Injuries and Rest in Basketball: Can Data Viz Help Find a Balance?
A Lesson on Radial Charts with Dr. James Naismith
It’s 1891. Dr. James Naismith is one of a burgeoning number of progressive physical educators dedicated to improving the lives of America’s urban youth through exercise. Saddled with a particularly rowdy group of young students at the Springfield YMCA where he teaches, Naismith is passing a chilly Massachusetts winter conducting athletic experiments. He’s in search of an indoor game that’s entertaining enough to persuade his kids to do aerobics without trying to kill each other. He has just stumbled upon an old broom closet full of salvaged peach baskets and now suddenly — woo boy — a new fitness craze is sweeping the nation!
It’s 1927. A new generation of basketball players and coaches have spread the game to high schools, colleges, and professional leagues across the country. No longer merely a vigorous gym-class workout, basketball has become a legitimate world-class sport. But — as Naismith himself notes in an editorial he wrote for The Athletic Journal — there are emerging safety concerns with his popular new pastime:
“…there has arisen an opinion that basketball is so strenuous that instead of being a benefit it has become…a menace to the participants.”
Naismith does not share this opinion. He believes that the problem lies with basketball’s critics: they simply aren’t paying close enough attention to his creation. Sure, a game of basketball can look frenetic, but only when you allow yourself to be mesmerized by the ball. Track the movement of an individual player around the court and you’ll find that moments of exertion tend to be interspersed with moments of rest and relaxation. And according to Naismith, this type of interval training can actually be a perfectly healthy routine:
“…prolonged sustained effort and then a prolonged rest is more severe than short periods of activity alternating with short periods of relaxation.”
To demonstrate the activity patterns of 1 hour and 9 minutes of basketball action — and to answer the research question: “Is Basketball Injurious” — Naismith logged periods of motion and rest for one representative player, subdivided into blocks of 3 minutes or less.
Here’s the explanation of the viz from Naismith:
“The black spaces represent the time that the team was in actual motion, the checked spaces representing the periods when there was relaxation but time was not taken out, the clear spaces where time was taken out and the mind as well as the body relaxed, because the game could not be started without warning.”
Naismith uses the pseudo clock face of his radial chart to make one point convincingly: that a player’s periods of activity tend to be broken up by brief respites of relaxation as well as longer game stoppages that allow for recovery.
And thus, with Naismith having defended the safety of his invention to its critics, the sport is allowed to continue its growth.
Bringing Naismith’s Charts to the Modern NBA
It’s 2019. The physical strain of basketball is an issue of national concern once again. “Load Management” are the buzzwords on everyone’s lips. NBA teams are strategically deploying rest days to help nominally-healthy players avoid injury. And it seems to be working! The best player on the league’s latest championship team missed 20+ games for preventative maintenance last year. Meanwhile, NBA commissioner Adam Silver is seriously considering plans to shorten the regular season.
So, now seems like a good time to revisit Naismith’s chart and ask, once again, whether basketball is injurious.
Naismith’s original circular chart had two shortcomings in the modern context. First, he was focused on the ebb and flow of activity over the course of a single game; whereas the current discussion is more concerned with exertion and recovery over the course of an entire season. Second, armed with only a stopwatch, Naismith was unable to evaluate the intensity of the player’s effort during the periods of motion and he had no way of measuring the potential detrimental effects that repeated exertion might cause.
But now, NBA teams are tracking everything their players do. Every step and every dribble of every game is captured from multiple camera angles and translated into an expanding universe of stats. The teams keep most of those data private (with good reason), but some sanitized and redacted version of this information does reach the public via the Speed & Distance page of the NBA statistics site.
We might use nightly distance-traveled stats as surrogates for activity levels, as is already done for soccer players in Europe. One drawback to this approach is that — for a basketball player — distance traveled is highly correlated with the team’s pace of play. That is, there is a sort of obligatory amount of running that is necessary for a player to keep up with the movement of the ball from one end of the court to the other. A game featuring 100 offensive possessions for each team would necessitate a player to take more steps than a game featuring just 60. However, day-to-day fluctuations in the pace of play would be unlikely to be a direct reflection of any individual player’s level of energy or activity and would be more likely to reflect coaching tactics or specific game scenarios.
To glean anything meaningful from these tracking data then, we must first normalize by the number of possessions for which the player was on the court during each game. Next, we will adjust our baseline by accounting for the distance that must be covered by every player on each change of possession — say the distance covered from one team’s 3-point line to the other — by subtracting 56’ from our per-possession metrics. Finally, because defensive movement should (in theory) be dictated more by the principle of checking your opponent’s progress and less by your own energy level, we will focus our activity metric exclusively on offensive movement. In the end, we are left to chart the distance traveled by the player in the front court during each offensive possession.
For our subject, we will consider the Philadelphia 76ers 7-foot All-Star center Joel Embiid. Embiid was the third overall pick of the 2014 NBA draft despite having a broken foot at the time. His injury prohibited him from playing in either the 2014–15 or the 2015–16 seasons and after 31 tantalizingly-good performances in 2016–17 he was sidelined once again, this time with a knee injury. The load-management plan adopted by the Sixers for Embiid is one of the most regimented in the league, as he often misses games for recovery, especially to avoid playing on back-to-back evenings. Philadelphia’s approach with Embiid gives us a unique opportunity to evaluate how a player’s activity level may change over the course of the season in response to periods of strain, rest, and injury. In short, Embiid epitomizes the current concerns about the dangers of basketball and he’s an ideal test case to demonstrate the benefit of load management. Here we examine his 2017–18 season, during which teams were still allowed to explicitly attribute a player’s absence to “DNP-Rest”.
