Data Science Hangout | Tegan Bunsu Ashby, Brooklyn Nets | Showing the Difference You're Making

Hi, everybody. Welcome to the Data Science Hangout. If you're joining for the first time, I'm Rachel. It's great to meet you. If this is your first Hangout, this is an open space for the whole data science community to connect and chat about data science leadership, questions you're facing, and what's going on in the world of data science.

So we always want to create spaces where everybody can participate and we can hear from everyone. There's three ways that you can ask questions. You can jump in live and raise your hand on Zoom, seems to work well. You could put questions in the Zoom chat, and feel free to just put a little star next to it if you want me to read it, or I could call on you to introduce yourself and add some context there too. We also have a Slido link that Tyler will share in just a second here, where you can ask questions anonymously too. She's ahead of me, I see it. I just like to reiterate that we love to hear from everyone, no matter your level of experience or area of work as well.

But today, I'm so happy to be joined by my co-host for today, Tegan Bunsu Ashby. Tegan is a senior software developer at the Brooklyn Nets and also co-founder of Women in Sports Data. And Tegan, I'd love to start off by maybe having you introduce yourself and share a bit about your work. Oh, well, we're so secretive, like there's the joke about how the Brooklyn Nets are kind of like the Kremlin, but you know, second rate. But at a high level, my job is to help win basketball games. And I do that by translating data into applications and insights and recommendations for how everybody within the Brooklyn Nets makes basketball related decisions.

So in front office, so how we evaluate players for draft acquisition or trade to the coaching staff of how do we best strategize how to defend the pick and roll.

Can I ask you what's something that you're really excited about in regards to data lately? Like data in basketball or data in general? Yeah, data science in general or data in basketball, whatever is most exciting to you.

I'm really excited about computer vision and in like broadcast technology. I don't know if that's so I'm a software engineer and so really more data science adjacent. But yeah, something that we really struggle with is getting rich enough data sets on college players or basically players who aren't on an NBA court because one of the primary data sets that we use comes from cameras in every NBA stadium that tracks the player location is tied with events, but we don't have that for our draft and something that would be really, really useful would be able to have that from television. But it's such a hard problem because the frames are jumpy. Sometimes like CV confuses like players on the on the wrong team, things like that. But having that would be immensely, immensely exciting.

Women in Sports Data

Cool. As we're waiting for everyone to jump in with their questions too, I'd love to hear a little bit more about women in sports data and how you started that too.

Sure. So women in sports data is kind of this groundswell event that a couple of my fellow amazing women in sports data from Houston Astro, Sarah Gallas, and Diana Ma, who was a data scientist at the Los Angeles Lakers, we just really felt that we needed a space where we could kind of showcase our abilities and have a platform to talk in a real technical manner about how data is applied to sport and just to have that opportunity. So I'm really fortunate that the Brooklyn Nets, you know, I came to RGM with this idea of like, we really don't have a space for this. This doesn't exist. It's kind of a unique environment to work with in data. How would you feel if the Brooklyn Nets hosted it? And he was all about it and just kind of started running from there.

That's great. I should probably like describe what the event is. Yeah, I'm a little bit in playoff fog, I have to admit. For those that aren't aware, I'm mourning my second loss in a row. And I'm a little, this is super off topic, but I'm a little paranoid because I used to work for the Philadelphia 76ers. And the second year that I was in the playoffs with the Sixers, we got swept by the Celtics. And so this is my second playoffs with the Nets. And get a little antsy, but different team, different parameters.

Anyway, so the Women in Sports Data event is going to take place in Brooklyn in late summer. We're going to announce our official date in 10 days. And it'll be this event with panels and talks by women working in sports across the technical spectrum. So not only analysts and data scientists, but software engineers like me who partner with our data team and turn those insights into applications for our vested stakeholders. Because it's such a unique environment where you have a small coaching staff, a small front office, you really want to get them to interact with the data in a much more immediate way than you might in a traditional organization where you have different layers to step through.

And so concurrent to our one day event in Brooklyn, we're going to be hosting a hackathon using what is looking like event level player location data from soccer sponsor to be announced. And that will take place over about a six to eight week sprint before over the summer. And we really wanted to have something that was supported with mentors for women and non-binary people interested in working with sports data to get an idea of what is going to be most impactful. And how do I look at this data? And even if I don't have sports domain knowledge, I have these technical skills. So how do I apply that to what is genuinely very interesting data at an abstract level?

Working with data scientists and the team

And I know there's a lot of data scientists on the call as well. So I think it would also be great for us to hear how you work together with data scientists on your team as well. Productionizing models is a large part of my job. And I think traditionally my favorite part of my job. But over the past couple of years, I've really been getting into like the very front of the stack. So I've been doing a lot of data visualization. And so what sometimes will happen is that a data scientist will build a visualization in Matplotlib or ggplot and doesn't necessarily want to productionize it in a tiny dashboard. We need it to be a little bit more robust. It needs to talk with a different database. So I take that and I turn it into D3 code. And it lives on our Planets application. And it's just a lot more accessible and intuitive for non data fluent users.

