I Have Grave Concerns About Marco Belinelli: Part 2

In Part 1 of this article I explained why the Sixers could use a guy like Marco Belinelli. But I concluded by suggesting that there is a major problem with the signing, which is that Belli does not appear to be a quality NBA player.

I’m going to focus most of the discussion on adjusted plus-minus statistics that calculate how the team did offensively and defensively when a player is on the floor, adjusted for whom he was playing with and against. These stats can be noisy, but they have the advantage that they account for anything a player does that helps or hurts his team, whether it be setting screens, spreading the floor, taking charges, picking up loose balls, failing to rotate defensively, as well as points, rebounds and the rest ... if it contributes to the scoreboard, positively or negatively, while you are playing, adjusted plus-minus stats like RAPM and ESPN’s RPM take account of it.

For those that don’t like adjusted plus-minus stats, let me just note that Belli’s traditional stats are not impressive; basically he looks like Jerryd Bayless’ little brother. For their careers they score about the same (small edge to Belli) but Bayless wins on per-minute assists and rebounds and steals and blocks. Belli shoots 38% from three, Bayless 37%. This year they are both right around their career averages by the way, so this is not some artifact of their long-ago play. Bayless has, I think, struggled for parts of this season with the wrist injury that kept him out last year, and despite his solid 3P%, he has in fact been bad, probably even worse than Belli. But they are similar-quality players; Belli has been better this year than the Bayless of the last couple months but not as good as the Bayless we saw in the early going. Which is already a warning sign because we all know that a guy who plays like Jerryd Bayless is not the solution to our problems.

OK, back to adjusted plus-minus. I’m going to use ESPN’s RPM, even though in a perfect world I’d prefer to use the version called RAPM. (RPM is annoying because ESPN is secretive about exactly how it is calculated, but the advantage it has is that it’s easy to access the data, whereas I don’t have a good source for RAPM.) I can confirm that based on past examinations of the data, the two measures are highly correlated.

Let’s look at the brutal numbers. Recall that this measures contribution in points per 100 possessions. For comparison, dominators like Joel Embiid and Giannis are around a +5, terrific players like Ben Simmons and Kyrie Irving are around +3, JJ Redick is around +.5, TJ near -.5, Justin Anderson -1.5, and TLC around -3. So replacing a near-zero player with a -3 has as much negative impact as replacing Ben Simmons or Kyrie Irving with an average player. Which is to say, a lot! And I mention -3 for a reason, here’s Marco, for the five years the stat has been kept:

ORPM measures offensive contribution, DRPM defensive, RPM is their sum.

The good news is, last year Belinelli played at the level of a poor backup, around -1.5; he was as good as say Justin Anderson. The bad news is, this year he’s played at the level of an awful backup like TLC. And the worse news is, it’s not this year that’s the fluke. He was awful in 2015-16. And awful in 2014-15. And also terrible in his big title season in San Antonio, 2013-14.

As I mentioned, this kind of statistic can be noisy, even over a full season. But when a guy is averaging a -3 RPM over five years, indeed has only one year among the five meaningfully above -3, we can’t explain that away with noise.

Now, most of the negativity here is coming from the defensive end, and many people like to point out that defensive statistics in basketball are not all we would wish. But understand that when people talk about defensive stats being lousy, what they usually mean is that we tend to count things like steals and blocks, and such counts leave out both great one-on-one D and great team D (switching, helping, etc.). All that is quite true, and is a reason we shouldn’t overweight the importance of steals and blocks, which can mislead. We’ve all seen players abandon their post to chase a block, leading to a bad outcome not held against them by traditional statistics.

And what most people call “advanced” defensive statistics don’t help! Measures like DBPM (Defensive Box Plus/Minus) generally just add up counting stats like steals, blocks, and defensive rebounds using some weighting scheme. They are useful and convenient, but could easily give poor numbers to a player who is playing quality team D. Could Marco be an example of such an error? The answer is yes: although his DBPM this year is a hideous -2.0, and although his career DBPM and the figure for all recent years is also massively negative, it is possible, not likely but possible, that his defense is merely bad and not exceptionally terrible as the counting stats suggest.

