NBA

The NBA MVP Voter Criteria is Deeply Flawed (Opinion Piece)

(📸 Business Insider)

Recently, I’ve scoured the internet for a clear and qualified description of what constitutes the definition of the NBA’s MVP Award, and unfortunately, these attempts have been fruitless. Perhaps this was done intentionally so that the ideas behind “value” could extend beyond thinking in the mass, but that still doesn’t stop people front searching for a universal criterion that can act as a “correct” interpretation of the “most valuable” player. However, in these efforts, there seems to be a widespread misunderstanding of what makes a player valuable. This post is admittedly and entirely an opinion piece, so while none of these ideas go without saying, I also won’t say the current state of the MVP voting rationale is without brokenness and deep flaw, and this is why.

Current Voter Tendencies

As discussed by a multitude of blogs, podcasts, and video before this, there are three main factors that go into how the voters will generally approach casting their MVP ballots:

  • Team success
  • Individual statistics
  • Narratives

As evident from talk shows like ESPN‘s “First Take,” it’s not uncommon to see the “best player on the best team” notion thrown around. Namely, some will say the MVP is the best player on the best team; and from what I can gather, it’s because the “best” player on the “best” team is supposedly impacting the ability to win at the highest level. While I’m diametrically opposed to this idea, and I’ll explain why later, it’s an undeniable factor in media voting.

Aside from knowing which teams are really good (which, in this case, takes a quick glance at the standings and nothing else), there also has to be some way to recognize the “best” players on those teams. This is where the box score comes in. While traditional box score statistics seem to be the most telling indicators of MVP winners (the historical MVP results that fueled Basketball-Reference‘s MVP tracking model found that points, rebounds, and assists per game were three of the four signaling variables in predicting voting outcomes, alongside team record), I will define this branch as more whole due to sparser but present references to advanced statistics like PER, Win Shares, etc.

Narratives are perhaps a less decisive, but still influential, part of the equation. Because it’s difficult to tell exactly how much impact these have on voting, we do have the examples of the noise that surrounded Kobe Bryant in 2013 and LeBron James in 2020. Both of these players were approaching their twilight years and, due to their ages, were garnering much more praise in major media outlets. Among others is Russell Westbrook having averaged a triple-double in the 2017 season. Although this is more of a statistics-driven case, there was a widespread significance to these numbers as Westbrook would be breaking a record set by Oscar Robertson back in 1962, so there were still strong hints of story-telling in this instance.

What does “value” mean?

Because the MVP is an acronym for the “Most Valuable Player,” it makes sense to vote for the award as it’s defined and choose the “most valuable” player; but what does that really mean? This, of course, means value needs to be defined. Even in basketball nomenclature, “value” is a loosely defined concept, which often leads to lots of dissenting opinions. Most recently, I’ve seen these types of discussions in the comment sections of MVP ladders and a delicately placed one of my own that I issued recently. Let’s look at some of the interpretations expressed in these forums:

  • “It’s flawed but the logic behind it is that if you can’t lift your team to a top seed, you’re not impacting winning at an MVP level.”

(This quote came from someone with a seemingly dissenting opinion, hence an “it’s flawed” beginning.) The logic outlined here suggests, for a player’s value to be validated, it has to materialize at the team level. But instead of that being a show of high lift, or larger differences in win pace or scoring margin with and without a player, it has to be the player’s team’s win percentage. This has been a concept I’ve struggled with for a long time, and that’s because it parallels a phenomenon I discussed in an earlier criticism of the work of the Instagram video podcast, “Ball Don’t Stop.”

Without going into the nitty-gritty of this event, the podcast’s host drew an unsound connection between scoring on the player level and scoring at the team level. Namely, there are many more ways other than scoring through which a player can positively impact his team’s point differential. The same logic applies to the improper connection between winning at the player level and winning at the team level. (Winning “at the player level” would simply be represented by a hypothetical parameter of exactly how many wins a player contributes to his team.) All it takes is a damning example in recent history to disprove this: Anthony Davis in New Orleans. From the 2015 to 2019 seasons, he was arguably one of the ten-or-so best players in the league, yet his team only managed to barely surmount the “average” mark (in SRS) two of those five years.

