Is longevity really that valuable? Our intuitions of championship equity

In the top-40 career series, I evaluated hundreds of single seasons for the game’s greatest players. Then I added up the value of all those seasons to see who had the most valuable careers. But can you really just add seasons together like that? And does that process underrate higher peak players who lack great longevity?

After the original top-40 series, one of my big takeaways was how much we intuitively downplay longevity. Philosophically, that’s fine — plenty of people only care about peak. But what’s striking is that people who care about “career value” also genuflect at lower peak players like Reggie Miller ranking so highly, even if they agree Miller was quite good. So this post is going to put the entire idea of Championships over Replacement Player (CORP) under a microscope and tackle some of the trickiest issues with modeling multi-year player value. And hopefully, add some clarity to the “Miller-Walton conundrum” — is it better to have a short MVP peak, or a long All-Star career?1

Can multiple stars replicate superstar value?

For many people, non superstar seasons intuitively feel much less valuable than superstar seasons. It’s as if there is a threshold that players need to cross to reach championship heights, and only a certifiable “Number One” can cross that threshold. “Second bananas” can’t take you there. The thinking seems to be that all secondary players require a superior superstar teammate to win.

But as I’ve mentioned before, superstars almost always require secondary stars to win! They go hand in hand. And I think this is where the first “mental downgrading” of secondary stars takes place for most of us. It’s almost like we think:

  1. a superstar can win with no star teammates, and
  2. a superstar can win with a secondary star, but
  3. a secondary star can never win without a superstar.

If we give each of those premises equal weigh, it sounds like superstars are orders of magnitude more valuable than secondary stars. But should each of those premises be given equal weight?

First, superstars rarely win without All-Star teammates. It happens occasionally, but “Lonestar” teams have accounted for just six percent of the titles in the shot clock era,2 and that’s if we’re generous and don’t call Tyson Chandler an All-Star in 2011.

Second, superstars don’t just win with secondary stars in tow. They win because of secondary stars. Adding Pippen to Jordan is what made Chicago so good. Adding Ginobili and Parker to Duncan made them a dynasty. And so on. Most championships are forged with this synergy of stars, because All-NBA or All-Star level players move the needle quite a bit, even if they aren’t as important as Grade A superstars.

NBA basketball follows the Pareto principle, or an “80-20 rule” — most of the impact comes from a small number of players. Look at this 25-year study of player impact: just 21 percent of the players look like positives. Only 6.8 percent of players moved the needle by 2 points per 100 possessions, and just 2 percent were at 4 points per 100. So All-Stars have big impact, and they’re really scarce!

Sure, 2 to 4 points per 100 is never enough to “carry” a team to a title with no other stars. But ensemble teams do win championships without a superstar, just like superstars occasionally win championships without secondary stars. So our third premise from above is actually wrong — teams like the ’04 Pistons will break through, and they break through with greater frequency than the Lonestar teams like the ’94 Rockets.

Yet, it’s hard to think many fans would trade LeBron James or any MVP for two or even three All-Stars. Our intuition tells us that swapping out MVPs for lesser players doesn’t end well…although I’m not sure an MVP has actually been traded for two All-Stars, let alone three.3 Since it’s hard to find an instance where a team gave up two stars for an MVP,  front offices might actually value two All-Stars over a single superstar. Then again, that could just be a byproduct of salary cap rules.

Either way, it’s hard for four quarters to equal a dollar in basketball because there are only five spots on the court. Consolidating so much value into one player frees up four lineup slots to provide additional impact. This is why, in practice, it’s very hard to construct rosters without superstars that outperform rosters with superstars. The majority of the highest performing teams in league history were built around Michael Jordan, Magic Johnson, LeBron and so on.

Don’t sleep on the “ensemble” teams though. The 2012-14 Spurs were fifth in our latest playoff evaluations (back to 1984) of team performance, and the ’04 Pistons just missed the top-10 single-season teams. So while it’s generally easier to build super teams with stars — and the ceiling of super teams is surely higher than one without any superstars — the ensembles of the last few decades have hit some ridiculously high heights.4

So…can multiple stars match a superstar’s value? Based on the above evidence, it might be close, but superstars offer a higher ceiling. Trading an MVP for two All-Stars might not be an equal swap, but — but!! — longevity calculations are not about a single season! They are about a single roster spot over multiple seasons. We are asking: “Would a franchise be better with one MVP season and two replacement-level seasons from Joel Embiid, or three All-Star seasons from Embiid?”