But understand this: defensive adjusted +/- stats like DRPM do not have this problem. As noted above, the measure simply counts how the team does in preventing baskets when you play, and adjusts to take account of the quality of your teammates and opponents. Such stats have random noise in them of course, but there’s really no fundamental flaw in the approach as there is with counting blocks and steals. If you are a great defender, you could give up lots of points to the player you guard, never get a steal or block, and you would nevertheless have, in general, a terrific DRPM because your great D would mean the team did well defensively when you were on the floor. Here are the top players in the NBA in DRPM, among those who play at least 20 minutes a game:

Hassan Whiteside

Rudy Gobert

Andre Roberson

Joel Embiid

Jusuf Nurcic

Luc Mbah a Moute

Clint Capela

Kyle Anderson

Robert Covington

The next group includes Draymond, Giannis, AD, Drummond, Ben Simmons, and Jimmy Butler. It’s pretty obvious this syncs up nicely with the views of fans and experts alike as to who is really good at defense. Not perfectly of course; the method suggests that Avery Bradley is hugely overrated as a defender. But it’s clear to me that, if we have enough data to wash out random variation, this is a reliable method for seeing who is good and bad at defense, as, indeed, how could it not be? People sometimes like to knock these methods by cherry-picking examples. Famously, a couple months into last season Kawhi Leonard had an unimpressive DRPM because of ten or twenty lucky threes that got hit while he was on the floor. But by the end of even a single season, that noise had mostly washed out and Kawhi showed as having an excellent DRPM for last year. And over multiple years, the odds of such an occurrence are minuscule.

That’s good news for us as people who want to know the truth. And it’s fantastic news for us as fans of Joel, Cov and Ben. But it is not good news for Marco Belinelli. The evidence that he is an awful defender is overwhelming. And this metric puts a number on what “awful” means: Marco Belinelli hurts your defense about as much as Joel Embiid helps it. Um, yikes! Defensively, having Joel and Marco is about as valuable as having Richaun and TJ at the same two positions; the increase from Marco to TJ helps you defensively almost as much as the Joel-Richaun drop hurts. Think about that for a moment.

Now, can offense make up for a bad defender? Sure it can; take a look at Lou Williams, who is playing Marco-level D this year but who has been so terrific as a scorer and playmaker as to more than make up for it. Or look at our own JJ Redick; JJ is around a -2 defensively but is over +2 on offense, so on net is a solid starter.

But that’s just not the case for Marco. He is playing offense at a level that is above “average player” but below “average starter.” His defensive negative is many times larger than his offensive positive. And with him turning 32 in a few weeks there’s no reason to expect any more than what we’ve seen.

I know the notion that Marco is just adequate offensively will surprise many, but I’m not sure why. He scores OK but not a ton; around 15 points per 36 for his career, 17 this year. He shoots 37 or 38% from three which is quite good, but that’s about the extent of his offense. His assists per 36 are about the same as Amir Johnson’s; is Amir your idea of a guy who’s terrific at creating for others? He gets hardly any rebounds. His free throw percentage is terrific, but he doesn’t get to the line that often so the impact is muted. Again, he’s a solid offensive player, at least this year and last. But he’s hardly something special. And when you combine “adequate” offensively with “horrifying” defensively, well, it’s not an effective combination.

Now, with this all said, is it possible our bench scoring situation is so dire that he helps us anyway? Actually, yes! It’s possible. Basically the problem we have is that Bayless has been unplayable (-4 RPM). And at the moment, Justin Anderson is injured. So that leaves us with only TLC to play wing off the bench, and TLC’s RPM is only a little better than Belli’s.

So, look, our options are so poor there is a case that Marco is actually the least bad of them. You have TLC, who’s about as bad as Marco. You have Anderson, whose numbers are quite a bit better (-1.4 RPM), but who, at 35% from three, may be shooting better than his true level of ability, and who moreover has been injured lately. You have Bayless, who has an even lower RPM than Belli and is probably hurt. And Markelle: oh, Markelle, please return to us!

So, there’s a decent argument for playing Marco, but not so far an argument that doing so will improve us more than a tiny bit. Yet there is enormous enthusiasm for his addition, because folks feel Belli fills a crucial need for bench scoring and creation. In Part 3 we will examine this theory.