But does that mean, because the Pelicans were a sub .500 team, Davis’s ability to positively impact an NBA team is invalidated? By no means is this true. I’m not one to throw around impact metrics without attempting to make some adjustments for confoundment, but between those five seasons Davis clocked in at no less than 7.2 Win Shares in a season, making him an extremely likely candidate for the title of a “valuable player.” Contrary to popular belief, there is a lot of historical analysis that suggests a player becomes more valuable to a team at it becomes worse. The premise is that because the team becomes more reliant on the player and his skills are being put to more use in that situation, the increased role would mean a team’s win pace, barring confoundment from variance and/or team circumstance, would actually be impacted to a higher degree by any player as the remaining roster’s quality is weakened.

  • “I think it’s [team quality in MVP cases] a pretty large factor. You can’t be on a bad team and be the MVP, that shows a lack of leadership, even if ur team is dogsh*t. It’s the ‘Most Valuable’, u might be the most valuable person on your team, but when your team is meaningless, you’re not exactly valuable.”

This argument implies an axiomatic truth that states a player, if he so bears the value of an MVP-caliber player, must be able to transform the worst team possible into a “good team” (as it’s stated an MVP can’t be on a “bad team.”) Leadership attributes aside, let’s design a method to determine the probability of We know that the most extreme cases of player impact, based on records of with-or-without-you data (which measures the difference between the team’s schedule-adjusted margin of victory with and without a player from game-to-game) and APM data (estimates of a player’s impact on his team’s Net Rating, controlling for the quality of teammates and opponents), would say a GOAT-level player can add no more than +10 points to his team’s scoring margin each game; and even that measurement is quite generous.

So if the Basketball Messiah set foot on the court, we would expect him to be worth about +10 points to his team per game. Because, as an individual player, he’s about to embark on the greatest peak season in league history, he “should” in this case be able to transform any team into a “good” team. The worst team in history per Basketball-Reference‘s SRS was the 1993 Mavericks, posting a woeful -14.68 SRS, surprisingly in a full 82-game season. So if this amazing player, we’ll call him Cornelius, is worth +10 points per game and his cohorts are worth (about) -14.7 points per game, would the new team be a -4.7 SRS team? Perhaps, but a significant factor we need to account for is trade-offs in roles and how these teammates will scale alongside Cornelius.

Most superstars will play about 75 possessions per game in the modern era (roughly 36 minutes per game), but because Cornelius is so good, let’s say he’s a bit more of a heavy lifter and plays 40 minutes per game, playing just over 83 possessions per game. Because he’s on the floor for 83% of his team’s games (league-average paces generally tend toward 100 possessions per 48 minutes) and there are five players on the court at a time, we can estimate Cornelius carves out roughly 2.44 points of influence from his teammates, which in this case, would be -2.44 points per game. That means the additive efforts of his teammates now equate to a -12.24 SRS team. Therefore, with the addition of Cornelius’s +10 points per game, the new team is now a -2.24 SRS team. This would equate to a 35-win pace in an 82-game season.

But we don’t have to stop there; we can continue exploring the possibility that Cornelius does, in fact, make this historically-bad team a good team through a significance test. Namely, we’re trying to determine if the -2.24 SRS estimate is convincing evidence that Cornelius doesn’t turn the previous roster into a good team. Without going into the nitty-gritty of how this hypothesis test works, here’s a takeaway from the final result:

  • Assuming Cornelius would improve the ’93 Mavericks to average levels, the probability they would have a -2.24 SRS is 30.47%.

