What can NBA advanced stats tell us about this year's MVP race?
In an unlikely twist, the utilization of all-in-one player value metrics has become a flashpoint in this year's debate between the Milwaukee Bucks' Giannis Antetokounmpo, the Philadelphia 76ers' Joel Embiid and the Denver Nuggets' Nikola Jokic.
Among the many bad-faith arguments made in this regard, one stands out as legitimate: Those utilizing advanced stats don't necessarily understand how they work.
With that in mind, let's take a refresher course on how what ESPN's Zach Lowe refers to as "the VORPs and SHNORPs" (inspiring Andrew Patton to create a semi-real SHNORP stat blending some of the others) work and what they indicate about the cases of the top three MVP favorites.
Throughout the NBA season, I answer your questions about the latest, most interesting topics in basketball. You can tweet me directly at @kpelton, tweet your questions using the hashtag #peltonmailbag or email them to peltonmailbag@gmail.com.
In addition to the main question, this week's mailbag tackles whether the international flavor of this year's MVP race should raise concern about American player development.
"Which analytics would you use in picking MVP and who would they choose?"
-- James
Certainly, the sea of acronyms and capital letters (to which I contribute) can be intimidating to navigate. The first thing to understand is there are, broadly, three ways to measure player value:
Using exclusively box score stats: Player Efficiency Rating (PER); win shares; box plus-minus; my own wins above replacement player (WARP) metric
Adjusted plus-minus systems that don't incorporate any box score data but only how a player's team performs with him on or off the court, adjusted for teammates and opponents (regularized adjusted plus-minus, the state-of-the-art version of this, can be found at NBAshotcharts.com)
Systems that synthesize both box score and plus-minus data (ESPN's own Real Plus-Minus and FiveThirtyEight's RAPTOR fall into this category, as do estimated plus-minus (EPM) and the LEBRON metric)
When considering MVP, we're no longer interested in predictive metrics like DARKO projections or DRIP. They utilize multiple seasons of data to be most predictive going forward, so they can't isolate a player's value within a single season as we want to do with MVP.
Because they were the first generation of player-value metrics, the ones using only box score stats tend to be most familiar and widely available. All of Basketball-Reference.com's metrics, for example, fall into this category. But they've struggled to keep up as it's become easier for big men to generate strong box score stats. Of the top 20 qualifying players in both PER and win shares, 13 are big men.
Box plus-minus does better with big men vs. perimeter players by utilizing positional adjustments for individual stats, but this method can break down at historical extremes like Jokic's unprecedented playmaking for a center. Although that's not a reason to discount the metric overall, when it comes to MVP those extremes are most important.
As a result, our focus in evaluating player value should be on the metrics that utilize both box score data and plus-minus data because of the complementary perspectives the two provide. Not everything a player does to help his team is captured in the box score, while plus-minus data tends to be extremely noisy from season to season, and box score stats help distinguish real value from randomness. The combo metrics have proven most predictive in out-of-sample testing.
Let's break down these numbers for the top candidates.
Offensive ratings
I think sometimes in the MVP conversation we tend to flatten differences among elite players. All three top MVP candidates have been outstanding offensively, but there's a general statistical consensus that Jokic has been the best of the three and Embiid the weakest despite leading the NBA in scoring at 30.4 points per game.
The first reason for that is although Embiid has the highest individual usage rate of this group at 37.6% (and in the league, just ahead of Luka Doncic), the advantage Antetokounmpo and particularly Jokic have in assist rate means they're responsible for creating about the same amount of their teams' offenses. Secondly, they've been a bit more efficient. Jokic's .661 true shooting percentage is first among high-volume scorers, while Antetokounmpo (.633) is ahead of Embiid (.612).
Defensive ratings
Here's where I acknowledge I don't buy the massive defensive impact both RAPTOR and RPM attribute to Jokic this season. From a box score standpoint, the only difference in Jokic's game is a massive increase in defensive rebound rate (a career-high 31.3% of available defensive rebounds). Yet RAPTOR has his box score rating component going from a modest plus-0.7 to the league's second-best mark this season (plus-5.9).
