I used to say that the dunk contest was dead. The days when the NBA’s elite players competed were long gone, I said, along with lots of basketball fans. Most of the modern dunks are rehashes of old ones. The event has plateaued.


That all changed in 2015, when Zach LaVine made me care again, and in 2016 when Lavine and Aaron Gordon turned in the best performance since the 1988 Jordan-Wilkins duel.

Entering 2017, we actually care: Aaron Gordon returns to the event he almost won a year ago, and while Gordon has to be the favorite, Derrick Jones Jr. has emerged as something of a dark horse, at least around these parts. The other contestants—All-Star DeAndre Jordan and Glenn Robinson III—are, meanwhile, certainly capable of winning the whole thing.


While we might not think of the dunk contest as the type of event that would benefit from an analytical lens, it provides a fun example of a situation where data can prime our expectations and teach us a few things, even if it doesn’t give us true predictive power. We have a lot of data on slam dunk champion winners going back to 1984, and it’s not too hard to create a quick and dirty statistical picture of what we might expect a winner to look like.

To do this, we collected a subset of data on all 31 slam dunk champions, including that strange 2014 season when John Wall was awarded “Dunker of the Night.”

We then asked a bunch of questions:


  1. What does the average slam dunk champion look like?
  2. Have slam dunk champions changed dramatically in any key characteristics over the years?
  3. Which of the 2017 participants looks the most like the average slam dunk champion?

First, we tried to get a broad picture of what the average winner looks like, in terms of both physical and basketball attributes. For the former, we used height and age. As far as we know, basketball players, as human beings, have little direct control over how tall they are, and age is, well, age; nothing we can do about aging, at least for now. If we are trying to get a large-scale picture of how the participants have changed through time, looking at two characteristics that we can’t control is a start.



For basketball attributes, we used player efficiency rating (PER) and win shares per 48 minutes (WS/48), two metrics that are commonly used when discussing the quality of a player. Each have their strengths and weaknesses, so we’ll use both. (Note that we used the PER and WS/48 for that winner’s entire season, even that portion of the season that followed their slam dunk champion performance.)

Thankfully, it is not hard to find data for each of these attributes for all slam dunk champions. Using these available data, we calculated the average attribute values across all slam dunk champions, from 1984 (Larry Nance) to 2016 (Zach Lavine). The results say the following: Mr. Average Slam Dunk Champion is just over 6’5 and about 22.5 years old, and in the year he won the contest, he ended up with a player efficiency rating of 16.88, and a WS/48 of 0.107. This PER qualifies him as a “third offensive option,” according to John Hollinger— an above-average player. This fits with intuition. Slam dunk champions tend to be young guys who play on the wing and are talented but not necessarily stars.

With all of this in hand, we calculated relative ratios for all past champions across the four attributes of interest—height, age, PER and WS/48— relative to the average slam dunk champion. A value of 1.0 means that champion had the same value as the average champion. Values higher than 1.0 communicates that the champion had a higher value than the average champion; values lower than 1.0, a lower value than the average champion.

(Note: Harold Miner in 1995 and Zach Lavine in 2015 had WS/48 values that were negative. For the sake of clarity, this graph’s y-axis stops at 0, and so these two instances have no bar for the WS/48 ratio.)


When we look at these data, several patterns emerge. The basketball attributes are far noisier than the physical attributes, which makes sense—both height and age are (presumably) largely fixed properties with a very finite range, while basketball ability is the product of many more variables and influences, and so individuals vary in those characteristics more than they do the physical attributes.

(Another random thing that jumps out at us: Michael Jordan in 1987 and 1988. Sheesh.)


Next, we addressed the question of whether times have indeed changed. We conducted a decade by decade comparison across these attributes.

Here we see that the Gen Xers among us have a point: The players who competed and won slam dunk contests in the 1980s were far better basketball players on average than those doing so today, at least as dictated by both PER and WS/48. From the 1990s onward, however, the data are mostly indistinguishable.


Today, though, we focus on the 2017 contest: Is there a way to use these data to predict who will win? The answer is “almost surely not,” because we are using a data set that goes back 30+ years. The characteristics of a champion in 2015 may be different than those of a 1985 champion, due to changes in the competition structure, judging, etc. Systems rule everything.

While we shouldn’t feel comfortable making a bold prediction, it might still be useful to speculate on which of the 2017 competitors “looks the part.” To do this, we conduct an analysis of the four 2017 competitors similar to the one we did all of the winners between 1984 and 2016. The only difference is that we will use the PER and WS/48 at the current All-Star break, rather than their PER and WS/48 for that entire season, because the whole 2016-17 season hasn’t happened yet.

We then calculate the ratio of the 2017 slam dunk competitor values to the average slam dunk champion values. (This is simple: divide the competitor values by the average winner values, as we did before.) The results:

Again: A value of 1.0 suggests that the player has the same value as the average winner going back to 1984; values higher than 1.0 mean the player has a higher value, and values lower suggest a lower value.



Here we can see that DeAndre Jordan is a unicorn in several respects. He is both a much better player and a lot older than the average slam dunk champion.
So, which of the remaining players seems to most “look the part?”

If we take the average of the ratio values across the four attributes, we will get a number. The closer the number is to 1.0, the closer that player is to the average champion. (We’re ignoring things like the standard deviation at the moment, and are taking the simple average.) When we do that, one player ends up with a value of .984: Derrick Jones Jr. That differs from 1.0 (the average slam dunk champion) only very slightly. The 0.016 difference, in fact, is far less than Glenn Robinson III and Aaron Gordon (both have values of about 0.80) and unicorn DeAndre Jordan (1.37).

There is a major problem with this analysis, however: Derrick Jones Jr. hardly plays! Our analysis here can’t account for a limitation of the PER and WS/48 metrics because they don’t intrinsically account for minutes played. As of today, Jones Jr. has played a total of six games, averaging 3.3 minutes per game. And so the PER and WS/48 minutes metric that I used is embarrassingly misleading, and tells us nothing about how similar Derrick Jones Jr. is to other players with related advanced-metrics scores.


This fragility in PER and WS/48 pops up in other areas of this investigation. Go back to the first graph: It does a beautiful job capturing Michael Jordan’s greatness in 1987 and 1988, and Vince Carter’s almost greatness in 2000. But what about Jeremy Evans in 2012? He boasts some of the higher PER and WS/48 scores of any recent slam dunk champion (19.6 and 0.193, respectively). The problem is that Evans only played 29 games in 2011-12, averaging 7.5 minutes per game.

That doesn’t mean that the entire analysis here is wrong. Perhaps Derrick Jones Jr. will win the 2017 contest; and perhaps he does harbor characteristics common to most slam dunk champions. Maybe the precise characteristic he shares with them is something we haven’t yet diagnosed. Maybe he’ll lose! The possibilities here are why data speculating is fun: Sometimes it provides definitive answers, but it almost always leaves us with better questions.

My own view is that the outcome of the 2017 slam dunk contest is really about one factor: Aaron Gordon. After all, he sorta already won this thing. And come Saturday, the competition will almost surely come down to the answer to a compound question: Did Gordon empty his cabinet in last year’s dazzling display, or does he have anything left in his quiver? No current algorithm can answer that.

All data in this study can be found at https://www.wikipedia.org or http://www.basketball-reference.com. Cheekay Brandon is a “Data shrinker” and computational biologist: @bigdata_kane.