<
>

Ranking top four MVP candidates, based on advanced stats

How can advanced metrics help frame the MVP debate? Bill Baptist/NBAE via Getty Images

Russell Westbrook or James Harden? Or, alternatively, LeBron James or Kawhi Leonard?

From win shares to RPM, NBA player value metrics were designed to help inform our discussion about awards such as this year's fascinating MVP race. Yet it can be tricky to navigate the alphabet soup of various metrics to determine what advanced stats really tell us about the players involved.

Let's take a look at how I would rank the top MVP candidates, based on five important metrics:


VORP (Basketball-Reference.com)

Historical context: Westbrook's plus-15.3 box plus-minus (BPM) is far and away the best since 1973-74, the first season for which the stat was calculated. LeBron James' 2008-09 season is next at plus-13.0, and there's a larger gap between Westbrook and James than there is between the No. 2 and No. 14 seasons all time.

Based on BPM, Westbrook is on pace for 12.4 VORP, which would surpass Michael Jordan's record of 12.0 in 1988-89.

What it means: Frankly, Westbrook's season has broken BPM, which is not quite as flattering to him as it sounds.

To improve the quality of the rating for most players, BPM uses interaction effects that multiply a player's assist percentage by his usage rate and his rebound percentage. As you might guess, Westbrook's season is off the charts historically by both measures.

Here's the leaderboard for assist percentage multiplied by usage rate back through 1977-78, the first season the NBA tracked player turnovers:

And here's assist percentage multiplied by rebound percentage:

Basically, Westbrook's season is way outside the sample on which BPM was trained to estimate player value, making its estimate of his value unreliable. BPM is treating Westbrook's versatility as exponentially better than anyone else's on record, and that's surely an exaggeration.

In short, Westbrook probably isn't having the greatest year in modern NBA history, just the one best designed to maximize BPM.


Win shares (Basketball-Reference.com)

Historical context: Win shares are relatively down on this year's crop of MVP candidates. Projecting Harden's league-leading total would put him about 100th in NBA history. Win shares are available through the first season recognized by the NBA.

What it means: More than other value metrics, win shares tend to reward team success and defensive-minded bigs. (Rudy Gobert of the Utah Jazz is second to Harden with 13.9 win shares.) Those factors work against Westbrook, who ranks fifth in the league, by far his lowest ranking in any common value stat.

Win shares also don't factor in the key concept of replacement level, so they tend to reward minutes played, where Harden leads the league, more than productivity.


EWA (based on PER)

Historical context: Because Basketball-Reference.com has estimated PER back to 1951-52, we can use the formula for EWA, the wins-based version of PER, which is a per-minute stat, to come up with an all-time leaderboard. There are, however, a couple of caveats. Even more so than win shares, PER suffers from the limited player stats tracked before the 1970s, and creator John Hollinger set replacement level by position based on the player pool in the modern game, not historic norms.

Based on my calculations, Westbrook's projected 27.8 EWA would rank 37th in NBA history.

What it means: It's worth noting that PER tends to favor offense more than other value metrics, so it's remarkable from that standpoint that Leonard is fourth on a per-minute basis, just ahead of Harden, but behind Kevin Durant and Anthony Davis.


WARP

Historical context: Westbrook is on pace for 23.1 wins above replacement player, which would put him 28th since 1977-78 in my metric.

What it means: Westbrook padding his rebounding stats has been a key point of the MVP discussion. Because I calculate WARP, I can play around with the formula to see the value of Westbrook's rebounding. If Westbrook had 450 rebounds -- 5.8 per game, or 4.8 fewer than he's actually averaging -- he'd still be slightly ahead of Harden for first in the league in WARP.

At the same time, in part because of his rebounding, Westbrook rates as nearly as effective defensively as Leonard, and both Westbrook and Harden rate as above-average defenders based on their box-score stats. That's tough to justify.


RPM wins

Historical context: Because it requires play-by-play data, ESPN's real plus-minus (RPM) has only been calculated back through the 2000-01 season. Even over that relatively short span, this year's totals are unexceptional. As of the most recent update last week, James was on pace for 18.3 RPM wins, which would rank 24th in that span. Westbrook was on pace for 17.7, which would rank 26th.

What it means: RPM combines the two typical methods for determining player value, using a BPM-like metric based on box-score stats for stability while also incorporating how teams perform with a player on and off the court, adjusted for teammates and opponents.

As a result, RPM is the metric of choice for James fans. Because the Cavaliers have struggled so much with James on the bench, being outscored by 8.0 points per 100 possessions without him, per NBA.com/Stats, James leads the league in RPM wins even though his box-score statistics have not been as strong as those of the other MVP contenders.

It's possible, though, that Westbrook will pass him for the top spot by the end of the season, particularly if James is rested down the stretch.

RPM is relatively low on Harden's candidacy because the Rockets haven't declined much with him on the bench; they still outscore opponents by 4.5 points per 100 possessions when Harden rests. In particular, Houston defends better without Harden, so his minus-1.7 defensive rating in RPM is the lowest among the MVP candidates by a wide margin. Westbrook rates minus-0.4 points per 100 possessions on defense, which is better than it sounds, because guards are typically worse defensively than frontcourt players.

Leonard's defensive RPM (plus-0.9) has declined dramatically from last year's plus-3.9 mark. As I've discussed previously, opponent 3-point shooting explains a lot of why the Spurs have defended better with Leonard on the bench. Subsequently, Bo Schwartz Madsen of Nylon Calculus found that this effect was larger for Leonard than anyone else in the league. If Leonard's defensive RPM was about 3.3 or better, he'd be leading the NBA in RPM wins, so there's a case to be made for him here, too.

It's intriguing that Westbrook and Harden have also seen opponents shoot 3s worse against their teams while they're on the bench, though the effect isn't nearly as large as with Leonard. Cavaliers opponents have shot better from 3 with James on the bench.

Summing up the MVP race by advanced stats

Depending on your metric of choice and assumptions, you can use advanced stats to make a case for any of the four leading contenders, so be careful to avoid cherry-picking in favor of your candidate.

After looking at all these stats, I find it useful to consider players in terms of their impact per 100 possessions at both ends of the court. BPM, RPM and WARP are all similar in terms of how they estimate this impact, and here's how the MVP candidates compare.

Don't flatten the difference in skills by saying all four MVP candidates have been great offensively. It's clear by all three metrics that Harden and Westbrook have been the best offensive players because of how much offense they create for themselves and their teammates. Based on these, my best guess is that Westbrook has a small edge on Harden, with James and Leonard two to three points per 100 possessions behind.

Defense is tougher to measure statistically, but I'm inclined to consider RPM the most telling metric for Harden and Westbrook, because my scouting suggests their box-score stats overvalue them defensively. Because of the 3-point effect, I think box-score stats actually do better with Leonard, and I'm a little more skeptical of James' regular-season defense than some of the metrics suggest.

Combining those two guesses with minutes played would yield the following wins above replacement based on my estimate of replacement level (about 2.6 points worse than league average per 100 possessions).

Even though Leonard rates second to Westbrook on a per-possession basis in my subjective evaluation, his lower minutes total relegates him to fourth in my MVP rankings. Meanwhile, Harden's defensive disadvantage relative to James drops him to third with Westbrook comfortably leading the other contenders in terms of wins above replacement.

To me, the most valuable player is the one who adds the most wins to his team's bottom line, and that's why after consulting the advanced stats, Westbrook would be my MVP.

My top four:

1. Russell Westbrook
2. LeBron James
3. James Harden
4. Kawhi Leonard