We can alter the parameters of these experiments to account for even more scenarios:

  • Assuming Cornelius would improve the ’93 Mavericks to the quality of an eighth-seed team, the probability they would have a -2.24 SRS is 31.81%.
  • Assuming Cornelius would improve the ’93 Mavericks to the quality of a championship contender, the probability they have would a -2.24 SRS is 4.92%.

These probabilities are obviously quite low. To increase the leniency of these situations, let’s look at how some of today’s players might fare in the current MVP race by plopping him on the worst team in the NBA right now: the OKC Thunder. Based on the latest APM data, a reasonable higher fence for a stable form of impact from the game’s best player is +6 points per game. Using the method from earlier, this new player (we’ll call him “Player B”) would alleviate 1.43 points of the Thunder’s SRS deficit en route to a -1.16 SRS team. While the higher quality of the current Thunder compared to the ’93 Mavericks allows for the lesser player to help them attain greater heights, no player in today’s game could lead the Thunder to a win-pace greater than a 9-seed team.

Using the probability method from earlier, let’s once again lay out some of the likelihoods for Player B:

  • Assuming Player B would improve the OKC Thunder to the quality of a Playoff team, the probability they would have a -1.16 SRS is 40.79%.
  • Assuming Player B would improve the OKC Thunder to the quality of an average team, the probability they would have a -1.16 SRS is 39.55%.
  • Assuming Player B would improve the OKC Thunder to the quality of a “good” team (one SRS standard deviation above average), the probability they would have a -1.16 SRS is 10.29%.
  • Assuming Player B would improve the OKC Thunder to the quality of a title contender, the probability they would have a -1.16 SRS is 7.97%.

Now, let’s take the whole landscape of “good” and “bad” teams for some final statistics. For context, Player B lifted each “bad” team to an average -0.25 SRS and no higher than an 0.37 SRS:

  • Assuming Player B improves a currently-existing “bad” team to the quality of a “good” team is 14.53%, the probability the currently-existing “bad” teams continue with a “bad” SRS is 14.53%.
  • Assuming Player B improves a currently-existing “bad” team to the quality of a championship contender, the probability the currently-existing “bad” teams continue with a “bad” SRS is 11.53%.

As we can see, the odds are not in Player B’s favor, despite being the very best player in the game. So what does this mean for this year’s MVP race? Aside from the likely candidates, there’s a heap of very valuable players that play for teams that aren’t particularly good, such as Luka Doncic, Nikola Vucevic, Zion Williamson, and even two of the strongest MVP candidates right now: Stephen Curry and Damian Lillard. But do the qualities of these players’ teammates make them any better or worse, any more or less valuable, of a player? All the evidence suggests not. Even the greatest imaginable player in league history wouldn’t improve the current OKC Thunder to a “good” team!

The major takeaway from today’s study:

  • DO NOT JUDGE THE QUALITY OF A PLAYER’S MVP CASE BY HIS TEAM CIRCUMSTANCES. SERIOUSLY, DO NOT DO THIS.

I’ve seen some rebuttals to reforming the NBA MVP voting criteria, both of which qualify as logical fallacies. The first is the Appeal to Authority: i.e. the voters think this way, therefore it’s the “right” way. But, of course, the fact that the voters may sway one way doesn’t confirm or deny any qualifications of what makes a player valuable. The second is the Fallacy of Sunken Costs, i.e. the voting has gone on this way for so long, so why change it? Namely, the continuation of this flawed criterion is for the attainment of some unsound form of achievement. But, as these arguments are heavily fallacious, why not spur change in your next conversation to actually change the topic to… the most valuable players?

The heavy inclusion of team success / three-pillar system as a focal point behind a player’s MVP case is deeply flawed, fallacious, and a massive embarrassment to the intellectual integrity of NBA basketball. To see such a complex topic be dumbed down to such measly levels is atrocious to me, and I hope this post helped reinforce the understanding of why I feel this way.

Leave a Reply

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