What is portability? Portability (sometimes called scalability) can have a variety of meanings. In the traditional meaning, popularized by Thinking Basketball, portability is how well “players carry more value on better and better teams.” (source: https://thinkingbasketball.net/2018/02/12/backpicks-goat-philosophy-of-player-ranking/ ).
A player’s value pretty irrefutably faces diminishing returns as their teammates get better and better. Different skills face different amounts of diminishing returns. Some skillsets are better at “floor raising,” taking a bad team to an average team. Other skillsets are better at “ceiling raising,” taking an average team to a championship-level team. Good defense is usually maximally portable (“most clubs need defenders”, per the article above), while offense can have a greater range of portability (“there’s only one ball”, as the adage goes). Maintaining more value on better teams is important, as generally only the most dominant teams in a season are capable of winning championships. Thus, if you’re interested in how much a player improves their teams’ championship odds, incorporating portability as one facet of your evaluation can produce more accurate results.
Of course, as proponents of portability readily say, it is far from the only factor. The overall value of a player is likely more important than that player’s portability. And the importance of a player’s portability might change depending on how good they are. In the traditional portability model, the importance of portability increases as the player gets better. This is backed up by team results throughout history: a team’s chances of winning a championship increases more as you go from 55 to 65 wins vs as you go from 45 to 55 wins, so facing fewer diminishing returns on those dominant dynasties can really improve your championship odds.
Still, there are opponents of portability who argue against certain facets of the model. Some refute how portability changes with player value, saying portability is more important for role players than star players. Others argue it’s not a significant variable (e.g. your evaluation would be no worse if you rated everyone as having +0 portability), and a minority of people even argue negative portability skills are better.
In general, I think there’s room for more detailed study of portability, the evidence for it, and how important it is in for good players and good team results. Here, I’ll provide evidence for the existence of portability and note a few interesting trends across several studies: A) trends in overall team results, B) trends in offensive team results, C) trends in player impact data, D) trends in superstar on-court offensive team results, E) film evidence, and F) expert opinions.
A. Trends in Overall team results
Good portability is supposed to be a measure of strong ceiling raising. Is this true? Let’s look at the Top 100 teams according to two overall team performance stats: ELO (which measures team record, adjusted for opponent, in the regular season + playoffs) and Sansterre’s Overall SRS (which measures team margin of victory, adjusted for opponent, in the regular season + playoffs). Then we can compare these teams’ impact metrics with the offensive portability of their best player to see if there’s a trend.
Why start with overall team performance and not offense-only results? We’ll do both. But among the top offensive players and top overall players, it’s not uncommon to have an offense-focused roster around a low-portability player or a defense-focused roster around a high-portability player. This could bias the signal, so we’ll look at overall team results first.
To be clear, I’m using a very crude model: it doesn’t include any other information besides the best player’s offensive portability score. I expected the trend to be insignificant. It turns out, there is a signal that portability improves ceiling raising.
1. ELO
-Data: Top 100 Teams according to ELO (1955–2023)
-Method: Linear Fit of portability score of top player vs Composite ELO score. Note: for this study I’m using Thinking Basketball’s older portability scores from a year or two ago.
-Slope: +6.82 [ELO / port].
The mean difference in ELO between neighboring ranks is 1.56, while the median is 1, so an improvement of +1 portability in your best player is estimated to improve a Top 100 team’s ranking by 4–7 slots.
Replacing a -2 portability with a +2 portability best player is estimated to improve a Top 100 team’s ranking by 17–28 slots.
-p-value: 0.02 (good). This is statistically significant! This is a real feature in the team results.
-R^2: 0.05 (bad).
This trend is not sufficient to explain all the variation in the team results. Which was self-evident, even before calculating. It seems the portability of a team’s best player is a statistically significant factor, but it is obviously far from the only factor.
-Change in slope across the data: +7.16 (top 50 teams), +0.09 (bottom 50 teams).
So the portability of your best player becomes more important on better teams. This is exactly what Thinking Basketball predicted: the most dominant teams are subject to the most diminishing returns, and so gain the most benefits if their best player (who has the most value on the team) has better portability (and so faces less diminishing returns).
2. Sansterre’s Overall SRS
-Data: Top 100 Teams according to OSRS (1955–2023)
-Method: Linear Fit of portability score of top player vs the OSRS ranking value (a combination of OSRS, OSRS standard deviation, and playoff series won, which Sansterre used to rank teams in his list). Note: for this study I’m using Thinking Basketball’s older portability scores from a year or two ago.
-Slope: +0.75 [OSRS ranking score / port].