Building Around Superstars

Even if superstars need secondary stars, it’s much harder to “build around” the 20th-best player in the league than an MVP-caliber talent. The great ensemble squads aren’t really constructed around a single player — those teams stack players that fit together and hope they’re greater than the sum of the parts. The way I calculate multi-season value for the top-40 careers series drops players on random teams — more on this below — but in practice, teams don’t randomly build around superstars.

For instance, with a ball-dominant superstar like LeBron, front offices can (relatively quickly) find 3-and-D players to space the floor around him. They then fill out the rest of the roster as needed to fill in the gaps. In this sense, teams with LeBron are less likely to play a bunch of rookies, less likely to sell for the future at the trade deadline, and generally more likely to field a competent roster that fits around James’s strengths. (As Lakers fans can attest to, that’s not always true.)

So while 85 or even 95 percent of players land on “random” rosters — meaning, they have no control over the team strength and construction around them — superstars actually dictate certain roster moves with their very presence. Even with someone like Jayson Tatum or Paul Pierce (roughly top-10 players), there’s some incentive to spend now and build around the franchise’s best player. The real question is, how much does this change the likelihood of playing on a bad, good, or even great team?

Here is the historical distribution of teams based on their strength, according to adjusted point differential (SRS):

Team Wins Frequency
64-73 1%
59-64 5%
54-59 9%
48-54 15%
42-48 17%
37-42 19%
31-37 14%
25-31 9%
21-25 7%
9-21 5%

One percent of all teams have played at a 64-win pace (or better) for an entire season, and five percent have played below a 21-win pace. But is it equally likely for an All-Star or MVP to play on one of these teams? Do Bradley Beal and LeBron James each have a 7 percent chance of playing with a 23-win supporting cast and a 1 percent chance of playing with a 65-win supporting cast?

Given the urgency to win that we just discussed, it seems less likely for superstars to play on such poor teams. However, these situations do pop up a ton: In the last 30 years, Hakeem Olajuwon, Charles Barkley, Kevin Garnett, Dwyane Wade, Chris Paul, LeBron James, and Anthony Davis have all registered superstar seasons in my book, while playing with supporting casts that would do well to win 25 games without them. If we loosen the criteria slightly, Kobe Bryant, Nikola Jokic, Allen Iverson, Deron Williams, Karl-Anthony Towns and Damian Lillard have all ended up in similar situations. Those situations might not last indefinitely, but they make me think that superstars play with really weak supporting casts at rates that aren’t too far off from normal players.

On the flip side, I do think it’s less likely for LeBron to have a 57-win supporting cast compared to some role player. This is tautological true — removing a role player from an elite team won’t change their quality much, but removing a superstar will bring them way back to the pack. Additionally, the economic structure of the league (soft cap, max salaries, free agency and so on) makes it less likely for stars to fall on “random” teams because they cost more.

But I’ve never been comfortable pairing economic modeling with player evaluations. If Dennis Rodman signed teeny contracts in the ’90s because no one understood his value, would that make him a better player? Not to me. Similarly, should we penalize good scorers because they are more likely to be overpaid as max contract players? Again, that seems arbitrary and separate from on-court play. Front offices worry about these factors annually, but I’m trying to retrospectively describe player performance, so the economics only complicate matters.5

Still, it’s clear to me that MVPs are slightly less likely to have 60-win supporting casts than role players in real life, even if it’s hard to say how much less likely. After all, Michael Jordan’s Bulls were at least a 50-win team without him. Kevin Durant joined a 70-win team. So great players do end up on great teams, even if it’s at a lesser frequency than supporting players.6

I actually think the more compelling advantage for superstars is the ability to optimize the team around them in a decidedly non-random way. When the NBA brought the 3-point line closer in 1995, it created more 3-point shooters to help Hakeem Olajuwon play 4-out offense. Any team he played on could pursue that strategy at the trade deadline or in the off-season. Since front offices have multiple years to build around an MVP, they can build for their specific skillset.

So while Lebron’s teams might only win 35 games without him, they’d probably win more with him than if he were added to a random 35-win team. All the other middling rosters around the league are probably not ideal for LeBron James. James might take an average 35-win team to 55-wins, but he can take his 35-win team to 60. And that’s a wrinkle that’s hard to model.7

Scarcity of stars or scarcity of seasons?

We’ve covered a lot of ground at this point. Let’s get tangible and illustrate how this longevity (CORP) calculation works, and why three All-Star seasons can eclipse an MVP season.

