This Week’s Chart Dump

Charts I've done but don't know what to do with them

Every week I usually end up with (hopefully) one text, and throughout that week I get a lot of different ideas that I try to write around and compose into a thoughtful post. To build on top of those ideas I create various charts to give an extra dimension. In the end, I’m just not happy or confident enough with the words I write around some charts, so I end up with a lot of charts/images that are sitting on my disk waiting to be used in the future, or (more often) discarded to the shadow realm.

I’ve figured I’m just going to recycle them and try to display them in a short text format.

Throwback to Stephen Curry’s All-Star Game Shooting Flurry

We’ve just had the first NBA vs WNBA shootout during this year’s All-Star Weekend, and even though Steph Curry managed to beat Sabrina Ionescu, I was more impressed by Sabrina’s shooting. Steph managed to win by knocking down 21/27 shots notching him 29 points in total. Sabrina made 18/27 3-pointers scoring 26 points in total (due to money balls) and that was impressive to me as she was shooting from way further away than she usually does (she shot from the NBA 3-point line). She (out)matched some competitors in a 3-point contest with her score.

That contest then reminded me of the 2022 All-Star game when Steph broke the record for most 3-pointers made in the All-Star games, as he went bonkers and shot 16/27 from the 3-point range.

Steph is the best shooter of all time in my opinion, and this performance is maybe more impressive than any 3-point shooting contest (he then scored more than Donovan Mitchell and Malik Beasley at this year’s 3-point contest for example - and probably many others in the past…).

The All-Star game is relaxed and fun of course, but the shots that Steph made in that game were impressive. Shooting on the move, scoring on step-backs, and when things started to heat up in the later stages of the games he even managed to get a bucket when his shots were contested.

Tatum reached 10 triples last year en route to his All-Star MVP title, we shall see tonight if someone manages to come near this ridiculous record.

Wemby’s Shot Selection

The idea for this chart came after watching several podcasts/videos about Victor Wembanyama from people who are way smarter than me regarding basketball. The one that made the biggest impact was Ben Taylor’s video on his channel Thinking Basketball.

The chart I’ve made basically confirms what Ben said about Wemby’s scoring in the video.

Victor’s game started to normalize after roughly twenty games (so in mid-December already) but ever since Tre Jones entered the starting lineup his efficiency, impact and Spurs’ records exploded. Exploded is maybe too strong of a word for a team that has 11-44 at the moment, but 6 of those wins came after Jones entered the starting lineup. The Spurs still look confusing overall, and the most confusing fact is that Wemby cracked 30 minutes (barely) just twice in this same period.

While the Spurs can’t reach either playoffs or play-in (mathematically they can, but let’s be real), ingraining a winning culture can only help boost their young team for the upcoming season(s). And they should start thinking about it very seriously.

Caruso’s Defensive Dominance

Alex Caruso is one of the players that every coach would want on his team. Given his current contract, I think GMs wouldn’t mind fulfilling that wish since he is on a very team-friendly contract, but the Bulls weren’t considering trading him unless an offer they couldn’t refuse came along at the latest trade deadline.

And that makes complete sense, given how Caruso transforms their defense from a below-average to an elite one when he is on the court.

To visualize this, I’ve created a chart that shows the top 150 players in minutes per game, and highlights (with their profile shots) the top 20 players in the defensive impact that they bring on defense.

To gauge an impact I’ve calculated a very simple “metric” for those players:

Def. Rating Diff = Def. Rating ON - Def. Rating OFF

Caruso comes out as 2nd overall in this metric standing at the difference of -8.6 pts per 100 possessions.

The thing (issue) with this sort of measuring impact is that it doesn’t reflect an impact for a single player that well. You can see that Kawhi (who is 1st in defensive rating difference) is among the top 20, but so are Zubac and George, and that makes sense since they share the court a lot (2nd most common Clippers’ combination after Kawhi-PG-Harden).

But the impressive thing is that Caruso is completely isolated when it comes to the Bulls' teammates. The only other player that has a negative defensive rating difference and more than 10 minutes per game is Jevon Carter, coming in at -5.9 pts/100 poss, however, Caruso and Carter shared the court for only 239/1274 minutes that Caruso played.

What I’m trying to say is that Caruso’s presence on the court is impressive and mind-blowing, and he is indeed one of the best defensive guards in the league poised to make 2nd appearance on the All-Defensive team.

The Other Deterrence Chart (KDE form)

This is directly connected to my latest post about Rudy Gobert’s deterrence. I mentioned this type of chart there, so I figured I’d display it here so it doesn’t become forgotten.

Let’s first break down what is Kernel Density Estimation. KDE is a statistical tool that estimates probability density function, so the final result can be a smooth data curve. In the 2d world, it’s not a curve but rather more of a heatmap. In other words, instead of plotting binned data (basically - a 2d histogram) we can plot smooth data that is visually more appealing. To create an on vs off chart, I’ve created the kernel estimation for when Gobert was ON the floor and one for when he was OFF the floor and simply subtracted the ON-OFF estimations. The result is the chart below.

If the area is blue, opponents take fewer shots, if the area is red, opponents take more shots. The chart is generally correct, however, KDE generally has issues with estimation at the boundaries of the data. Our data is very sparse, as the grid (basketball court) is very large and the locations aren’t filled - especially the areas in the mid-range. So that amplifies those issues. There are almost no shots behind the backboard, but the KDE creates a very visible estimation which isn’t really correct.

Another issue is that this isn’t as interpretable as a regular histogram, and once again, especially when the data is sparse.

When you compare the KDE chart with the original raw chart from the text you can see that it overall fits very well and the smoothing makes sense. But if we want to be entirely correct, which is my goal, the histogram-binned chart is the one to use.

If someone has any insight or additional comments regarding this topic, feel free to hit me up in the comments, or through mail/twitter/other social media.

---

This will be it for the very first edition of This Week’s Chart Dump. I want to make this a habit and try to share everything I’ve either posted somewhere else, or just created but didn’t tweet/write about.