Amen Escaped

Amen Thompson is breaking out using the tools he's got. Plus an interview with Shri Khalpada, creative coder and software engineer for Cleaning the Glass

In the 1956 film A Man Escaped, a French Resistance Fighter named Fontaine is captured and sent to a German military prison during World War II. Over the course of the movie, Fontaine uses scraps of linen, the wire from his bed frame, and a rusty spoon to engineer an elaborate escape. Without any conventional tools, Fontaine breaks out and walks away a free man.

It’s a great movie. Bill Hader loves it. Cate Blanchett loves it. Paul Dano loves it. The Safdie brothers love it. It’s your favorite movie nerd’s favorite movie.

There are some similarities between Fontaine and Amen Thompson, the do-it-all Swiss army knife for the Houston Rockets. Despite lacking the conventional tools thought necessary to succeed in the NBA, Thompson has found a way to break out.

On Monday, against the Boston Celtics, he scored 33 points (a career high) and hit the game winner over Jaylen Brown. If he wasn’t already, he’ll soon be your favorite NBA nerd’s favorite NBA player.

Thompson was already on the shortlist for the league’s most exciting young players and then Jabari Smith Jr. fractured his hand. That created a vacancy in the Rockets’ starting lineup and head coach Ime Udoka tapped Amen Thompson to fill it.

Since naming Thompson a full-time starter, the Rockets have blasted into the top ten in Offensive Rating. In the month of January alone, the Rockets have had the league’s 4th best offense, trailing only the Cleveland Cavaliers, Denver Nuggets, and Memphis Grizzlies.

Thompson can’t shoot for shit and he knows it. He’s made just 24 percent of his threes on a meager 56 attempts this season. That’s about as many threes as Tyrese Maxey has taken in his last five games.

Still, the Rockets are a better offense when Thompson is on the court and that’s because he’s one of the most prolific opportunistic scorers in the NBA. Thompson is one of six players this season that have scored 100 or more points off turnovers, second chances, and fastbreaks.

Watching Thompson create second chance points out of nothing has been one of the great joys of the 2024-25 NBA season. Watch as he sidesteps a would be box out to sky for an offensive rebound and putback. Poor Payton Pritchard never had a chance.

The Rockets are 26th in three-point percentage. So, like Thompson, the Rockets as a team can’t shoot for shit. So how have they managed to be so effective on offense despite their lack of shooting?

It’s because they give themselves more chances to score than anyone else. They’ve paired the league’s best offensive rebounding rate with one of the league’s lowest turnover rates. They’ve discovered that if they can’t shoot better than their opponents then they can try to shoot more than them.

Over on Jared Dubin’s Substack he tracks which teams are giving themselves more extra chances to score than their opponents on a per game basis. I’ve charted the results below to illustrate just how much of an outlier the Rockets are to the rest of the league.

Thompson is a big part of that. He rebounds teammates’ misses (and his own) at rates usually reserved for big men. Meanwhile, his steals numbers are near the top of the leaderboards. So even though Thompson and the rest of the Rockets aren’t great shooters, they’re still generating great offense by giving themsleves more bites at the apple than anyone they face off against.

The big question for the Rockets is what happens when Jabari Smith Jr. comes back from injury. Few teams will strip a player — let alone a high draft pick like Smith Jr. — of his starting status just because of an injury. But if the Rockets want to be the best version of themselves they’ll need to find a way to keep Thompson and Smith Jr. on the court together as much possible.

The table above shows this season’s on/off numbers for Amen Thompson (abbreviated as AT), Jabari Smith Jr. (JS), and Jalen Green (JG). The numbers paint a clear picture: the Rockets are at their best with Thompson and Smith Jr. on the court and Green off.

Smith Jr. is expected to return from injury in a month or so. Meanwhile, the trade deadline is a little more than a week away. All eyes will be on the Rockets’ front office the next few days.

---

A Refreshing Interview with Shri Khalpada, Principal Software Engineer for Cleaning the Glass

There was a time on the internet where people made cool stuff just because it was cool. Not so much these days.

Shri Khalpada is an exception.

I first became familiar with Khalpada’s work through his website, PerThirtySix.com, that he founded with his friend, Rob Moore. The two of them created interactive visualizations on NBA shot data, passing wheels, and hustle leaderboards.

His work on PerThirtySix led to a job with Cleaning the Glass where he now builds data tools for NBA teams. When he’s not busy with his day job, Khalpada is rooting for his hometown Wizards and developing personal projects, like the Communal Plot which asks users to place themselves along two distinct axes that change every day.

I talked to Khalpada about how he explains his job to DC Hillterns, blending creativity and coding, and his biggest data viz pet peeves.

This Q&A has been edited for length and clarity.