To begin with, we need to know how likely a team is to win a title based on how good it is. Given their regular season strength, we can calculate the typical quality of their playoff opponents and their odds of winning four series against those opponents. When we do that, we get a table like this:

Team Wins Frequency Title odds
71-73 <1% 89%
70-71 <1% 83%
68-70 <1% 75%
67-68 <1% 66%
65-67 <1% 55%
63-65 1% 44%
60-63 2% 32%
58-60 3% 22%
55-58 4% 12%
53-55 5% 4%
50-53 7% 1%
47-50 8% 0%
0-47 69% 0%

If you aren’t a 53-win team or better, you basically aren’t winning an NBA title. It’s a long shot even for 50-win teams, requiring luck and other elite teams to fold (from upsets, injuries, or some other factor).8 So pushing teams over that magical threshold is really important, but how many extra titles does it yield in practice?

For simplicity, let’s model one strong MVP season (high peak) and two All-NBA seasons (better longevity). For 2022, this could be something like Steph Curry (MVP) and Jayson Tatum (All-NBA). If you prefer comparing a player to himself, it could be 2006 LeBron versus 2015 LeBron. Either way, just think of an All-NBA player, then make him slightly better so he becomes an MVP.

Since 48-win teams never win titles, adding an All-NBA player to a 35-win team won’t cross the threshold into title contention. But when the MVP is added, the team will cross the threshold and contend for titles, even if they rarely win them. On those kinds of teams, two years of the All-NBA player doesn’t do much good because it’s almost impossible to win a title like that. In that case, one MVP season is better than two All-NBA seasons. It looks something like this:

Team Without Star Team w/MVP MVP title odds Team w/All NBA All-NBA title odds
38W 53W 2.0% 50W 0.3%

On an average 38-win team, our MVP’s championship equity is nearly seven times greater than an All-NBA player. And that’s based on just three extra wins of value. But I think this is where our intuition usually stops! We think of the case where a lesser player effectively can’t generate titles, and forget about all the teams where the lesser player also bumps his team into contention.

In the table below, I’ve presented all of the scenarios where an All-NBA player and a slightly better MVP player help teams win championships. From left-to-right: the first row starts with a 65-win team (1 precent of all teams), adds our MVP to take them to 73 wins (which has an 89 percent chance of winning a title), then adds our All-NBA player to take them to just 71 wins. A 71-win team has an 83 percent chance of winning a title, so over two seasons they have a 97 percent chance of winning exactly one title:

Without Star (win quality) frequency Team w/MVP player odds of 1 ring (w/MVP) Team w/All NBA player odds of 1 ring (w/All-NBA for 2 years) odds of 2 rings (w/All-NBA)
65W+ 1% 73W 89% 71W 97% 69%
62W 2% 71W 83% 70W 94% 57%
60W 3% 70W 75% 68W 88% 43%
58W 4% 68W 66% 67W 80% 30%
55W 5% 67W 55% 65W 68% 19%
53W 7% 65W 44% 62W 54% 10%
50W 8% 62W 32% 60W 39% 5%
47W 8% 60W 22% 58W 23% 1.5%
44W 8% 58W 12% 55W 9% 0.2%
41W 9% 55W 5% 53W 2% 0.01%
1-year Expected Value 0.19 2-year Expected Value 0.22 0.07

One MVP season with a 41-win cast more than doubles the titles of two All-NBA seasons with the same supporting cast. But as the team quality gets stronger — once the All-NBA player crosses the threshold and brings his team into title contention — having two bites at the apple becomes more and more valuable. And the really good teams start picking up a second title.

Now go back to the odds of playing with supporting casts of varying quality (the “frequency” column above) — a sizable chunk of teams will be 50-win quality, and the rest are so bad they won’t compete with titles even with an MVP. In that sense, a peak Bill Walton season doesn’t “guarantee” a title, and multiple Reggie Miller seasons will add plenty of championship equity in all the cases he’s on a 50-win team or better. The result is that two All-Star seasons produce more Expected Value than one MVP season.9

Takeaways

If we calculate the overall title increase from these players (giving them equal odds of playing on different quality teams), the MVP brings home a title in 19 percent of seasons, and the slightly worse All-NBA player in 12 percent of seasons. That’s a pretty sizable jump in title odds, even though we’ve been questioning whether MVP players are even better than the top-40 project thinks they are. Using a totally different method — one that, I believe, was based on empirical results — 538 actually came up with an incredibly similar model for calculating title odds between All-Stars, All-NBA players and MVPs:

That makes me think, despite some of the possible tweaks and issues we’ve discussed, that these multi-year value calculations are generally in the right ballpark, and that our intuition truly has a bias against longevity. At the end of the top-40 series, I calculated careers with slightly more value placed on MVP seasons and slightly less on All-Star seasons, and the results weren’t that different. Lower peak players all fell a little bit — John Stockton seven spots, John Havlicek eight spots and Miller nine spots — while higher peak players moved up, such as Steph Curry (eight spots) and Walt Frazier (seven spots). But longevity still reigned supreme. It’s hard to find All-Stars.

