actual RAPM hints to me that they’re ability to stabilize noise works. Not RAPTOR tho.
If you view RAPM as impact = player impact + noise, the composite metrics are gonna be player impact + noise + the composite part, but noise is gonna be mitigated. Composites generally are more predictive which would imply it’s a better measurement of player goodness
The argument against it I’d think is that using different things seperately in their own context has more value, I kind of see that but at the same time, in a general sense they’re gonna be more accurate outside of outliers for players that don’t fit patterns
That is probably the bigger issue with composite metrics, as far as I know I think composite metrics are usually box score priors + RAPM + any other adjustment here or there
There’s nothing wrong with box score priors neccessarily but maybe you can argue assigning a linear value generalized to each player isn’t gonna be max accurate because of different roles
I think Rapm doesnt meaningfully look at combinations fit which is kind of a big deal to me if it’s true though, and the justification is on a large scale it doesn’t mean much, that’s not something that you’re gonna be able to see that’s way because it’s very much on an individual basis.
At the same time again I think luka easily lol, bro nice. Tatums been great but I think I’d wanna see him in the playoffs