---

F5: How'd you get involved with Cleaning the Glass?

SK: This was pretty serendipitous. I have a computer science degree and after five years working tech jobs out of school. I started having some second thoughts about my career path. The work I was doing was interesting and impactful enough, but it felt creatively stifling and unaligned with any of my interests. This was right around when the pandemic hit in 2020. I found myself with a lot of new free time, so I wanted to use it to rekindle my love for coding and creating.

I got together with a friend of mine who's an engineer with particular expertise in data visualization (Rob Moore), and started a website called PerThirtySix. We found a niche in building interactives with basketball data. Rob would create beautiful bespoke visualizations, and I built out a pipeline to have those charts update daily throughout the season. A few months into that, we were pretty proud of the work we were doing, so I reached out to some prominent people in the basketball analytics space to get their thoughts on where we could take PerThirtySix next.

Ben Falk was one of the people I reached out to. We had a great conversation about the work we were doing and where we could take it, and around that time he was also looking to hire for Cleaning The Glass for the first time. A lot of the skills I learned working on PerThirtySix were exactly what he was looking for help with at Cleaning The Glass. After some more conversations and interviews, we both agreed it would be a great fit, and I've been there ever since.

Any other basketball-related projects in the works for PerThirtySix?

Unfortunately not! The existing projects are still up and running, but it doesn't make sense for me to work on new basketball projects while I'm working at Cleaning The Glass. I've had fun building out content around other topics for PerThirtySix though, including football. Most recently, I built an interactive Scorigami visualization for the NFL, which was a lot of fun. That project was fun for learning some of the quirks of football data -- it brought me back to the beginning of the journey doing the same thing for basketball data a few years ago.

I think you’re in the DMV area. I spent a couple of years there and I wouldn't call it a big basketball town unless Georgetown is good. What’s the typical response when you tell people what your day job is?

Definitely not a big basketball town sadly, especially with the buzz around the Commanders lately and Ovechkin nearing the all-time goal record. But I have hopes for the future!

It's fun to have a unique answer to that question, even if I haven't quite figured out the easiest way to explain it. I usually tell people I'm a software engineer working with sports data, and that's either met with confusion ("that's a job?!") or genuine excitement and interest. Another layer of confusion is that my work errs more on the side of building tools rather than doing the actual data analysis, but one metaphor I've had luck with is "if the data analysts for teams are the accountants, I'm trying to build Excel for them."

I don’t know how much you can say about your work building tools for teams, but how is it different from building tools for fans? What are the types of things that teams use that the average fan wouldn't?

At a high level, it's not so different. Ultimately, we're all in the business of analysis—finding ways to process and visualize data to represent the truth in a way that's useful or interesting. In my experience, the key differences are in the technical constraints. Teams usually require a much higher level of detail and precision, which introduces different challenges throughout the stack, from data engineering to communication.

I love the Communal Plot. It reminds me of my favorite visualization of all time from the NYT, which asked readers to guess the relationship between Parents Income and Percent of Children Who Attend College. Why do you think these kinds of interactive charts work so well?

At its simplest, I think they're just fun. Being an active participant with data can be so much more engaging than being a passive observer. Whether that means you're contributing to the data, like in the Communal Plot, or whether the data is gamified in another way, like guessing the shape of the chart in the NYT example you mentioned, it's just a more engaging and involved experience. Analogously, all of the lectures I remember from college at this point are the ones where the professor actively demonstrated a concept, rather than just reciting a slide to a group of sleepy students.

At a deeper level, I'm fascinated by technology as a medium. If we went back in time and could show our phones to anyone at basically any point in history, they would be absolutely blown away. Imagine a stone tablet or papyrus scroll whose contents can instantly transform, access almost all of human knowledge, and be manipulated based on what someone else far away is doing. It's like the Arthur C. Clarke quote, "Any sufficiently advanced technology is indistinguishable from magic."

So in some ways, it's upsetting to me that the web, with all of the potential to leverage this magic in positive ways, seems to be going down this route of stagnation and "enshittification". Communication of information, in particular, is a space where I think we can do much better in terms of innovation and ethics. That's not to say that every visualization needs to be deeply interactive or some type of experimental piece... but I think there's a lot of room for exploration in how we communicate information. The Communal Plot is hopefully a fun exercise in that!

When you're making something, do you usually have a sense of what's going to get a good reaction online?

I've had things I thought would be a hit that went nowhere, and things I thought were throwaways that got a lot of attention. It seems to be some combination of timing, luck, and the algorithm.

