The NBA is certainly making a habit of running the little green playbook
of the NFL – turning an already absurdly long season into a year-round show, or more appropriately to some
, a year-round work
. One season ends and immediately the next begins; the echoes of Sunday night's dramatic, season-ending climax are quickly engulfed by the draft theater. It should come as no surprise the NBA Draft was held last night – the same day of the week the NFL's draft bonanza begins. With the draft now in the books, the 2016-2017 season is officially underway. Therefore, instead of recapping the Finals – I think this picture
sums it up – this week we will explore historical results of the draft to calibrate expectations for those still and soon to be on rookie contracts this season.
One powerful tool in measuring a draft is win shares
. Inspired by the legendary sabermetrician Bill James and drawing upon techniques from our previously heralded visionary, Dean Oliver, win shares (WS) attempts to divide team wins amongst the individuals. In doing so, the combined individual win shares of the players on a team very closely approximate the team's actual wins. For context, the highest single-season win share in NBA history belongs to Kareem Abdul-Jabbar with 25.37 WS in 1971-72. Kareem also holds the all-time career record at 273.41 WS. At a fundamental level, win shares measured over a career encapsulate the overall impact of a player on his teams. Statistically armed, we can take a look at the expected career win shares for each draft slot during the lottery era (since 1985). I exclude active players from the analysis because their careers are unfinished. The results are pictured below:
As you can see, the number one overall pick can be expected to generate around 70 WS in his career – the equivalent of TNT's Brent Barry, a number 15 pick in his own right. There are galaxies between that number and the highest number one pick in the data set – Shaq's 181 WS. Therefore, the average does not tell the whole story. The color code shows that the number one pick also carries with it the highest standard deviation, a measure of risk. Treating a pick like a financial asset, teams are compensated for the risk assumed in the number one pick with a higher expected yield. Examining the pattern of the graph, the returns for each pick begin to drop rapidly – logarithmically if you want to sound smart. The second pick returns just 80% of the value of the first pick, while the tenth pick returns 50% of number one overall value, a steep decline. As you go further down the draft order, the expected WS begin to level off – some mid-second round picks equate late first rounders – while the standard deviation shrinks. This fits with the conventional wisdom of drafting for potential early in the draft and taking more polished, limited upside players later in the draft.
One of the most intriguing things about the draft is the analysis of the prospects. While the league and the media have jumped aboard the advanced analytics train, draft prospects are still often presented with their basic stat lines from college or overseas – points, rebounds, assists, etc. Just as with baseball, ridding the public conscious of its entrenched understanding of statistics in the game takes time if even possible or appropriate. Therefore, the goal of modern analytics should be to complement and build upon what fans already know and understand about the game. In this light I wanted to find the prototypical stat line for a particular player based on draft slot. As an added wrinkle, I drilled down into the data based upon position – guard, forward, and center as defined by basketball-reference.com
. At each draft slot, the varying trajectories of each position are worth exploring. However, the number of players drafted at each slot at each position is too small to have meaningful results so I combined the picks into groups of five. Below you can see the statistical line – points/game, assists/game, rebounds/game, and career WS – of a player at each position in each draft slot group (I only displayed groups up to pick 40 for the sake of space and because there is not much change in the later pick groups).
Just as in the graph, we can see a dramatic drop-off after the top five. On average, top five picks are around twice as valuable as the subsequent five picks in terms of career WS. It is interesting to note that the forward position (F) has a significantly lower average WS than the other two – approximately 2/3 their value. Part of this is certainly due to the nature of having to pigeonhole dudes into only three categories. Forward is such a general position whose definition has changed over time, but nevertheless it is interesting to see such a large disparity, especially since it largely falls in line with the average in bins outside the top five. Thanks, or rather shame, should be given to some rather catastrophic busts including Adam Morrison (#3, -0.02WS), Darko Milicic (#2, 7.1WS), and Kwame Brown (#1, 20.8WS).
There is also a pattern that all three positions follow – a dip and then immediate increase in the numbers in the following group. For forwards and guards, the dip is in the 16-20 pick bin and then you can see a slight increase in the 21-25 pick range. For centers, the dip occurs in the subsequent group. Intuitively, we can wrap our heads around this if we consider players getting drafted in the middle to the end of the first round are probably role players – perhaps fringe starters. In order to best maximize the talents of these players, they need to be surrounded by superior talent, which means getting drafted by an already talented team in the late first round as opposed to a mediocre team in the middle, who likely doesn't/won't have the talent to allow that player to flourish. This is part of the motivation behind “tanking.” If you are not a title contender but not bad enough to get a top five pick, then the likelihood of drafting an impact player in the middle is quite low and not proportionately as valuable to later picks as the top picks are to yours.
To put some of these numbers in perspective, the career profile of a top-five guard resembles that of Penny Hardaway
. He is a perfect example of the inherent risk of drafting any prospect in the form of injury. A potential Hall of Fame career was thwarted by knee troubles, which left his win share totals lower than expected or hoped for given his talent. But that is part of the game and a reminder that for all the historically great players drafted, particularly in the top five, there are an equal amount, if not, more examples of those who either busted entirely or suffered at the end of a trainer's table.
The most analytical amongst you may object to the inclusion of per-game statistics, as they do not account for the difference in minutes played or the pace of play. I would respond saying much of my analysis here is arbitrary and ripe for objection. Using average over median, grouping picks by fives, or classifying players into three positions are all examples of choices I've made to paint the narrative in a direction I find interesting. I am simply adding awnings and statues to the tall buildings built before me as a continuation of this incredibly elaborate architecture of analysis we as sports fans have constructed to further our enjoyment of the game. We should not seek to end this analysis by imposing boundaries upon it, but chase the infinite horizon by re-imagining the past through an updated lens. Yet, we constantly seek to limit ourselves, especially when it comes to evaluating performance – so much emphasis is placed on the short term. I hope you can take what I've presented here and consider the long-term impacts of the draft selections made last night, in the previous few seasons, and on into the future.