Painting MVPs by Numbers: Comparing Six MVP Candidates

You've heard MVP debates dominating the radio waves for months, and you've seen the names Russell Westbrook, James Harden and Stephen Curry especially, splattered and slathered all over the blogosphere like the handiwork of a bad graffiti artist during his "cry for attention" phase.

That's just the standard operating procedure for the NBA in April — especially when the league features so many bonafide MVP-caliber performances. But April 2015 will be more contentious than most, as many fans and experts seem legitimately torn. Few MVP races feature more than two viable candidates, but you could make a case for as many as six NBA players right now. SIX!

Using platitudes, anecdotes and feel-good-isms is a sure-fire way to chase our own collective tails, so let's refrain, shall we? For this article, let's take a look at the MVP race and determine a winner without using phrases like, "makes his teammates better," unless it can be objectively backed up by stats. We'll similarly dump tired truisms like "he's a winner," and "he's a good face for the league," and "he carries a bigger load." Again, we're trying to quantify things in order to reach a defensible conclusion. Save those other ideas for your creative writing workshop.

All that said, you should (will) know that I'm a man of science: I crave understanding and proof. I will proceed with the assumption that we are of like mind.

A quick note before we dive in: we're disregarding monetary value, obviously, because otherwise we get a wacky result. Obviously Anthony Davis for $7 million this season is a better value than James Harden at just under $15 million. I doubt even Rockets fans will argue that. And while we're at it, let's disregard everything else that isn't purely related to playing the game of basketball. Stephen Curry has hung out with Barack Obama and Jamie Foxx this year, so while his Klout score is at an all-time high, I don't think it's fair to include that in a discussion of on-court basketball impact. Sorry, Bo and Sunny.

Most Valuable Real Plus Minus

So then what is on-court basketball value? No one can agree, so let's look at a few metrics. First, the plus-minus school of NBA stats. NBA basketball gives us plenty of scoring chances (wish we could say the same for the NCAA tournament...),  so it's easy to track which players have a positive points-impact for their team, and which players are a negative just by counting the team's points and points allowed while they're on the floor or off. Further, we can determine precisely how positive (more points for) or negative (more points allowed) players were by point differential. There are a number of variations, each tweaked and tuned in an attempt to produce more accurate data. For example, regularized adjusted plus-minus (or RAPM) attempts to improve predictive power by valuing data which falls within expected boundaries, and undervaluing outlier data (hence: it is 'regularized').

You, the reader, don't need to know every detail. Just where to look: you have RPM (real plus minus) statistics freely available on ESPN.com. They measure per 100 possessions (roughly a basketball game), so we can extrapolate to compare players regardless of their playing time. The result is Defensive RPM, Offensive RPM, and Total RPM: each number is the number of points said player is worth, compared to league average, per 100 possessions. League average is 0.

MVP Candidate	Defensive RPM (per 100 possessions)	Offensive RPM (per 100 possessions)	Total RPM (per 100 possessions)
Stephen Curry	1.82	7.34	9.16
James Harden	0.72	7.94	8.66
Russell Westbrook	-0.64	7.43	6.79
Anthony Davis	3.74	3.78	7.52
LeBron James	1.99	5.74	7.73
Chris Paul	0.61	6.82	7.43

LeBron James is an easy reference point for most fans. And although he's spectacular, the oft-cited "best player in the world" is not the most valuable to a team in 2015, by this measure. A few more observations: an ORPM over five is pretty much hall of fame worthy, yet we have three players over seven! And a fourth (Chris Paul) over six. As stated earlier, this is a really ridiculous MVP vote.

More observations: Russell Westbrook really, really stinks at defense. He's noticeably below average.

It's exceedingly rare to be truly great at both offense and defense. Even the King is merely "good" at defense according to DRPM.

Most importantly: Stephen Curry kind of kicks all sorts of butt. And yes, his TRPM is tops in the NBA (Harden and ‘Bron are 2^nd and 3^rd, respectively). Curry's 9.16 score is the highest in the two years ESPN has tracked the stat. While he isn't as valuable to a team's offense as Harden or Westbrook, his significantly better defense closes the gap, and then some, as far as RPM is concerned.

Common Counterpoint

Harden and Westbrook supporters will turn to one painfully relevant tidbit, no matter what number they hear: my guy does more with less. And it's tough to ignore that thought, especially with the way the Warriors have been pulling productive players out of seemingly every crevice. And to be sure, if I were a fan of another team, I'd be pretty darn jealous of a team bringing Andre Iguodala and David Lee off the bench.