I think there's a strong case Jokic is an underrated defender. His on/off page at Cleaning the Glass shows the Nuggets have been substantially better defensively with him on the court in every season of his career but last year. (That he's gone from a massively negative team impact to a massively positive one this season is a reminder how noisy those stats can be.)
Still, I don't believe that he's the best defender of this group, and certainly not by a wide margin. I think LEBRON's ratings hew closest to my assessment -- a narrow gap between the three players on defense, much smaller than on offense.
Value over replacement
Looking at per-possession ratings is useful to understand how metrics are evaluating players. When it comes to MVP, however, we want to turn to the value version that factors in minutes played and typically an estimate of replacement level to see how much a player has helped his team while on the court. This is particularly important for the cream of the crop because they're so much better than the alternative options behind them on the bench.
Here we probably need to add a fourth player to our group.
Embiid and Jokic are in the top three of all four value metrics, but Giannis isn't higher than fourth in any of them. All of them have the Boston Celtics' Jayson Tatum either second or third, due in large part to Tatum having played 2,705 minutes, far more than any of the top three candidates (Jokic is at 2,476, Embiid 2,259 and Antetokounmpo 2,204).
Even with that advantage, box only stats tend to put Tatum behind not only this group but also Luka Doncic. Tatum is fifth in my WARP metric and sixth in value over replacement player, the value version of BPM. I think it's possible that on/off metrics give Tatum a tad too much credit for the strength of the Celtics' team defense. So he'll still be fourth on my ballot.
As for the top three, Jokic is a clear No. 1 to me based on his advantage in both quality and quantity of play. You'd have to give Embiid a massive defensive edge to get him ahead of Jokic in terms of total value. Embiid slots in a strong second with Giannis in third.
The last spot on my ballot comes down to Doncic vs. Stephen Curry. Although the box only metrics value Doncic more, the combo ones favor Curry because of how the Golden State Warriors have struggled without him. Despite his recent injury, Curry has played basically the same number of minutes as Doncic.
I think this is the rare case where the idea of a player being downgraded by comparison to himself is actually true. There's been so much focus on Curry shooting the worst percentages of his career that it's overshadowed him still being a more efficient scorer (.601 TS%) than Doncic (.570) is at higher volume. So Curry grabs the last spot.
"As we look at the top three MVP candidates and a quickly rising Doncic, you could argue that today's top four NBA players all developed their game outside the United States and our programs. Even the most electric young American (is there any question? Ja Morant!), grew up a little outside our star development path with two years at Murray State.
Is this an anomaly or do we need to review how we are developing young basketball stars in this country?"
-- mb
I don't think so. Looking at the top three, there's not exactly a common development path to favor. Jokic, like Doncic, developed through the European youth system. But Giannis was on the fringes of that in Greece and Embiid didn't get serious about basketball until coming to the U.S. at age 16 -- hardly the kind of development path you'd expect others to emulate.
To some degree, I think we're dealing with a math problem. Besides Doncic, the other three players are all 6-foot-11 or taller and the two latecomers to basketball are incredibly athletic for his size. (Jokic has his own degree of skills and unparalleled skill for his own size, but it's tougher to say whether that was innate or a product of his development.)
Although Morant wasn't a highly touted recruit, I wouldn't say he was outside the American development path. He famously played AAU basketball with top recruit Zion Williamson in South Carolina. Morant's development might have been different at a major college program where he wouldn't have had the freedom to do as much with the ball as at Murray State, but that doesn't seem like any reason to rethink things.
Of course, the way NBA players develop in the U.S. continues to evolve all the time with Overtime Elite, the G League Ignite and international options available to players who want to bypass the traditional college path. We'll want to continue monitoring how these alternatives are working out to see what we can take from each of them.