So even with some wiggle room in these models, it still looks like three All-Star seasons is comparable to a single MVP season. Once we have conversion rates like that, the value of longevity becomes clear. It’s great to have five MVP seasons to “build around” and create a contender, but a slightly lesser player is smashing championship equity by playing at an All-NBA level for a contending team over and over.

I do see a few potential areas to tweak that might change these year-to-year calculations slightly. The first is the issue of playing on good teams, where the odds seem to change as players grow better and better. This requires an economic component, and I’ve never landed on how to model that easily.

The second issue is how players impact better and better teams, or the “scalability curve.” I wrote about the complexity of this recently, where some players probably impact good teams and bad teams quite differently, and combined with the other factors discussed above, this makes for some tricky use cases to model. (We might call this the “Draymond Green rule.”) Either way, that issue seems more player-specific than something disrupting the peak-longevity calculus.

The final issue is simple: Most of our estimations of player impact are rooted in the regular season. While I try to use as much playoff data as possible — and most of that data echoes the regular season — there’s a possibility the overall scale of player impact is slightly broader than what the regular season suggests. If that’s true, and a typical MVP is worth 6-points per game instead of 5 (on a scale where an All-Star is worth 3), that would make a bigger difference in longevity calculations than any other issue raised in this article. At least according to my current CORP calculator.

For me, the major takeaways from the original research a decade ago still stand: (1) Floor-raising tends to be overstated, and thus overly glorified. (2) Secondary players are actually super important (team building!) and (3) thus holding value (scaling) in stronger environments is critical for title odds. And yes, (4) we underrate quality longevity. Trucking along as the 100th-best player in the league does nothing, but All-Stars are hard to come by, and All-NBA seasons are critically important. So while a sidekick is less valuable than a superstar, having 10 smaller bites at the apple instead of 3 big ones can certainly yield more championships…as counterintuitive as that may be.10

Team Records when MVP candidates play (OnWins and WOWY)

As discussed on a recent episode of the Thinking Basketball podcast, there’s been a lot of talk about the importance of “winning” when it comes to MVP voting (and even All-NBA) this season. With the regular season behind us, for posterity, I’ve compiled the team records for every candidate in games said candidate has appeared in.

Additionally I’ve added a column called “OnWins,” which is the record for each player based on his individual plus-minus in a game. (Shoutout to realgm’s Doctor MJ for creating this.) In other words, if the player finished with a positive plus-minus in the game, he is credited for a “win” individually, because his team won his minutes for that game. Here are the 13 candidates from the latest ESPN straw poll:

Player In Wins In Losses Win% Pace in lineup Win% Pace Out line Win% change OnWins OnLosses OnWin %
Devin Booker 56 12 68 47 21 49 19 72.1%
Chris Paul 53 12 67 53 14 45 19 69.2%
Jayson Tatum 49 27 53 27 26 54 22 71.1%
Nikola Jokic 46 28 51 21 30 50 20 67.6%
Giannis Antetokounmpo 45 22 55 33 22 44 23 65.7%
Joel Embiid 45 23 53 35 18 45 22 66.2%
Steph Curry 45 19 58 36 22 44 18 68.8%
Luka Doncic 44 21 56 39 17 38 27 58.5%
DeMar DeRozan 43 33 46 41 5 43 32 56.6%
Trae Young 40 36 43 41 2 42 33 55.3%
Ja Morant 36 21 52 62 -10 35 21 61.4%
Kevin Durant 36 19 54 24 30 36 17 65.5%
LeBron James 25 31 37 25 12 24 30 42.9%

Disclaimer: Again, I’m posting this for people who like to reference “team success” for individual awards. This should not be an end-point in individual player analysis if you are trying to determine how much impact or value a player has. In order to do that with this data, you should also try to account for teammates who are in or out of lineups, opponent quality, role, and so on, to say nothing of using on/off, adjusted plus-minus data, or hybrid metrics that combine this with box score factors in order to help determine impact, value, and/or winning. /disclaimer

Finally, I should note that the leader in Wins and OnWins this season was Mikal Bridges, with 64 wins and 62 OnWins, 17 OnLosses and 3 ties. Here’s the full podcast episode: