Tuesday, October 13th, 2009...4:03 pm
Ilardi on Durant
The adjusted plus/minus crowd is underwhelmed by the early returns on Kevin Durant. Wayne Winston, via TrueHoop, said that Kevin Durant’s APM performance was “in the lowest 10% of all NBA players.” Wayne Winston was not the first statistical expert to make this observation.
48 Minutes of Hell will feature a statistical expert of its own this season. Dr. Stephen Ilardi, who is an advanced stats expert, especially in the area of adjusted plus/minus, is set to contribute occasional articles to 48MoH. I couldn’t resist asking Dr. Ilardi what he thought of Monday’s Durant scuttlebutt.
His view of Durant, poor APM fully in view, is considerably rosier than one might expect. Ilardi acknowledges the force of Winston’s observations, but prefers to see them in a broader context:
- Steve writes, “Young, inexperienced players like Durant (who was only 20 last season) have long been viewed by insiders as posing a major defensive liability for their respective teams, and this explains why savvy coaches like Phil Jackson are loathe to give them major minutes. In this context, it should come as no great surprise to anyone that Durant hurt his team considerably on the defensive end last year.”
- “Traditional stats have no way of accounting for the magnitude of a player’s defensive impact (for good or ill), so it’s all too easy for fans and insiders alike to succumb to the “halo effect” on a player like Durant, wherein one highly visible attribute (scoring prowess) colors the perception of all other attributes, especially those like defensive play that are less readily quantifiable.”
- “Even on the offensive end, Durant has several widely acknowledged limitations: he’s not a good passer, he’s somewhat turnover prone, and he generally fails to make his teammates better. Thus, despite his high TS%, he doesn’t yet make a net positive contribution to his team’s offensive efficiency. In tandem with his defensive limitations, this rendered him one of the worst starters in the league last year with respect to his impact on the game’s bottom line (i.e., net efficiency).”
But Ilardi doesn’t stop there. One of his current behind-the-scenes projects is developing an adjusted plus/minus model that allows statisticians to project a player’s numbers forward. It’s a work in progress, but much of the heavy lifting is already accomplished. It’s exciting stuff, and you’ll learn more about it in the coming weeks. For now, Ilardi passes along this modest projection:
Durant will undoubtedly improve on both ends of the court in the years ahead. According to an analysis of APM trajectory over time, we could expect an overall APM improvement of a whopping +3.3 points/100 possessions over the next 6 years: and given Durant’s undeniable scoring talent, it is not at all unlikely that his improvement will be much greater . . . i.e., he could be contributing like a bona fide All Star within a few years. (Carmelo had a similar trajectory: major APM liability his first couple of seasons, yet All-Star caliber APM last year.) There’s no guarantee, of course, but it would be unwise to think such a scenario unlikely.
Using Carmelo Anthony as an example allows us to put handles on our expectations of Kevin Durant. Steve Ilardi paints the picture like this:
Dan Rosenbaum published a decent set of APM numbers for the 03-04 season (when Carmelo was 19), and he had Anthony at -8.1 for the year. He’s come a long way since! In my low-noise estimates for last season, for example, ‘Melo was still a pretty big defensive liability (-2.95), but his offensive contribution was so extraordinary (+7.73) that his net APM (+ 4.78) still rated as All-Star caliber (though nowhere near All-NBA level, as might be inferred merely on the basis of his offensive APM).
According to Ilardi’s 2008-9 low noise APM, Kevin Durant registered a -6.12 last season.
All of Dr. Ilardi’s previous APM estimates have appeared on 82games.com. His currently-under-construction low noise APM numbers are not yet public, although we provided our readers with a taste earlier this summer–his low noise APM is that scarce resource which gives a measurable account of Bruce Bowen’s defensive brilliance. You’ll be hearing more from Steve in the coming weeks.
17 Comments
October 13th, 2009 at 4:14 pm
Hopefully we can get some projections on our young guys too.
October 13th, 2009 at 4:16 pm
Matt,
All in due time, my friend.
October 13th, 2009 at 5:50 pm
*warm fuzzies*
fantastic addition the the blog, guys. welcome Dr. Ilardi and i look forward to future insight!
October 13th, 2009 at 6:21 pm
Nice additions to an already fantastic blog. Look forward to Ilardi statistical analysis and Bruno’s international perspective.
Any other wrinkles in the works?
Keep up the good work!
October 13th, 2009 at 8:58 pm
excellent post. Can’t wait for the season. Keep it coming fella’s. Would like to see what george hill’s numbers were last year and a projection for him mooving forward. I think he is going to be a very postive player this year.