The mean difference between neighboring ranks is 0.22, while the median is 0.09, so an improvement of +1 portability in your best player is estimated to improve a Top 100 team’s ranking by 3–8 slots.
Replacing a -2 portability with a +2 portability best player is estimated to improve a Top 100 team’s ranking by 14–33 slots.
-p-value: 0.04 (good). This is statistically significant! This is a real feature in the team results.
-R^2: 0.04 (bad).
This trend is not sufficient to explain all the variation in the team results. Which was self-evident, even before calculating. It seems the portability of a team’s best player is a statistically significant factor, but it is obviously far from the only factor.
-Change in slope across the data: +1.02 (top 50 teams), +0.18 (bottom 50 teams).
So the portability of your best player becomes more important on better teams. This is exactly what Thinking Basketball predicted: the most dominant teams are subject to the most diminishing returns, and so gain the most benefits if their best player (who has the most value on the team) has better portability (and so faces less diminishing returns).
So in both cases, teams with a more portable best player were likely to produce more dominant results. An improvement of +1 in your best player’s portability score correlates with an improvement of +3–8 ranks in the Top 100 lists, and the importance of portability seems to increase as you get to better teams. The trend is statistically significant: better portability does produce better ceiling raising.
B. Trends in Offensive team results
In Part A, we looked at trends in the best overall players’ offensive portability scores compared to the overall team results, among the best overall teams. Here, we look at trends in the best offensive players’ offensive portability scores compared to the offensive team results, among the best offensive teams. Teams: As before, we need a large enough sample that we’re not dominated by small-sample noise, so let’s look at the Top 50 offensive teams ever. We’ll use relative offensive rating, as the best teams ever by raw offensive rating will be primarily modern teams, which would make samples across the variety of top teams/styles/players smaller and thus noisier. As an aside, the recent boost in raw offensive ratings have been improved first and foremost by the improvement in three point shooting, which happens to be one of the most portable skills.
Players: We’ll use Thinking Basketball’s portability scores. To find each teams’ best offensive player, we’ll use Thinking Basketball’s offensive evaluations. In most cases, the most valuable offensive player on each team is clear. Generally, quibbling over if Player X is actually better offensively than Player Y in the small sample of teams where it’s ambiguous won’t change the overall results.
Prediction: The average portability score among players is 0, and we might expect it to be roughly normally distributed (so most players have +0 port, fewer players have +/- 1 port, fewer still have +/- 2 port). If portability exists, we would expect the top offensive teams to skew slightly towards positive portability when looking at the top offensive teams ever.
Among the Top 50 Teams according to RS rORTG:
-Number of best offensive players with +2 Port: 7
-Number of best offensive players with +1 Port: 8
-Number of best offensive players with +0 Port: 21
-Number of best offensive players with -1 Port: 10
-Number of best offensive players with -2 Port: 0
-(Number of best offensive players without Thinking Basketball Port scores yet: 4)
-Average port score: +0.3.
-Positive vs negative scores: 13 have positive port (7 have +2), 10 have negative port (0 have -2).
We can do a statistical test called a (one-sample) t-test to determine if the mean portability in this sample is significantly different (in the statistical sense) from the expected mean of 0. If the (one-tail) p-value from this t-test is less than 0.05, we can conclude there is a real shift towards positive portability, and that this is not just noise.
-p-value: 0.038 (statistically significant!)
Among the Top 50 Teams according to PS rORTG (minimum 10 games played):
-Number of best offensive players with +2 Port: 7
-Number of best offensive players with +1 Port: 12
-Number of best offensive players with +0 Port: 14
-Number of best offensive players with -1 Port: 10
-Number of best offensive players with -2 Port: 0
-(Number of best offensive players without Thinking Basketball Port scores yet: 7)
-Average port score: +0.4.
-Positive vs negative scores: 19 have positive port (7 have +2), 10 have negative port (0 have -2).
-p-value: 0.011 (statistically significant!)
So among the top offensive team results ever, as predicted, we find the most common value is neutral +0 portability, with the distribution skewing more towards positive portability than negative portability. In both the RS and the PS, the average portability is slightly positive, and there are more positive portability best offensive players than negative ones. Better portability in your best player seems to produce better team offenses. This trend is statistically significant. This supports the idea of portability.
The general qualitative takeaways don’t change if you change the sample (e.g. shrink the sample to the Top 25, or change the playoff minimum games played filter), and there are similar trends if you do a similar study across team playstyle (e.g. movement-based offenses vs two-man offenses vs heliocentric offenses, etc.), although the exact numbers would obviously change.