Beyond that, I think it's important to create things that you like. One of the best pieces of advice I've heard around this comes from the songwriter, Elliott Smith. There's an interview where he talked about how if we try to create things to maximize engagement but we don't actually care for the work ourselves, we actually run the risk that nobody will like it, not even us. But if we create things we like, then at least we like them, so there's something to it that other people will probably resonate with too. It sounds simple, but that philosophy has stuck with me.

There's a tendency among sharp people I know to be wary of data when it's presented in a pretty way. Have you run into that kind of thinking before?

That's funny -- I've definitely heard that and am probably guilty of that type of skepticism myself. That bias does come from an earnest place though. For example, I'd be more inclined to trust a bank app that hasn't changed its UI in 20 years than a bank app that looks super sleek and modern, because the first one has probably worked reliably for two decades.

I tend to think about design in terms of two axes: aesthetics and usability. There are more, but this is a good place to start. We can probably all think of examples of apps and websites that prioritize form over function, or that are functionally sound but have a rough experience. I'm really impressed by the ones that manage to nail both, which is what I aspire to in my work. Some apps that I think strike a good balance are Linear and Notion.

For the "it looks pretty so I don't trust it" bias, I think it's important to earn trust. There are lots of small ways I try to do that: always citing your data source near the visuals, not overwhelming the user with too many levers and knobs at once, and proving out the value of making it pretty. If we're asking somebody to consume data in a brand new way, it's on us to show them why it's worth it.

A lot of the stuff I see you post falls into the category of "creative coding.” Was there something that inspired you to explore that side of coding in the first place?

My first introduction to creative coding was through a friend of mine that I make music with (playing the guitar is my other lifelong hobby, alongside this nerdy data stuff). His Master's thesis focused on collaborative music-making using technology, and through chatting about that work with him, I learned about tools like Processing.

It wasn’t until relatively recently that I decided to try some creative coding myself. There wasn’t a specific impetus—just an idea I’d been sitting on for a while and being curious about the process.

I've completely fallen in love with it since then. It feels like a direct bridge between my technical side and my creative side, and there are clear throughlines with my other interests, like music and data. A lot of projects I've done are just for the sake of my own learning and enjoyment, but I've also been able to directly apply a lot of what I've learned to the work I do at Cleaning The Glass.

These days, if an idea pops into my head that sounds like "wouldn't it be cool if ___ existed", I try to make it. Sometimes life gets in the way, but that's been the goal.

Do you have any hot takes on data visualization in general that you want to get off your chest? Things that annoy you? Chart types that need to be retired? Feel free to get spicy.

Oh man. This might get into the weeds, but my biggest hot take is that the way we teach data viz is all wrong.

To take a step back, I think data visualization is more of an art than a science. Of course, there are non-negotiables, like "don’t intentionally misrepresent data," but broadly speaking, a good data visualization balances many qualities: precision, aesthetics, emotional impact, accessibility, cultural relevance, and more. The data viz world, in my experience, focuses way too much on just precision.

For example, there are countless debates about pie charts. Most arguments focus on perceptual precision: it’s easier to compare positions (as in a bar chart) than angles (as in a pie chart). Research shows that people can estimate values more accurately with an unlabeled bar chart than a pie chart. Early data viz pioneers used this to declare, "Pie charts should never be used." But that ignores the fact that pie charts are better at other things, like showing that all data adds up to 100% at a glance and showing part-to-whole comparisons.

As an example, imagine you're tasked with coming up with a visualization about a natural disaster. If your goal is to build a dashboard for first responders, you'll probably want to stick with something super precise, like a table or a simple graph. If your goal is to spread awareness on social media, you might give it a more striking design. If your goal is to write an article diving into the human aspect of a disaster, you might use visuals that tell the stories of the people affected. These are all very different goals that would produce very different visualizations.

So instead of this "precision above all" mindset I've seen, I'd prefer a different approach. I'd like to see us think about some of the "primitives" of data visualization (color, space, encoding, etc) and how we can combine these primitives to achieve different goals. The approach that works in the business world might not be what works best in journalism. Of course, I don't think we should throw out the line chart or anything radical like that! I would just think about that as a premade combination of primitives to generally serve the goal of "precision in showing change over time".

Since I think data viz is an art, I try to draw parallels with other art forms, like writing. It just wouldn't make sense if writers used the same best practices for technical writing as they do for poetry.

What's one thing you can't live without during the NBA season?

As someone immersed in the data side, I love content that ties the stats into bigger narratives. I think the F5 does a great job of that.

When I started working at Cleaning The Glass, I realized that my X's and O's knowledge lagged way behind my data skill, so I've really enjoyed Ben Taylor's Thinking Basketball channel on YouTube. Content like that does a good job tying together the stats with what we're actually seeing in games.

And even though he hasn’t done much basketball content lately, I have to mention Jon Bois. I’m not sure there’s a better storyteller out there today.