October 14th, 2009 at 6:29 am
Great work here, guys. I’m excited for Dr. Ilardi’s periodic posts on my favorite website.
I’m intrigued by the fairly recent statistical revolution going on in basketball, although skeptical at times. I don’t always buy what Hollinger is selling. Correlating numbers to the team I love the most is going to be great.
October 14th, 2009 at 6:58 am
Bryan,
About Advanced Stats: no one offers them as a replacement for the things we see with our eyes; they’re offered to help us better understand those things, correct misconceptions, and add to the what we already know. I’m sure you know this, but we’re not pitching a big bag of numbers as gospel truth apart from other types of evaluation. In that regard, we’re not subtracting anything from what we already do. We’re just adding another layer.
October 14th, 2009 at 9:58 am
Completely understand, Tim. And that is what I’m excited about.
October 14th, 2009 at 10:37 am
[...] checks in with Ilardi on the whole Durant Debacle, while Pelton digs a little deeper. These are actually much better [...]
October 15th, 2009 at 7:50 am
Its really odd all these people tripping over themselves to prop up Durant. No one accussed him of being an axe murderer just being very ineffecient and having a negative impact on his team. And they were accussations with quite a bit of evidence. The guy is a major talent but enough already.
October 15th, 2009 at 9:03 am
Great post here. Really, he’s pointing out the obvious – he’s young, he’s an absolute monster in terms of scoring the ball, but he’s not fully developed yet. He’s using numbers to do so, though, which is nice. Just give the kid a few years.
October 15th, 2009 at 7:47 pm
I personally prefer the PER numbers when evaluating players. The +/- stuff seems a little off to me.
October 15th, 2009 at 10:13 pm
Jeff- PER is pretty much the worst “advanced” stat out there. Rewards shot-taking (think: Iverson, Maggette, etc.) and ignores defense. It’s an archaic and mostly useless stat.
October 16th, 2009 at 10:34 am
I don’t care what anyone says, KEVIN DURANT IS THE MAN!!!!
I think sometimes we take these plus/minus stats a little too far and this is the perfect example.
Is it possible that the other team slacks off when
Durant is not in the game since he is really the Thunder’s only established scorer and playmaker? I think so.
Durant is the future and I would pick him to start a team with over anyother player not named Lebron!!!
October 22nd, 2009 at 7:51 pm
Hello from Russia!
Can I quote a post in your blog with the link to you?
October 26th, 2009 at 2:11 pm
I have the same problem with WW’s version of APM as I do with most of the BCS inputs: They’re closed formulas (as in, not publicly accessible). I get the whole trade secret thing, but as a means of advancing the statistical understanding of basketball it’s a little bass-ackwards.
Not to mention there’s a kind of hidden assumption in computing any version of APM, and that’s that APM adds linearly. In other words, if you have a team with five APM = 0 players, you’ll tie the opponents, on average. Replace one of those with an APM = +5, and you’ll outscore the average opponent by a certain amount (typically either 1 or 5, depending on the precise formulation of the APM). Replace two of them with an APM = +5, and you’ll double the margin. And so on.
Except that it’s not clear that that’s actually so. Not only is it implicit in the use of APM, it’s also thickly involved in the computation of APM. And because it’s a natural expectation of the way APM works, it’s not so easy to notice. But once you notice it, you wonder if it’s really been firmly established. Got a blog post on it here:
http://thenullhypodermic.blogspot.com/2009/10/adjusted-plus-or-minus-more-or-less.html
PER is sort of the flip side of APM. APM is non-constructive in that it purports to tell you which players contribute most to winning, but can’t tell you why or how. PER is constructive; it’s an explicit linear combination of basic statistics. However, it doesn’t have APM’s appealing characteristic of short-circuiting the whole “which statistics are really all that important to winning” thing (the spectre of non-linearity notwithstanding). Got a blog post about that, too:
http://thenullhypodermic.blogspot.com/2009/06/superstars-and-per.html
Sorry if that seems like link-spamming. But the careful use of statistics is a big bugaboo with me and I think the points I make there are pretty relevant to the discussion.
November 2nd, 2009 at 10:47 am
Brian,
I appreciate your concerns about APM, and I’ve addressed the bulk of them – e.g., “trade secret,” non-linear player-by-player interaction effects, etc. – elsewhere: http://www.82games.com/ilardi2.htm. Stay tuned, as further clarifications will follow in upcoming 48MOH posts/articles.
- Steve Ilardi
Leave a Reply