Time To Make The Sausage

There aren’t many certainties in today’s NBA, but beginning the month of May with MVP controversy is one thing you can always count on. There are no standardized qualifications for becoming the league’s official Most Valuable Player, and that creates a huge amount of inherent wiggle room, allowing voters to weigh different criteria in whatever way they see fit. That loose flexibility was shoved into the spotlight yesterday when Boston Globe columnist, Gary Washburn, revealed himself to be the lone voter who didn’t put LeBron James at the top of his ballot. Washburn went with Carmelo Anthony, and made his case public as part of yesterday’s announcement.

LeBron had an absolutely dominant season and it’s nigh impossible to find any reliable statistical metric by which he wasn’t the most productive player in the league this season. Washburn actually seemed to agree, and his argument was that although Anthony may not have been the better player, he was more important to his team. I’m not here to argue the merits of Washburn’s argument. But I would like to point out that this is an extreme example of separation between decision-making based on the power of statistics and the power of narrative. LeBron’s season presents some incredibly compelling storylines as well, but while there’s little space to argue against his statistical case, there’s plenty of room to argue about stories.

I don’t mean to imply that Washburn’s choice is somehow immature or incorrect because he gave more weight to the narrative elements of Anthony’s case. Stories are part of basketball; how we watch it, understand it, talk about it, and certainly how the media covers it. Stories are important and have always been a part of how the MVP award is decided. My own experiences as a basketball fan and amateur analyst are a constant balancing act between the narrative and the numeric. It’s an indelicate art and the line between the two moves constantly. One of the questions that the whole Washburn rigamarole raised for me was, exactly where that line falls for MVP voters in the aggregate. How much of MVP voting is based on statistics, literal or implied, and how much is based on a compelling story?

Narrative is an extremely complex idea to measure, but tracking the statistical case for MVP candidates is a little more straightforward. I began at Basketball-Reference’s Award Page, looking at the players who have received MVP votes over the last 10 seasons. Basketball-Reference is nice enough to include a limited statistical profile right alongside each player. The listed categories are age, games played, minutes per game, points per game, rebounds per game, assists per game, steals per game, blocks per game, field goal percentage, three-point percentage, free throw percentage, Win Shares and win shares per 48 minutes. My intuition is that any MVP voter who does include statistics in their decision making probably doesn’t look much further than these categories, and so they seemed like a reasonable place to start.

The one category which is conspicuously absent from a voting perspective is team win percentage, which I added. The other changes I made were dropping total Win Shares, keeping just the per 48 minute version, and converting total games played to percentage of games played, adjusting for the lockout shortened season. I then regressed those categories onto the share of total possible points that each player received from the voters. The result was an R^2 value of 0.516, which means just over half the variation in MVP voting can be explained by players’ performance in those categories I mentioned above.

While that explains a significant block of variability, it still leaves nearly half of the story untold. That 0.484 is where the narrative comes in. The results of the regression analysis also include an equation by which you can project the share of possible MVP voting points a player should have received, based on those numbers. I did that for each of the top five vote-getters from those 10 seasons and put them in to this Tableau Visualization, along with the actual share of MVP vote they received.

[iframe]<iframe src=”http://public.tableausoftware.com/views/NBAsMostValuablePlayerResults/MVPVotingDashboard?:embed=y&:display_count=no” style=”border:0px #FFFFFF none;” name=”MVP Voting” scrolling=”no” frameborder=”1″ marginheight=”0px” marginwidth=”0px” height=”663px” width=”663px”></iframe>[/iframe]

You can play around and sort by year, looking at how each race shook out. The higher a player is on the vertical axis the more compelling their statistical case was. I’m making an assumption here, but the implication is that the difference between a player’s projected share and their actual share represents the power of their narrative. Player’s who fall low on the vertical axis, but far to the right on the horizontal axis would appear to be the ones with the most compelling narratives.

I put this visualization together for you to draw your own conclusions, but I’ll share I few seasons I found particularly interesting.

2013

2013 MVP

This was a year where the narrative component of the MVP voting went hand-in-hand with the statistical rationale. LeBron and Durant had big statistical edges and it was clearly reflected in the results. But those numbers also fell in with the storyline of two dominant stars elevating their games and leading their teams to a new level. I also thought it was interesting how much of a difference narrative made in the case of Carmelo Anthony. We already discussed how his story swayed Gary Washburn, but he apparently wasn’t the only one. Anthony finished third in this year’s voting despite a weaker statistical component to his case than either LeBron, Durant, Kobe Bryant or Chris Paul.

_____

2011

2011

This was one of the most memorable MVP votes for me and really exemplified the divide between analytic-minded decision makers, who advocated for Dwight Howard, and those drawn to the compelling one-against-the-world narrative of Derrick Rose’s season. In the end the award went to Rose, by a healthy margin. Amazingly, the regression equation seems to indicate that LeBron had a much stronger statistical case than either Rose or Howard, despite finishing third. This is a case where the negative narrative of the Heat’s ‘front-running’ and the ‘post-Decision’ backlash probably kept LeBron out of the top two spots.

_____

2008

2008

2008 was another fascinating year in terms of balancing narrative and production. There was a lot of push for Chris Paul who jumped several levels in production, leading the New Orleans Hornets’ to the second-best record in the Western Conference, along with building the most compelling statistical resume of the candidates. In the end he lost out to Kobe Bryant, who trumped Paul’s narrative with a career of dominance, that had at that point been unrecognized with an MVP award. Kevin Garnett finished third for his work in coalescing the Big Three in Boston and leading the Celtics to the best record in the league. LeBron James finished fourth, with the second-most compelling statistical resume but no enticing story to attach it to.

_____

2006

2006

This is another infamous award season. It was Nash’s second consecutive MVP, despite being the worst for the Nash-D’Antoni Suns, both in terms of wins and offensive efficiency. But it was a remarkable and, at the time, almost unbelievable duplication of what they had done in their first season together. This was especially true when you consider that Amare Stoudemire played just three games all season long. That the ‘Seven Seconds or Less’ philosophy was able to sustain into a second season and prove a viable offensive strategy that wouldn’t dissipate once it was “figured out” by NBA defenses was the narrative that drove Nash to this award. LeBron finished second in the voting, but he was one of three players, along with Dirk Nowitzki and Chauncey Billups who had a more compelling statistical case.

_____

People on both sides of the narrative-numerical divide often seem to get their hackles up around the MVP Award, depending on which side prevails in a given year. While middle ground we currently walk always leaves someone frustrated, it’s by far preferable to the alternative. There is a place for logic and reason in the NBA and no one would be satisfied by a world where postseason awards were handed willy-nilly with no verifiable, objective reasoning to support those decisions. At the same time, making decisions with a formula only denies our human instinct to create, tell and consume stories. It may be a bumpy ride, but you can enjoy the MVP award both for what it is and for what it is not.