Is reducing workload an effective way to increase value?
Tim Duncan averaged over 39 minutes per game in his career before the age of 26. Since then, his minutes have steadily decreased, reaching a career low of 31.3 MPG last year. Despite this decrease in minutes, his production per minute is quite comparable before and after age 26. Clearly, much of this reduction in minutes can be largely associated with age and the wear and tear often associated with aging.
Duncan also averaged over 37 MPG in the playoffs last year, so this reduction in minutes also includes some level of preservation. Soon the Spurs will need to weigh their level of wear and tear for not only Duncan, but Parker, Ginobili and even the Spurs young talent as they balance current contributions with long term conservation.
The issue of measuring career preservation is a tricky task to tackle. In the parallel issue of pitch counts and pitcher burnout in baseball, the debate has often been heated. Over 10 years ago, Rany Jazayerli created a system introducing Pitcher Abuse Points (PAP) in which he suggests weights of abuse on pitchers when they throw over 100 pitches in a game (he seems to focus on young pitchers). Many major league teams appear to make decisions roughly in line with these guidelines.
However, individuals such as Steve Treder have expressed doubts regarding the current pitch count system in place. Even Bill James chimes in, saying that Jazayerli and Woolner’s “research is so flawed that it is virtually useless.” Although it seems logical to assume that fatigued pitchers are more likely to encounter injury problems, there is not a consensus whether or not this tendency should be a significant factor considering taking starting pitchers out of a game.
The problem with analyzing the impact of fatigue on career longevity is that the best players tend to play more games, play longer in each game and have longer careers. This means that players who play more games and minutes (or pitches) also tend to have longer careers, so player ability must be considered at the same time as player “abuse”, which can be easier said than done.
The Method – Predicting Longevity
In order to try to account for this problem, I selected players with at least 2,000 minutes from the ages 28 and 29 and predicted Mean Expected Wins Added (MEWA) beyond age 29 (Mean Expected Championships Added was too variable to yield significant results) using regression analysis.
In this projection, I used an estimate of value per minute (basic statistical plus/minus or SPM), an estimate of perceived value and health (MPG) and a combination of the both (MEWA/gm). The formula I used to predict MEWA after 29 was 85*MEWA/GP + 0.72*MPG + 3.05*SPM +0.95*Max(0,SPM)^2; not below 0.
MEWA, MPG and SPM were cumulative for age 28 and 29. I summarized the data, excluded all players who were expected to have fewer than 10 MEWA after 29 and included only players who turned 30 after 1980 and “retired” before the 2010 season.
Whose Longevity Defied Expectation?
The following table lists MP during ages 28 and 29, predicted Mean Expected Wins Added after 29 (based on performances at 28 and 29), actual Mean Expected Wins Added after 29 and the respective ratio of [Actual Wins/[Expected Wins] entitled “Longevity Ratio”. Games played before 28 and MPG before 28 are also included.
NBA MEWA vs Expected after 29
|Nick Van Exel||4752||21||28||132%||378||34.8|
The upper part of this list is littered with names that came into the NBA with little expectations and went on to have long, successful careers. Perhaps a late start tends to produce a later than expected finish. Could this be driven by matured players who enter the NBA with fundamentals that last late in their NBA lifetime?
The bottom of the list is full of players with injury-riddled careers. Note that I excluded players with fewer than 10 actual MEWA after 29. Many such players either returned to Europe before their skills or health greatly diminished. Others saw their careers end under tragic non-basketball related circumstances. I accounted for neither of these factors explicitly in my projection of MEWA after 29 and foreign players could especially skew the results. Increasing the minimum actual MEWA effectively eliminated these types of individuals.
The following is a graph displaying the averages of MP during ages 28 and 29 and Games Played and MPG before 28 for the top and bottom halves of the initial list (labeled “exceed” and “burnout”):
Notice that although average minutes during ages 28 and 29 are higher for those who exceed expectations, cumulative career MPG AND GP up to age 27 are actually LOWER by a statistically significant margin. In fact, I estimate that saving a seasonâ€™s worth of games before 28 corresponds with an increase of about 12% career effectiveness after 29 (using regression analysis weighted by expected MEWA). Similarly, saving 5 MPG before 28 typically results in an increase in career effectiveness of roughly 6% after 29.