But sports nerds insist of using statistics for the same reason scientists do: you can all but remove bias and emotion from the result. The media and your friends say that Harden and Westbrook are doing more because they have to carry a lesser team: but what does the data say?

Basketball-reference.com has an extremely handy, free-to-use tool called the ‘Plus/Minus Finder.' As you guessed, it allows us to sleuth out raw plus-minus numbers, sorted by any sample size, and filtered through any comb we see fit. Yay!

But first, what are we looking for? The raw plus minus is the actual, unaltered differential between how many points a player's team has scored with him on the court, minus the amount of points the team has scored without him on the court, each delivered per 100 possessions. In a sense, we're looking for the number which combines a player's defensive and offensive plus-minus contribution. We expect a positive number for all of our sample players because they are, of course, highly valuable. A negative net-point differential per 100 possessions would mean that the team was literally worse by playing said player. Speaking of which...

However, there are a couple problems with using raw plus-minus to weigh an MVP vote. One, as the best players on their own teams, an MVP candidate will play a lot of minutes. If a lesser player, hypothetically, gets all of his playing time with the MVP candidate, he'll have similar plus-minus metrics. If the two players always shared the court together, they would have the exact same plus-minus differential. For this reason, many NBA fans will dismiss plus-minus entirely, as it's dominated by seemingly every player on the team with the best plus-minus. For this reason, Draymond Green, Andrew Bogut, Klay Thompson and even Harrison Barnes all have elite double-digit plus-minus scores per 100 possessions.

There is a solution, however! Baskettball-reference.com also has an on/off stat listed for every player, by year. This number takes a player's total plus-minus contribution (points and points allowed), and compares it with the team's plus-minus score without the player on the court. With this number, we can get closer to seeing how valuable each player is by calculating how good the team is without him.

You've heard Harden and Westbrook supporters say that they're carrying their lesser teams to the playoffs. But you may be surprised to see the truth of the matter.

MVP Candidate	(On) Net Points per 100	(Off) Net Points per 100	On/Off Differential
Stephen Curry	+16.5	-1.0	+17.5
James Harden	+5.3	-2.9	+8.2
Anthony Davis	+5.5	-5.5	+11.0
Russell Westbrook	+4.3	-2.0	+6.3
LeBron James	+10.8	-6.3	+17.1
Chris Paul	+12.8	-7.6	+20.4

First thing's first: every single team on the list is below .500 without its star (even the 66-win Warriors are outscored without Curry). Literally speaking, everyone on this list is carrying their team to wherever they're going.

Secondly, and most shockingly, it looks like Westbrook actually has the best supporting cast outside of Golden State. The Cavaliers (!) and the Clippers (!!!) are significantly worse without their stars. If the award were only about carrying a bad team to the playoffs, we would disqualify Stephen Curry, James Harden and Russell Westbrook right now. Anthony Davis, LeBron James and Chris Paul are carrying much worse teams, according to on/off. Or more accurately, teams that perform much worse without their stars on the court.

The on/off differential paints a very different MVP ballot. It's Chris Paul, followed by Curry and LeBron per-100 possessions. They mean the most to their teams, and if that's how you define MVP, so be it. Long story short, it's nigh-impossible to make a case for Harden or Westbrook based on these numbers.

Stephen Curry, on the other hand, is worth an additional 17.5 points per game in scoring margin every 100 possessions. Last season, that was roughly the gap between the best team in the NBA (the San Antonio Spurs, with a 7.72 margin of victory), and the worst team (the Philadelphia 76ers, with a -10.45 MOV). Curry isn't literally carrying the worst team in the NBA. But when he's on the court, the numbers say he's just as valuable. And Chris Paul, by this metric, is even more valuable than that.

For sake of completeness, there are some more advanced statistics that fans will toss out there. Let's go ahead and acknowledge them now.

PER Pete's Sake!

John Hollinger's baby, Player Efficiency Rating (or PER) is an estimated evaluation of a player's efficiency, controlled for pace so that all players, regardless of minutes, can be compared on equal footing. It is a distant descendant of box score stats, using a variety of common statistics to punch out a single number representing a player's efficiency. League average is always set to 15.0.

PER is an ambitious idea that isn't so much flawed as it is regularly misinterpreted. It's common to see pundits refer to a player as the MVP because of his PER, or equate PER to value. However, PER is purely an efficiency metric, and it only allows for efficiency as can be quantified in actual player statistics - in other words, while RPM uses actual outcomes to determine each player's value, PER determines efficiency based on that player's points, assists, rebounds, blocks, and so on -- regardless of the team result. John Hollinger actually says this himself: the statistic is meant to help simplify stats so we can look at the rest of the picture -- not end a discussion.

