Using Numbers to Predict the NBA MVP

Premise

This year, I was curious if a player’s box scores, his team’s record, and miscellaneous factors could accurately predict how likely he is to win the MVP. Wielding the results from past award voting—supplied by Basketball-Reference—I got around to establishing patterns among these types of indicators to forecast who’s most likely to win the regular season’s most prestigious award.

Voting considerations
  • Player value: This should be obvious—the building block of MVP voting. If there’s no indication you’re providing a high degree of value, you’re not in the running for the Most Valuable Player.
  • Team strength: Some voters have argued from the perspective that if you’re not contributing to winning basketball—in other words, if your team is low-seeded or doesn’t look like a contender—you don’t have a good case for MVP. There are more than enough anecdotes and testimonials through the years to support the idea that team record is a relevant factor.
  • Team weakness: Conversely, there are situations in which players are credited when playing with poorer teammates. I’d say it’s likely the ideal MVP in the minds of voters is able to carry teams to extreme heights without another All-Star teammate, thus reducing the error in assuming he’s an incredibly valuable player. Think of Stephen Curry and Kevin Durant in 2017. Both of them could’ve easily been called MVP-caliber players on clearly the best team in the league (67 wins), but received 5.1% and 0.2% of shares, respectively.
  • Historicity: Russell Westbrook’s triple-double averages were a major factor in his winning the 2017 MVP. He proceeded to do the feat an extra two times before being traded to Houston; and yet, he wasn’t an MVP candidate in those seasons. Westbrook earned 88% of voting shares in the 2017 race, a (comparatively) measly 7.5% in the 2018 race, and a negligible 0.8% in 2019! What does this suggest? The rarity of certain events and/or statistical achievements are factors at least somewhat independent of the achievements themselves.
  • Voter fatigue: Part of the spirit of the award—as an outsider observing voter tendenciesis not to simply reward the best players in each season. Otherwise, players like LeBron James and Michael Jordan might’ve racked up even more than they already have. It’s a little more interesting to see new faces and names at the tops of ballots, and makes for interesting story-telling when discussing the legacies of these players.
Modeling factors

There are obviously real factors in MVP voting that can’t be quantified. This year, Ja Morant is sixth in these projections despite it being unclear whether he’ll make an All-NBA team. Historicity and voter fatigue—while quantifiable and thus potentially relevant factors in the modeling process—were tested in the construction of this model and were concluded to be insignificant predictors in relation to the other variables. (When it doubt, went with simplicity.) But otherwise, a player’s MVP likelihood was estimated with the following:

  • Box scores: While raw box scores probably account for more of how voters intuit the process, adjustments were made based on how common each statistic was in a season. (This is analogous to “inflation-adjusted” statistics.)
  • Team record: This biases the results toward players on winning teams in such a way that predicts MVP likelihood with more accuracy than if it weren’t included.
  • Yearly adjustment: Because season strength varies, a player’s likelihood is adjusted based on who the other strong candidates are in a given season.
Historical accuracy
Season Pred. Winner (Actual Rank) Actual Winner (Pred. Rank)
2021-22 Nikola Jokic (1) Nikola Jokic (1)
2020-21 Nikola Jokic (1) Nikola Jokic (1)
2019-20 Giannis Antetokounmpo (1) Giannis Antetokounmpo (1)
2018-19 James Harden (2) Giannis Antetokounmpo (2)
2017-18 James Harden (1) James Harden (1)
2016-17 Russell Westbrook (1) Russell Westbrook (1)
2015-16 Stephen Curry (1) Stephen Curry (1)
2014-15 Stephen Curry (1) Stephen Curry (1)
2013-14 Kevin Durant (1) Kevin Durant (1)
2012-13 LeBron James (1) LeBron James (1)
2011-12 LeBron James (1) LeBron James (1)
2010-11 LeBron James (3) Derrick Rose (2)
2009-10 LeBron James (1) LeBron James (1)
2008-09 LeBron James (1) LeBron James (1)
2007-08 LeBron James (4) Kobe Bryant (4)
2006-07 Dirk Nowitzki (1) Dirk Nowitzki (1)
2005-06 LeBron James (2) Steve Nash (10)
2004-05 Tim Duncan (4) Steve Nash (11)
2003-04 Kevin Garnett (1) Kevin Garnett (1)
2002-03 Tim Duncan (1) Tim Duncan (1)
2001-02 Tim Duncan (1) Tim Duncan (1)
2000-01 Shaquille O’Neal (3) Allen Iverson (2)
1999-00 Shaquille O’Neal (1) Shaquille O’Neal (1)
1998-99 Karl Malone (1) Karl Malone (1)
1997-98 Shaquille O’Neal (4) Michael Jordan (3)
1996-97 Michael Jordan (2) Karl Malone (2)
1995-96 Michael Jordan (1) Michael Jordan (1)
1994-95 David Robinson (1) David Robinson (1)
1993-94 David Robinson (2) Hakeem Olajuwon (2)
1992-93 Michael Jordan (3) Charles Barkley (2)
1991-92 Michael Jordan (1) Michael Jordan (1)
1990-91 Michael Jordan (1) Michael Jordan (1)
1989-90 Michael Jordan (3) Magic Johnson (2)
1988-89 Michael Jordan (2) Magic Johnson (2)
1987-88 Larry Bird (2) Michael Jordan (2)
1986-87 Magic Johnson (1) Magic Johnson (1)
1985-86 Larry Bird (1) Larry Bird (1)
1984-85 Larry Bird (1) Larry Bird (1)
1983-84 Larry Bird (1) Larry Bird (1)
1982-83 Moses Malone (1) Moses Malone (1)
1981-82 Moses Malone (1) Moses Malone (1)
1980-81 Julius Erving (1) Julius Erving (1)
1979-80 Julius Erving (2) Kareem Abdul-Jabbar (2)
1978-79 Moses Malone (1) Moses Malone (1)
1977-78 Bill Walton (1) Bill Walton (1)

The model is right more often than it is wrong, correctly predicting 31 of the last 45 MVPs; and 96% of who the model predicts to win have at least been finalists. For the most part, there doesn’t seem to be a relationship between the year and the model’s accuracy, suggesting that these factors have been similarly relevant across time. The box score—as a traditional, unchanging snapshot of player tendencies, despite the analytics movement—has not lost popularity in higher circles. Team record and winning has always been important. 

Looking through the past results, the only player to really “break” the model was Steve Nash in the mid-naughts. Not only are his MVPs some of the most controversial to date—with opposers citing the heroic efforts of Kobe Bryant and Tim Duncan in these seasons—but Nash is reputed as a player whose skills and value are impossible to capture in the box score. Otherwise, every MVP winner since 1978 has been projected to finish in the top-4, which means one of this season’s group of Giannis Antetokounmpo, Luka Doncic, Joel Embiid, and Nikola Jokic is highly likely to snag the trophy this spring!



Leave a Reply

Your email address will not be published. Required fields are marked *