C. Trends in Player Impact Metrics
Is there a signal in the impact data of individual stars? If so, including portability in an evaluation of a player should produce a better fit to available impact metrics than evaluations that ignore portability.
Since we’re using Thinking Basketball’s concept of portability, let’s use their evaluations to filter players. Let’s take every season that’s worth 4+ in Thinking Basketball’s evaluation since 1997. There are 28 players and 164 seasons total.
To test if portability is a better model, let’s compare Thinking Basketball’s valuation and CORP without including portability vs their CORP including portability. The best model should produce a better fit with impact metrics, such as plus minus, on-off, and Augmented Plus Minus. As a reminder, lower p-value is better, and higher R^2 is better.
Fitting with Plus Minus (on):
-Raw valuation (no portability): 0.0004 p-value, 0.07 R^2
-CORP (no portability): 0.0006 p-value, 0.07 R^2
-CORP (with portability): 0.0003 p-value, 0.08 R^2
Fitting with On-off
-Raw valuation (no portability): 0.0006 p-value, 0.07 R^2
-CORP (no portability): 0.0006 p-value, 0.07 R^2
-CORP (with portability): 9e-5 p-value, 0.09 R^2
Fitting with Augmented Plus Minus (AuPM)
-Raw valuation (no portability): 6e-13 p-value, 0.28 R^2
-CORP (ignoring portability): 5e-13 p-value, 0.28 R^2
-CORP (with portability): 3e-15 p-value, 0.32 R^2
So in all three cases, including portability produced a clear improvement in how well the player evaluation fits the impact data. There’s more rigorous ways to compare models if people would like. And there’s other possible explanations (maybe higher portability players do better than lower ones for non-portability reasons). But as a first-pass, it seems if two players are otherwise equal in value, the one with higher portability tends to have better impact metrics than the one with lower portability.
D. Survey of Offensive team results when the Superstar is on-court
Overall team offensive data can be biased by bench offenses, so it’s worth looking at the offensive team data just when star players are on the court in the playoffs. It would be best to look at rORTG; but since it’s quite painstaking to calculate the proper rORTG when a star’s ON in the playoffs with differing defenses (and could be biased by e.g. starter vs bench defenses), let’s use a first-pass proxy: raw ORTG when the star’s ON. It’s imperfect, but if you remember the uncertainty bars, it’s still informative.
Since era differences will make a major difference in raw offensive rating, let’s look just at the top 8 offensive players of the pace and space era: Curry, Jokic, LeBron, Durant, Chris Paul, Harden, Luka, Kawhi. (Purely to satisfy my own curiosity, I’ll include Nash, but feel free to ignore him if you’d prefer). For context, here are the top players’ portability scores: Curry +2, Jokic +2, Durant +1, Kawhi (+1 Spurs, +0 post-Spurs), LeBron (-1 pre-2017, +0 post-2017), Harden -1, Chris Paul -1, Luka -1, Nash -1.
-Best 3-year PS ORTG when ON: Durant +120.4, Curry +120.3 > Jokic +118.5 > LeBron +118, Kawhi +117.8 > Luka +117.1 > Harden +116.1, Nash +116.1 > Chris Paul +111.6
-Best-fit linear Slope: 1.54 (better portability produces better results). p-value: 0.047 (it’s statistically significant). R^2 value: 0.51 (portability does a good job explaining the variance in the data)
(note: these fits are done removing Nash to provide a more fair same-era comparison.)
-Best 5-year PS ORTG when ON: Curry +119.4, Durant +119.3 > LeBron +117.7 > Jokic +117.3, Kawhi +117.3 > Nash +116.3 > Luka +115.8 (4-year avg) > Harden +114.7 > Chris Paul +111.4
-Best-fit linear Slope: 1.50 (better portability produces better results). p-value: 0.038 (it’s statistically significant). R^2 value: 0.54 (portability does a good job explaining the variance in the data).
(note: these fits are done removing Nash to provide a more fair same-era comparison.)
This peak team rating almost exactly follows the order predicted by portability, which itself is just assigned based on playstyle. The positive portability players are very clearly on top, the neutral portability players fall in the middle, and the negative portability players clearly lag behind.
There are similar trends if you look at other peak time ranges, like 2 or 4 years. Adding regular season data does not significantly help the negative-port players. Again, this is a crude metric. A larger sample and more context (e.g. checking relative on-court ratings) would be necessary for this to be more conclusive. But in a survey of the best recent offensive stars, the more portable stars seemed to have better on-court offensive team results. This is supportive of the existence of portability, especially when taken in conjunction with the rest of the evidence in this post.
