lessthanjake wrote:Okay, so a few foundational problems with what you’re trying to argue here:
1. These measures are either RAPM or based on RAPM with a box prior. RAPM is quite noisy over single-season samples. Adding a prior makes it less noisy in small samples, but it is still inherently noisy, primarily because the sample of data when people are on or off the court isn’t high in small samples. So yes, with noisy data, you get some weird results (though, the results also do *mostly* look right even in such small samples, and I don’t see you denying that). Some of those results you think are weird might actually be because you are wrong in your individual assessment of players, but a lot of the time you’re just going to be pointing to the result of RAPM being noisy in small samples. The problem you have here is that RAPM gets a lot less noisy over larger samples, and I didn’t just give you a single season of Jokic’s and LeBron’s data. I gave you analysis across a bunch of seasons in a bunch of different metrics. So yes, one can always point to weird results in single-season impact data, caused by statistical noise. But the chances of the results just being caused by statistical noise when we are looking at data across a bunch of seasons (and a bunch of different metrics) is way lower, and pointing to idiosyncratic single-season results doesn’t change that.
2. You do try to claim that career DRAPM must be wrong, based on things you assert are “abnormalities” but you seem to not understand what the time horizon is on that career DRAPM data. Specifically, it starts at the advent of play-by-play data in 1996-1997. So yeah, Hakeem only looking slightly better than Jokic in defense is really not particularly abnormal at all, because the data is only encompassing Hakeem’s age 34-39 seasons—which are miles away from his defensive peak. Same thing with Jordan, whose seasons in that data set are just him at ages 33, 34, 38, and 39. The data on Gary Payton doesn’t include his DPOY year, and almost half of it is from years where he was old and not making all-defensive teams anymore. The career RAPM data isn’t looking at the whole career for older guys like this, so you shouldn’t draw opinions about the data based on a feeling of how those players should rate based on their whole career.
I mentioned a lot more players than those legacy guys. And I could have mentioned a lot more. Hartenstein being more impactful than Anonoby for example.
3. Meanwhile, you mention Dwight Howard—who does have complete data for his career included. But the thing is that we also know his year-to-year DRAPM (as calculated by the same guy who did the career RAPM calculations I linked to) was very high—typically in the 3.0-4.5 range—during the years he was being recognized as a great defender. The rest of his career—which includes a lot of years where he was nowhere near as good of a defender as before—obviously weigh him down, such that the overall career DRAPM isn’t much higher than Jokic. But DRAPM actually does indicate that peak Howard was a significantly better defender than Jokic, which actually is suggestive of DRAPM being accurate to your expectations, not inaccurate.
You said these metrics get less noisey over the course of a players career. Howard was a better defender in his twighlight years than Jokic is now. Which one is it?
So yeah, there’s really not a lot to your criticisms, once we recognize that single-season impact data is very noisy (and that I’d provided less noisy multi-year data validating my point), that career DRAPM data only starts at 1996-1997, that you partially relied on completely false information, etc.
Again, I mentioned players who played after 97. What’s more, he ranks above Kawhi, Porzingis, Chris Bosh, Jason Kidd (which I mentioned), Jarett Allen, Andre Kirlenko, Dwayne Wade, Kobe Bryant, OG Anonoby, Mitchel Robinson, Marcus Smart, Serge Ibaka, Brook Lopez, Iman Shumpert, Billups, Josh Hart, Shawn Marion, Klay Thompson, Ilgaskaus, KCP, Noah, Tayshaun Prince, Nic Claxton, D’eandre Jordan, Jaren Jackson, Trevor Ariza, Ben Simmons and is a few points below Giannis, Patrick Beverly, Anthony Davis, Shane Battier, Dwight Howard (Marc Gasol is rated higher than Dwight Howard), Tyson Chandler, PJ Tucker, Marcus Camby, thybulle. These are all players praised for the defense being rated worse or near to a player who is hunted on pick and rolls.
Jerome James managed to beat Jokic though and Isaiah Thomas is a better defender than Deshawn Stevenson. This ain’t just a few anomalies. This metric is ****. And this just one of these advanced stats.
