Baller2014 wrote:I'll say this with regard to KG. If you want to convince voters like myself that he's coming up soon, then the pitch needs to be different (for me anyway). Right now the pitch is overwhelmingly "advanced stats, advanced stats, advanced stats!" What about if you're not a big fan of advanced stats? What if they're wrong, as everyone agrees they sometimes are? You need an argument that can be made independently of citing advanced stats (not one that rips on his coach either). Advanced stats are a "tool in the toolbox", but when the impact they claim someone is having doesn't seem to show up in the team results there needs to be more analysis. I made several posts on this in the last few threads, focusing particularly on KG's help in 2002. Nobody made any attempt to seriously reply to them.
I've made arguments outside of advanced stats. If you don't want to believe them, well, I'm not sure you want to be convinced. There are so many role players with uncertain values that it's really hard to pin down what was wrong with those Timberwolves. But even with some pretty basic information you get the sense that he was a monster player.
Someone else made a great case based on an evolution of his defense and how he was used.
ElGee made a post about how they were a 9 SRS team with Garnett on those Celtics. Is that too advanced? He could have cited wins/losses. I think it's pretty simple to say when Garnett played, the Celtics were X good. When he didn't play, they were only Y. That's something even casual fans could understand.
shutupandjam wrote:Yes, to some degree but it's designed to try and isolate those factors. It can be difficult to do that in some circumstances though like if one player plays a huge portion of his minutes with another player. And just because a player has a good rapm doesn't mean he necessarily would in a different situation. For instance, Matt Bonner might have a good rapm in his role as a stretch 5 off the bench who takes open shots created by his teammates, but if the Spurs asked him to be the primary offensive threat, you'd expect his rapm to drop.
This is a good rabbit hole.
If you asked Shaq to be your point guard and he had to run the offense and didn't post up, his RAPM would plummet. But that doesn't make him a bad player, per se, just misused (Diaw is an example of a guy who was misused early in his career.)
However, value is value. Bonner sometimes has sky-high RAPM marks, but it is meaningful. Kidd didn't shoot often, but he was still very valuable. If we didn't have assists in the box score, he'd be underrated.
One caveat is the situational player who only plays when he's needed. But this is rare in basketball. It's not like baseball. Jason Collins is probably the best example. Near the end of his career, he was starting to be used only for defending big centers. He didn't start every game in Atlanta, for instance. This masked his deficiencies.
But at this point, we're all looking at guys who play heavy minutes and don't cherrypick useful situations. Garnett changed teams and teammates several times and he was consistently rated high.
I've been working on a project I've always wanted to do. When people post stats of a guy versus good defenses, it's not anything ... say, robust. It's just an average versus defenses better than X and an average versus defenses worse than X. But that's not a stable comparison because some guys are lucky and the best defenses they've faced are all rated like -5 to -6. Or maybe the average defense the player has faced is -10 or so.
So I grabbed gamelogs of Olajuwon and Shaq (and Ewing) to see what the marginal change is when an opposing defense, or team strength rating (SRS), improves by a point. It's just simple regression right now.
I guess I'm calling these ...
V-ratings? For variable player stat ratings....
(edit: fixed per possession stats)
Hakeem (1987 to 1997, playoffs included):
PTS/100 coef 0.217
PTS/100 -5 def -1.084
Ortg coef 1.008
Ortg -5 def -5.040
Usage coef -0.102
Usage -5 def 0.512
Shaq (1995 to 2005, playoffs included):
PTS/100 coef 0.392
PTS/100 -5 def -1.960
Ortg coef 0.713
Ortg -5 def -3.567
Usage coef -0.138
Usage -5 def 0.690
(The 100 stands for 100 possessions. It's so minutes or pace are not factors. The coefficient is a regression result should be on average 1, except for usage, because you'd expect a one-to-one relationship with defensive efficiency and, say, points or efficiency. The -5 def rows are for the change you'd see in points, usage, or offensive rating from facing an average team or a great defense rated 5 points below the league average. Positive is good. It means you're, say, adding points versus a great defense.)
One observation and this includes Ewing is that these big men keep up their points production because their usage slightly increases against better defenses. They're all almost the same. However, efficiency is the difference. The coefficient for efficiency should be on average 1.0. Instead for Shaq it's under 1, meaning that his efficiency holds up against good defenses better than average. That's not the same for Olajuwon. His efficiency drops as much as you'd expect. And remember this is for ten years of his career. Shaq's points per possession, however, actually increase when the defense gets better. He's taking on a larger share of his team's offense. Hakeem's slightly decrease.
But the story about Hakeem versus elite defenses is mainly about his prime. So let's be kind and try 1993 to 1995 and compare it to Shaq's best three years.
Hakeem (1993 to 1995, playoffs included):
PTS/100 coef 0.214
PTS/100 -5 def -1.069
Ortg coef 0.498
Ortg -5 def -2.489
Usage coef -0.171
Usage -5 def 0.857
Shaq (2000 to 2002, playoffs included):
PTS/100 coef 0.319
PTS/100 -5 def -1.596
Ortg coef 0.970
Ortg -5 def -4.852
Usage coef -0.099
Usage -5 def 0.497
Hakeem scores more against elite defenses because he shoots more often and his efficiency increases less than expected given the opposing defense. His Ortg coefficient was chopped in half. Shaq actually looks a little worse during his peak. On an interesting note, Shaq's coefficient for '95 through '98 is 0.67, meaning his efficiency held up better when he was a young player. Are we underrating him there?
In summary, Olajuwon was not a magical performer against elite defenses for most of his career. He does quite well during his three year peak, although it's nothing too earth-shattering and it's only three years. Plus Shaq, for instance, starts from a higher base as he shoots more often and is more efficient.
edit:
This is fascinating:
http://shutupandjam.net/nba-ncaa-stats/ ... ed-impact/One critique: when I made my own all-time ranking based on stats alone, I found the David Robinson/Karl Malone problem. They were ranked really high, higher than in reality. But it's because we often just use regular season stats even for things like estimating championship odds. Any way of using playoff stats, or an approximation thereof, for this odds method? Like estimate a player's average drop in impact in the playoffs and apply it to every season. By the way, Kareem led the league in that estimated impact ten straight seasons. I know the 70's were an odd decade, but jeez....