E. Film Evidence (and various other studies)
The primary meat of this post was intended to be the new statistical analysis. However, it's worth noting that there’s also film evidence for portability. Skills like shooting (with its efficient finishing and spacing benefits), off-ball action (e.g. movement, screening, rebounding), passing, and efficient finishing tend to have high offensive portability, while traits like isolation scoring and ball-dominance tend to have low portability. Unfortunately there’s no one spot where Thinking Basketball goes through a comprehensive study of the portability of all different skillsets. However, there are a variety of places where he or others go in-depth on facets of portability. Here's a few of them.
An analysis of the value of off-ball motion:
Spoiler:
A case study on how a variety of portable skills (motion, playmaking bigs, handoffs with screening, shooting, and pace) can produce one of the best raw Offensive Ratings ever:
Spoiler:
The value of three point shooting should be pretty obvious to everyone on this board. It’s more efficient than long twos and so makes finishing with the help of a playmaker more valuable. The spacing makes it easier for players to drive or attack the rim, harder for the defenses to help certain actions, and more punishing when they do help with wider passing lanes and a longer distance to recover. Still, for completeness, it’s worth including some content on the three point revolution.
Spoiler:
A review of how recent offenses use motion (off-ball cutting) to strain defenses
Spoiler:
F. Reported Evidence
There are numerous quotes from players, coaches, and analysts in support of portability or related ideas. Quotes are far from conclusive. Indeed, just because a player is good on-court does not mean their on-court skills translate to being good off-court analysts. Even players with good on-court basketball IQ can be faulty off-court analysts, or be biased in their evaluations by personal feelings, relations, or loyalties. Nonetheless, since it’s fun, I’ll include a survey of quotes in support of portability and related ideas.
Wilt on Bill Russell’s fit:
So the ability to fit often requires not needing the ball in your hands so you don’t take the ball away from your teammates, being versatile with good passing and rebounding, and also defense. Defense faces minimum diminishing returns, and so always receives a maximum portability rating from Thinking Basketball, but too much individualism on offense can lead to diminishing returns and limit how much you win.“I picked him [Bill Russell] as the number one center of all time, because he was a complete basketball player…. I would pick him over me, because he also helped his team to win maybe a lot more than I could help my team to win. Sometimes the mere power of you makes you more individualistic. I have said this before: Wilt Chamberlain on the Boston Celtics might not be the same. Because I would take away from Bob Cousy, from Tom Heinsohn, because I was a scorer too. And then all of a sudden they would have to pass the ball to me, and that would take the ball away from them. So sometimes less is better…. The man could score. He averaged 17 ppg, which is only about 7 points less than Kareem Abdul-Jabbar. He was the supreme rebounder. He got the ball for his team, which made it happen. He’s high in assists, and we all know he played defense. You can’t ask for any more from his position.”
Moses on rebounding:
So being more active off-ball on the board can tire opponents out.“You keep moving, and it’s going to pay off. He’s going to be so tired fighting you, he won’t have anything left. Things change in that money (fourth) quarter. A man figures he has boxed you out for three quarters, he’s confident. He’s dead-tired, so he lets down a little. Now he kinda forgets about me, but I’m still coming. The only time I have trouble getting a rebound is if they put two or three guys on me. If it’s only one guy, I can get the rebounds.”
Cooper on Bird:
So off-ball action requires greater endurance, defensive attentiveness, defensive IQ, and gives an offensive greater versatility, which can be harder to guard.“People always ask me who’s the hardest player I ever had to guard…. I always say Lard Bird…. With Larry, when he passed the basketball is when he became more dangerous. He was either setting a pick, coming off a pick, catching the ball, passing the ball, so he was the one you always had to stay attentive to the whole 24 seconds of that offensive play.”
Phil Jackson on Triangle and Jordan:
So sharing the ball, reducing too much volume scoring, and running off-ball actions helps teammates grow and stresses the defense at multiple spots on the floor, ultimately leading to winning basketball.“Basically I was planning to ask Michael, who had won his third scoring title in a row the previous season, to reduce the number of shots he took so that other members of the team could get more involved in the offense…. I told him that I was planning to implement the triangle... ‘You’ve got to share the spotlight with your teammates,’ I said, ‘because if you don’t, they won’t grow.’…. Sometimes I would tell him that he needed to be aggressive and set the tone for the team. Other times I’d say, 'Why don’t you try to get Scottie going so that the defenders will go after him and then you can attack?’ “
LeBron and Redick on 2011 Miami:
So 3 point shooting, having floor spacers, and broadening your game to focus on skills that complement with your teammates, rather than focusing on putting up big counting stats, is essential to winning championships.LeBron: “Obviously my first year there, played great basketball, got all the way to the finals, lose in the finals. I play like ****…. When [Spo] came back to us, he knew that in order for us to reach our potential, one I had to be 10x better than I was in that previous June finals, but [two] Chris Bosh had to go to the five…. And we had to spread. He had to start working on his corner 3 faithfully….”