However, considering both of these variables simultaneously reduces these variables to about 10% and 2%. To try to put this in perspective, I did a quick estimate and found that the typical 82 games “saved” before age 28 tend to add, over the course of their career, about 1/4 of the wins contributed by that player for an 82 game season.
Clearly, resting the second of back-to back road games or games when nursing an injury would be expected to have a more significant impact on career longevity. Most games in which such a player is not required to rest, but considered for rest by the team or player, fit these circumstances. Therefore, the above estimates really suggest that a season’s worth of borderline, “Should I play or should I rest?” games before age 28 tend to improve career effectiveness by 12% for the age of 30 and beyond.
Additionally, one might suspect that as players age, the 12% improvement in career effectiveness increases; but since the time available to make up the lost player contributions for a season are lessened, these players might “only” be expected to add 1/4 or less of one season’s wins contributed over the course of their career. (See the end of last week’s post for a basic expectation of the Spurs player’s future contributions.)
The aforementioned decrease of the impact of MPG when combined with GP in the regression model indicates the correlation of MPG and GP. The fact that GP is less affected suggests to me that MPG might be largely reduced because of injury, when minutes are most significantly reduced in line with a reduction in games played. Maybe players who give injuries more time to recover instead of pushing through last longer. Perhaps these players just know how to take care of their bodies. The tendency for players recovering from injury to perform at lower levels adds to the list of reasons to not rush a return.
From a coaching or organizational perspective, much else must be considered in addition to games and minutes played for each individual. Just to name a few, factors such as how a player is feeling, team need, medical advise all play significant roles. Unfortunately the available data doesn’t seem good enough to shed any more light on potential factors causing the results in my study.
Reducing Player Minutes
You might be thinking “Ok, great. So now there is support for the widely accepted theory that saving players prolongs careers, now what? Sitting the best players hurts the team for the current game and season.” There is at least one more important factor must be considered. Not all minutes are created equal.
Clearly, playing Duncan at the end of a 30-point blowout does more harm than good, so in order to maximize his impact given his “allotted minutes” (not that there should be a hard number), he should play in as many high leverage situations that fit his skill set as possible. My article on Clutch Adjusted Plus/Minus displays the comeback probabilities given the deficit and time remaining (see Comeback Probability Table).
Weighing minutes played by the impact of 3 points on comeback probability, we do find that the best players typically play a higher percentage of the teamâ€™s minutes. The following chart shows the differences between the Spurs player’s percentages of team minutes and clutch-weighted minutes:
And finally, here is a table of some select players:
Notice how the better and/or older often are at the top of the [clutch-weighted minutes â€“ actual minutes] list. It is also interesting to note that two of the most historically statistically-oriented teams are in opposite ends of the spectrum. The Mavs are arguably the team whose best playerâ€™s clutch-weighted minutes increase the most. Meanwhile, the Rockets best player’s clutch-weighted minutes seem to increase the least (or even decrease).
This could be somehow related to Daryl Moreyâ€™s comments on last yearâ€™s Sloan Sports Conference that the Rockets donâ€™t make decisions accounting for clutch performance, to which Mark Cuban chimed in saying that his team has accounted for this factor in the past. Or it could just be because the Rockets have more role players who donâ€™t always fit in â€œclutchâ€ situations while the Mavs have older players on a more top-heavy team.
In summary, the Spurs best players can be effectively utilized over the long haul if Coach Popovich adequately focuses on resting key players during injury-related times, typically coinciding with events that have a low probability of impacting the Spurs’ 2011 results. I’m sure Coach Pop would not be surprised at all by this by this effect, but attaching some magnitude to this result, even a vague magnitude, can help with adjusting or assuring instinctive decision making. With the future in mind, perhaps Duncan, Ginobili and Parker shouldnâ€™t play elite minutes for the Spurs this year, but still play elite â€œeffective minutesâ€ if healthy.