The updated radial chart features a new dimension. Whereas Naismith used black, checked, and white spaces of a consistent height to distinguish periods of motion, relaxation, and rest; I am showing motion using colored bars of varying heights and rest using white space. The three colors indicate the distance covered by Embiid in the front court on an average offensive possession, by game: less than 35’ (bottom tertile, in red), from 35 to 40’ (middle tertile, in yellow), and more than 40’ (top tertile, in green). The bar height reflects the precise distance covered by Embiid during the average possession of each game. To highlight the games Embiid missed for rest or recovery I used black circles.
Embiid’s chart shows a gradual relaxation of the Sixers load management program from the beginning of the season to the end. Through January 2018 you will find ZERO examples of side-by-side colored bars, indicating that Embiid did not play any back-to-back games during that stretch. In the six instances when Philadelphia had back-to-back games scheduled during the first half of the season, Embiid skipped either the first game (once), the second game (three times), or both (twice). However, leading up to the All-Star break in February, Embiid did play in two sets of back-to-back games.
After the All-Star Game, as the Sixers maneuvered for playoff position, Embiid’s rest pattern condensed further. He played 18 games in 32 nights including appearances in four more pairs of back-to-back games. This run was characterized by some sluggish play from Embiid as you can see from his tracking data — 11 red bars, 5 yellow bars, and just 2 green bars. Then, on the 19th game Embiid broke the orbital bone in his face and was forced to miss the remainder of the regular season as well as two playoff games.
A moment of reflection and self-evaluation
We’ve done a faithful job of reviving Naismith’s radial chart by sticking to his original theme — the balance of basketball exertion and rest — and the results look pretty slick. But I’m not totally satisfied with the amount of information that’s being conveyed by the presentation. It’s easy to see that the bars are bunched together in March, but are they closer together than the bars in November? At what point during the season was Embiid’s body most overtaxed? It’s hard to tell.
I tried to clarify things a bit by tacking-on a line chart to emphasize the trends in Embiid’s accumulated workload over the course of the season (as characterized by a 10-game weighted moving average of the amount of rest he had before each game). And with this additional line chart we can confirm that Embiid’s extra DNP days functioned as intended, allowing him to bank some extra rest amid the rigors of the NBA season. Moreover, the chart can pinpoint the moment — after the All-Star break and leading up to his face injury — when Embiid’s accumulated-rest level cratered to a season-low.
But is that good enough? Initially I thought it might be possible to connect the dots between an increased load, decreased activity, and serious injury. Now I’m not so sure. Realistically, Embiid’s eye injury was a fluke occurrence caused by an accidental collision with a teammate. It’s unlikely that his level of fatigue had much impact on how the situation unfolded. Furthermore — if the relationship between accumulated strain and diminished activity level was abundantly evident — we would expect to see gobs of energetic, green games after periods of recovery (at the top of each peak of the line chart) juxtaposed with bunches of lumbering, red games after periods of exertion (in the troughs of the line chart). This is, more or less, what we see for Embiid in March of 2018, but the pattern is not consistent across the rest of the season.
For a more definitive answer to Naismith’s question of whether basketball is injurious, we may yet require even better data. The stats on distance traveled — at least in the form that is available to the public — are probably not sufficient to link rest days to activity level or injury prevention in a rigorous way.
But what might the ideal dataset look like to answer this question? And what might go wrong if we try to create it?
It’s already possible to outfit players with onboard GPS computers that measure changes in direction, pace, and orientation to provide real-time tracking of distances traveled and calculations of the peak speed, acceleration, and force achieved on the court. These in-game performance metrics could be augmented with compulsory pre- and post-game biomechanical measurements designed to quantify the precise amount of strain and recovery experienced by every joint, muscle, and fiber in the body. Taken together with comprehensive biomonitoring of sleep patterns, nutritional intake, heart rate, blood pressure, and biomarkers of stress, we might anticipate injury threats and mitigate risk with load management, prophylactic medications, nanobots, what have you. What could go wrong?
Well, of course, the teams would be generating all the data, so they would own the information, and they would have an incentive to use it to gain a competitive advantage. The result could be that the benefits of personalized basketball medicine — ostensibly pursued for altruistic purposes of injury avoidance — could end up being outweighed by the tilting power dynamics between players and management.
It’s 2027. In the one hundred years since James Naismith first asked whether basketball was injurious we have come a long way in the field of injury prevention. Tensions sparked by the mismanagement of some high-profile injuries brought discussions of health tracking to the fore during the collective bargaining process in 2023. Player reps pushed to democratize the injury-prevention system by putting the players in positions of power. As a result, a player can now choose if and when he wants to wear tracking sensors, undergo biomechanical testing, or provide biological samples for monitoring. Even if the player consents to participate in health tracking for the purposes of injury prevention, he is still entitled to the protection of his privacy. Any health data collected are kept entirely confidential and it is the player’s prerogative to decide who he will allow to access the information — whether the circle of trust is restricted to unaffiliated advisors (his agent, personal trainer, independent doctor, etc.) or extended to include representatives of his team (the performance analyst, medical staff, coaches, etc.). Moreover, players have become empowered to help inform the injury prevention process, by indicating which types of health data warrant tracking on a case-by-case basis. Professional athletes have exceptional body sense and this collaborative, bottom-up health-tracking strategy maximizes injury prevention efforts in a way that would not be possible with a one-size-fits-all solution.
And so it has been that — only by listening to the players themselves — have we achieved Dr. Naismith’s original goal of creating a healthy (and very-much not injurious!) game.
Note: The radial chart shown above was created in R using the ggplot2 library. The code is available upon request.