“The Bosh spacing, what did that sort of unlock?” -Redick
“The cutting. Slot cuts…. It unlocked exactly what myself and D Wade thrive on….
It changed the whole team. Yeah we added Ray. Added Shane. Added Mike Miller. We added the Spacing….” -LeBron
“…. Basketball is a very organic thing. And the players and their skills have to complement each other. And Chris Bosh is a great example of that. The sacrifice to figure out, how can my skills (and maybe I have to develop some of those; you mentioned the three point shooting), how can I figure out how to complement. It’s going to make me better, it’s going to make LeBron better, it’s going to make D Wade better, and it’s going to make our team better. And that’s basketball.” - Redick
“And that’s basketball. But that also comes from (to go back to episode one) basketball IQ. Him having the basketball IQ and the knowledge of saying ‘yeah, I could still be in Toronto averaging 25 and 12, but I didn’t come here for that ****. I came here to win championships. And we **** lost in year 1. What can I do to complement my teammates? What can I do to broaden my game out to where we don’t lose in year two?’ “ - LeBron
Greg Popovich on motion offenses:
So ball movement and unselfish player movement can improve shot quality.“We’re always trying to move the ball from good to great (shots). Penetrate for a teammate, not necessarily for yourself.”
Alex Caruso on Curry:
So when a star has good off-ball motion, good shooting, and good chemistry with smart teammates, all in a motion system, it becomes borderline unguardable.“That dude’s a menace. He’s just, he’s impossible. It’s especially the way they run their offense and the way that him and Draymond have this telepathy. Like the guys that are back this year… they’re actually understanding how to play with him now, how he operates, how he moves, when he gives up the ball and continues to run in circles until you fall asleep and even then shoots it from 30 feet. It’s borderline— at times it’s unguardable.”
Curry on Jokic:
So when a player has portable skills like 3 point shooting, playmaking, offensive rebounding, combined with motion and shooting around them, it makes it harder to defend.“He puts pressure on you all game and you got to respect, not only his ability to post up and put pressure on the paint. He knocked down 3 threes tonight. He is obviously an amazing playmaker so when they’re doing all that motion around and him, everybody has to be on alert to take away passing angles and lanes because he sees the court so well. And obviously even just offensive rebounding. There’s tip ins… you play good defense and he still finds his way right to the basket…. So he’s got kind of the full package, and guys are hitting shots around him and it makes it super tough to figure out how you want to defend him.”
Conclusion
There are definitely limitations to a study like this. Portability is qualitative concept. There’s general principles and common ideas, but there isn’t a single universally agreed definition. Reasonable people can disagree on the exact weights for how portable different skills are (see e.g. Thinking Basketball vs CraftedNBA’s portability scores). Even if you agree on the weightings, you could still disagree in your evaluation of a specific player’s portability in a specific season (e.g. Thinking Basketball has even updated their portability ratings for certain players).
Still, a survey of data across overall team results, individual impact metrics, offensive team results, and on-court team performance finds that portability (such as the model proposed by Thinking Basketball) exists and is statistically significant. And the portability of a team’s best player seems to become more important as the overall team quality. We can back this finding up, provide deeper explanations, and explore the necessary context using additional film analysis and player discussions.
There’s also room to disagree on exactly how important portability is (although the statistical studies above give us an approximate sense for a good team’s best player). Most everyone would agree it’s far from the only factor, or the most important factor. It’s far from deterministic: portability measures skills that tend to fit better with better teammates, but fit is complex and there can definitely still be ways to maximize fit on great teams with neutral or negative portability best players. If your best player has neutral or negative offensive portability, you might still produce dominant team results and minimize diminishing returns if you e.g. (1) focus on adding defensive costars, (2) be especially cognizant of offensive fit when upgrading your offensive talent, or (3) de-synchronize your stars’ minutes so both can play more of a floor-raising role.
Nevertheless, it seems portability is a real, statistically significant feature in team results, impact metrics, and film, and dominant ceiling-raising team performances benefit from having a best player with better portability.