2. How much does his rebounding and deflections contribute to his defensive impact data looking good? Probably quite a lot! As it naturally would, since those are very important things that genuinely have a massive effect on how much opposing teams score. Like, sure, if you want to define everything Jokic does really well defensively as not being part of defense, then you can get yourself to a conclusion that he’s not a good defender. But that’s just obviously ridiculous. Jokic is an *incredibly* impactful defensive rebounder, and that’s a huge part of why he’s a good defender. And you seem to not be understanding that if defensive rebounding is helping Jokic in this data, it’s because his rebounding leads to opponents scoring less, not because the data is arbitrarily weighing defensive rebounds highly. Impact data is not grounded in weighing box stats. It is grounded in isolating out an individual player’s effect on plus-minus data, regardless of any particular box data stats. That’s what makes it so good for measuring defense—which is an area where box stats don’t encompass much!
I don’t know why you’re saying I’m not understanding how much his rebounding impacts his defense when it’s one of the reasons I put forth as him ranking so favorably in advanced stats.
3. As for how much his offensive impact drives his defensive impact, there’s an argument that that might be the case to some degree. I’m not sure it’s really true, because, while being better offensively limits the transition opportunities the other team has, being better offensively *also* makes your team more likely to be ahead by a lot instead of behind by a lot, and teams defend a lot worse when they’re ahead by a lot. Either way, though, it’s not really relevant to this comparison, because we are comparing to LeBron, and LeBron has elite offensive impact too, so this same factor would be helping LeBron a similar amount. It cannot be some major factor allowing Jokic to look similarly good to LeBron defensively while actually not being as good—rather, it’d just be a potential factor that would make them *both* look better defensively than they should.
Jokic is more efficient a better FT shooter and is averaging more assist in his prime. All while being a much better rebounder. These aren’t marginal difference, and maybe these are things that explain why a player who’s hunted on pick and rolls has a nearly identical drapm than someone who almost won DPOY. You don’t hunt good defenders on pick and rolls. You don’t seek out Draymond Green, Joakim Noah or Keving Garnet on PNR
Please go on and tell me what measure of defense is better than impact measures? They’re just clearly the best, since everything else is hopelessly simplistic in measuring defense. Again, please understand that “advanced” box stats (like PER, win shares, etc.) are not the same as impact data. Impact data is not weighing assists or rebounds or other box stats to spit out a number. It is isolating out the effect of an individual player on what happens to the scoreline. That’s obviously the best way to look at defense, and it is a method that completely backs what I have been asserting and contradicts your position entirely.
You can actual watch the game. Again, Jokic himself admits to not being a good defender. Is it a good metric to determine someone’s defensive impact, maybe, but there are clearly abnormalities that don’t track with the actual game and Jokic ranking so high appears to be one of those abnormalities.
And, since this part of your post was not particularly clear, to the extent you’re referring to the 2006 series against the Pistons, please note that the Cavs offense was bad in that series too. It was the same story. In that 2006 series, the Cavs scored 99.4 points per 100 possessions against a team that had given up 103.1 points per 100 possessions in the regular season. In other words, the Cavs had a genuinely bad offense in that series. So how was the series close? Well, the Cavs gave up 106.4 points per 100 possessions to a team that scored 110.8 points per 100 possessions in the regular season. So, in that series, the Cavs had a bad offense and a good defense, and the result was that the series was close. And, by the way, to the extent you might point out that the Cavs got to 7 games despite actually being outscored a good bit and that that suggests doing well in business end of close wins, please note that LeBron had a putrid 45.6% TS% in the 4th quarters of the three games Cleveland won (along with 5 turnovers). The Cavs won those close games by giving up only 52 total points combined in the 4th quarters of the three games they won (17.3 points per quarter) and overall holding the Pistons to incredibly low 98.5, 86.1, and 99.8 points per 100 possessions in those games, while the Cavs offense averaged like 100.3 points per 100 possessions—again, bad offense carried by great team defense.
I was referring to the entire season where they had an identical record to the one they had in 07. The defense wasn’t the driving force to them winning 50 games. So what was?