Another huge problem with using PER as a catchall stat: it virtually ignores fundamental defense. If it can't generate a box score statistic (steal or block), it will not help a player's efficiency (this kills Draymond Green and Andre Iguodala's score, two of the league's very best team defenders, but it helps the score of a player like Anthony Davis, who blocks shots at a prolific rate but is otherwise a lesser defender). Let's take a look at some players up-and-down the efficiency spectrum, as defined by PER.

MVP Candidates	PER	Other Notables	PER
Anthony Davis	31.0	Brandan Wright	20.94
Russell Westbrook	28.8	Marreese Speights	18.66
Stephen Curry	28.0	Draymond Green	16.52
James Harden	26.8	Joakim Noah	15.58
LeBron James	26.0	Nerlens Noel	15.05
Chris Paul	26.0	Kent Bazemore	9.70

PER certainly has its uses. But determining a most valuable player is usually not one of them. But for what it's worth, Anthony Davis is indeed the league's most efficient offensive player, which sounds fair given his high productivity, relatively low Usage Rate (for an MVP candidate, at least), as well as his healthy TS%.

A Minute Blurb on Minutes

It's true: Paul, Westbrook and Harden each play more minutes per game than Stephen Curry. Westbrook's case is insignificant: by missing games, he's actually played significantly fewer minutes than Stephen Curry - so let's put that aside.

Harden has played a significant bit of extra time: 365 extra minutes. That is impressive, but ultimately, and unfortunately, it doesn't tell us as much as it seems. First of all, we don't know how Curry would play in an extra four minutes a night: maybe he would collapse from exhaustion (highly doubtful given last season), maybe he would go human torch for an extra few minutes a game (equally doubtful). However, we do know that the data supporting James Harden for MVP is cumulative, not per-minute (rate) based. Harden leads in categories like points per game, rebounds per game, and some other advanced box score stats (WAR, Value Added, Expected Wins Added). However, Curry leads all of these advanced numbers on a per-minute or per-possession basis, too.

At best, this would be relevant if Curry and Harden were mirror images of each other in the advanced stats profile. But since the large majority of rate statistics point to Curry, it's tough to hold four minutes a night as the only reason for voting one way or the other in an MVP vote. Especially when the reason for the fewer minutes is as follows:

Stephen Curry has been so good in 2015 that he literally wasn't allowed to play 15 whole times! For a country easily outraged by what it perceives as ‘running up the score,' it's funny that some voters would hold not running up the score against a player.

Per-36 Minutes Box Score Stats

Like I said: minutes aren't a great way to decide this debate. But if you're curious, here's the per-36 minutes comparison of the two presumptive MVP-favorites, from Basketball-Reference.com.

Harden: 26.8 points, 5.5 rebounds, 6.8 assists, 1.9 steals, 0.7 blocks, 3.8 turns (on .604 TS%)

Curry: 26.2 points, 4.7 rebounds, 8.5 assists, 2.2 steals, 0.2 blocks, 3.4 turns (on .639 TS%)

Pretty insignificant difference, don't we think? And that's why we have advanced stats - because the box score isn't always enough.

So who do I choose? Stephen Curry, though it's close. He leads Harden in every rate statistic I can think of, and it's essential that we use rate statistics because Harden and Curry don't play the same amount of time. Normally, I would agree with holding time played against a player (and for that reason, I penalize Anthony Davis and LeBron James for not being available all year). But this year, it is blatantly obvious that Stephen Curry is playing fewer minutes because he's so good--not because he's hurt or otherwise unavailable. Penalizing him for not playing garbage time is ludicrous.

In my opinion, the real debate is between Chris Paul and James Harden. I really struggled to decide on this piece, even now. Harden has the traditional numbers, Paul has the advanced numbers and defense. But Stephen Curry has both. Those are my top-3 regardless, because LeBron, Davis and Westbrook don't make up for missing more than 10 games each.

Whatever the second place decision, Stephen Curry will win the MVP because it's voted on by sports writers, and sports writers love Stephen Curry. He is the golden child, after all. He lead an epic resurgence ('surgence?) in Golden State, his family is well-known and respected, and he is outrageously popular. But make no mistake: he is the deserving MVP based on merit, too. Still disagree? Let me know all about it in the comments section, along with your own ballot.

My MVP Ballot

5. LeBron James (tie)

5. Anthony Davis (tie)

3. James Harden

2. Chris Paul

1. Stephen Curry

What WILL happen

5. Anthony Davis

4. Russell Westbrook

3. LeBron James

2. James Harden

1. Stephen Curry