A Graphical Analysis of Stephen Curry

WHEN CURRY SLUMPS, SO DO THE WARRIORS

Given what you’ve read about the phenom from Oakland, does it surprise you at all to know that when he has a poor night, the Warriors team suffers tremendously? The Warriors have lost a total of eight times when Curry has played this year. Eight. Of those games, only two have come when Curry is shooting above his average eFG%.

Even further, the Warriors see significant hits to their major offensive metrics when Steph is not in the game.

The Warriors are an historic basketball team, but without Stephen Curry, they couldn’t dream to even sniff 73 wins.

A MAN OF MANY TALENTS

Much has been said (with good reason) about Curry’s other-worldly efficiency, but truthfully, it’s only a small piece of what makes him transcendent. In fact, the man can do it all. Below is a heatmap with a dendogram attached to each axis. With colors ranging from yellow (lowest value) to red (highest value), this heatmap compares the top 25 NBA players with the highest usage rate in terms of advanced metrics. Simply put, the redder the box, the better that player ranks in that metric compared to the others in the graph, and the yellower, the worse. The dendograms along the y-axis depict how similar each player is to the others.

In looking at colors of the heatmap, and the way in which the dendogram hierarchically organized the players, it is pretty clear that the players that rank highest in these categories fall within the area between LeBron James and Chris Paul (especially in terms of analytics favorite metrics such as VORP, WS, PER, and BPM). This grouping includes Stephen Curry.

The heatmap above is interactive, so you could easily zoom in to get a closer look, but in order to get a more defined look at how these top players stack up against each other specifically, take a look at the subsetted comparison below.

From the first heatmap, we see clearly that Steph is better than the high usage offensive players in the NBA, but in looking at this second heatmap, it is clear that even against the best talents in the league, he still outperforms in many categories. In fact, the only observable yellow portions in Curry’s heatmap row are rebounding metrics (which are typically low for guards), turnovers (for which lower is actually better), assists (he is a score-first point guard), free throw rate, and most shocking of all, minutes per game. He has better numbers than most of the league, and he does so in fewer minutes. That’s incredible.

Last year, there was some debate as to whether the league MVP should to go James Harden, LeBron James, Russell Westbrook, or Stephen Curry (who did eventually win it), but this year, there is no question - Stephen Curry is the MVP. These heatmaps only reinforce that point. In fact, some have been making the case that Curry, the league’s reigning MVP from last year, should win both the MVP and Most Improved Player.

At first, I thought it was a bit of a knee-jerk reaction to his remarkable season, and that people might be getting caught up in the excitement of the moment. But then I gave it a closer look…

Those knee-jerk reactionaries might actually have a point.

NO, SERIOUSLY, HE’S GETTING BETTER

I decided to construct a heatmap to compare Stephen Curry against his aging self to see just exactly how much better he has gotten over the years, and the results were actually quite surprising. In this graph, pink represents the highest value, and blue represents the lowest value, with white being the middle-most value.

Remarkably, Stephen Curry really has been getting better across the board, with the exception of some defensive metrics and tendency to shoot 2-Pointers (but given his 3-Point prowess, why would he?). Most Improved Player apologists might have found themselves a new member of their campaign.

And before I got too carried away in my (admittedly over-zealous) Curry fandom, I thought that perhaps this progression is typical of all NBA greats, so I decided to compare the same career trajectories of some of the premier perimeter players in league history. Shockingly still, Curry’s constant upward progression is unparallelled when compared with even the greatest of the greats.

Sure, some guys show marked improvement in some aspects of their game, but across the board progression in nearly every major offensive category? Forget it. The evolution of Stephen Curry has been one of a kind.

I am certainly not using these career trajectory heatmaps as a causal reason to say one player is better than the other, but rather as an interesting comparison between some of the league’s greats throughout history. Think of it as another visual way of saying: “This Curry kid is like nothing we’ve ever seen before.”

HE HAS MY VOTE FOR MVP

(And He Has Twitter’s Too)

One of the many fun capabilities of R is that it has a way to explore Twitter through the RCurl, twitteR, and tm packages to get some sense about how people feel about a certain topic or global issue. You can read about it here. One can use the functions within these packages to mine the text from tweets containing a certain keyword or hashtag, and from there, get a public opinion on certain issues.

Being that I am cheap and didn’t want to pay for the service, the following was done only with the first 85 tweets (that’s all Twitter provides for free), but the results are convincing nonetheless.

I searched Twitter for the first 85 tweets containing “#NBAMVP”, stripped them of any hyperlinks and non alphanumeric characters, and aggregated them into one big text corpus. Then, I took the top 10 likely MVP candidates according to basketball-reference.com and used the stringr package to count how many times each player’s name appeared in that hashtag. The results were as follows, and as you may have expected, Curry was the runaway favorite.

##                 Name #NBAMVP Twitter Mentions
## 1      Stephen Curry                       43
## 2  Russell Westbrook                        8
## 3       Kevin Durant                        0
## 4       Lebron James                        2
## 5      Kawhi Leonard                        2
## 6     Draymond Green                        1
## 7         Chris Paul                        0
## 8       James Harden                        2
## 9         Kyle Lowry                        0
## 10 LaMarcus Aldridge                        0

The fans love him, the author of this article loves him, and certainly, the statistics love him.

As far as the math is concerned, I don’t think the game has seen a player quite as unique as Stephen Curry, and frankly, I don’t think we’ll see one for quite awhile. Take the time to enjoy the Golden State Warriors in the 2016 NBA Playoffs - it’ll be a team that gets talked about in every basketball conversation for years to come.

REFERENCES

Data Sources

Sports Reference LLC. Basketball-Reference.com - Basketball Statistics and History. http://www.basketball-reference.com/. (04-10-2016)

National Basketball Association. NBA.com/Stats. http://stats.nba.com/. (04-18-2016)

R Packages Used

Jeff Gentry (2015). twitteR: R Based Twitter Client. R package version 1.1.9. https://CRAN.R-project.org/package=twitteR

Joe Cheng and Tal Galili (2016). d3heatmap: Interactive Heat Maps Using ‘htmlwidgets’ and ‘D3.js’. R package version 0.6.1.1. https://CRAN.R-project.org/package=d3heatmap

H. Wickham. ggplot2: Elegant Graphics for Data Analysis. Springer-Verlag. New York, 2009.

Hadley Wickham (2015). stringr: Simple, Consistent Wrappers for Common String Operations. R package version 1.0.0. https://CRAN.R-project.org/package=stringr

Ingo Feinerer and Kurt Hornik (2015). tm: Text Mining Package. R package version 0.6-2. https://CRAN.R-project.org/package=tm

Articles Referenced

Assad, Al-Ahmadgaid. “R: Text Mining on Twitter #PrayForMH370 Malaysia Airlines.” R Bloggers. N.p., 21 Mar. 2014. Web. 15 Apr. 2016.

“Dendrogram.” Wikipedia. Wikimedia Foundation, n.d. Web. 18 Apr. 2016.

Maia, Eduardo. “Who is the next Rugby League player to follow Jarryd Hayne to the NFL according to the data?” The Data Game: Data Science and Sports Analytics. N.p., 26 Aug. 2015. Web. 10 Apr. 2016.