I'm curious what that transition looks like or that handoff from the different teams or how you work together. So I think everyone is very surprised by how small our teams are. But we're six, we're six people. And we have we're about 5050. We're three data scientists and three developers.

And really, I wish I could say that it was very formalized. But because we operate more on the basketball calendar than a development calendar, like I have project manager friends who probably scream at our process. But often it's kind of like, oh, like, so and so asked me to do this, or this is our development plan. Do you have time in your schedule to turn it into something a bit more robust?

I realized that wasn't very nice of us to do this in the middle of playoffs for you. But thank you for joining us. Maybe it'll turn turn the corner.

Building community and navigating secrecy in sports

Yeah. So, when I started in sports, I was pretty much explicitly told, like, don't talk to people on other teams. Like, don't disclose what you're working on. Every single thing that you do is a competitive advantage. And on the technical side, like, that's somewhat true because different teams have different investments and different capacities for investments and how they're able to use data. So, for example, when I was with the Sixers, I was also on a very large team, but there are teams in the NBA that might have maybe only one or two people that are basically data unicorns, and I'm sure very, very tired. So, my experience in sports has only been to specialize, but I didn't really know the landscape then.

And as I got further in my career, I noticed that one of the advantages that my male co-workers, both in research and development as well as basketball operations, had was they knew people on other teams and once COVID hit, you know, I kind of realized, oh, my God, like, I'm all alone. And I was very fortunate in the Sixers that I worked with an amazing data scientist, Savannah Serrett. She's now with Overtime Elite running their research. And I just kind of felt like I really wouldn't want anyone to feel like an island, to feel like they were alone. And because women, especially on sports teams, and with a data background, like, already tech is so male-dominated, and then to intersect that with sport, it's hard, right? Like, you have such unique experiences.

So about a year ago, I started a Slack community for women working in basketball, specifically, to just kind of get to know each other. And it just blossomed into this amazing community, so supportive. Like, we've been on Slack all today. But yeah, it was kind of that, and that first seed, and then Diana, one of my co-organizers, and I got to talking of, like, how can we grow this? Why didn't we have this before? How can we make sure that, you know, the next woman that works for a team and doesn't have, like, not only another woman in her department, but another woman in the entirety of basketball?

And it was also something that I really felt compelled to do when I joined the Nets. I'm the only woman in the front office here, and I'm one of the only women in a front office in the entirety of the NBA. And to me, like, that's literally insane, because data and software skills, like, that should be one of the more equitable ways to enter into sports. And it's such a high visibility, high cultural impact role. I mean, like, you look at my former boss, Daryl Morey, or Kweisi Odofomensa, who's the new GM of the Minnesota Vikings, and they both have technical roles. And this is kind of a pathway to build the resume and the experience of being in sports and getting to the very, very top of leadership.

like, that's literally insane, because data and software skills, like, that should be one of the more equitable ways to enter into sports.

Open data and hackathons

Yeah. So, I'm more of an NFL fan. I know there's the big data, as I'm sure you've heard of, which is the NFL's open source data community, where they give out, like, GPS data, helps, like, give access to more people to show off their skills. Is there stuff in, like, in the NBA? Because I'm presuming that's one of the big things in helping show off more diversity and allow more people access to this information. Because, of course, traditionally, it's just, like, whoever's in the teams have access to this data.

Right. And it's very expensive, too. I know, so our main data provider right now is a company called Second Spectrum. And I know that there are some people in the media who have subscriptions, but really, you need to have, it's minimum five to six figures just for access. So, typically, that type of data is not open source. I'm not sure if older data sets from Second Spectrum or the previous company, SportView, are available. And that's really just more from, like, my privilege of having a different bank account responsible for paying for that API.

But the NBA used to hold a data competition pre-COVID, and I really hope they do it again. But this is kind of the goal of the Women in Sports Data Hackathon, is to get different providers to donate that tracking data, or at least event-level player location data. And I'm being very coy about this because we haven't officially signed our contract for that. But, yeah, like, these rich data sets that show, like, where are players, you know, what events are associated with an action, and how do you build a model on top of that? And also, critically, I think, for the NFL Big Data Pool is that there's a very large prize associated with it. So, it's very motivating to dig in. It's freely available on Kaggle. But you also, you know, like, okay, like, I'm doing this work. It's definitely going to be valuable either for a team or a league. How do I make sure I'm not exploited for it?

Prioritizing work and protecting time

So how do you, what are the, what's the identifiers for, you know, quick versus strategic?

It's so tricky, because I feel like working in sports, we have multiple full-time jobs, right? So my full-time role is being a software engineer. And then I have another full-time role of being a front office person. So, you know, I have to understand my domain of, you know, having an opinion on basketball and the league and upcoming games.

So I don't think that this is a solved problem by any means with my team or around the league. But I've been talking to a lot of, like, project managers, like, how, exactly, like, how do I solve this question? Because we don't have enough room to, you know, appoint, you know, like a scrum master or, like, a guard dog, essentially, to say, like, no, like, I protect my engineers' time, like, let them do their thing. And something that's kind of wild right now. So this is not actually my office. My office is in something we call very cutely the war room. I think it's, like, very common in sports to call it the war room. And I have strong opinions on this. But because it's kind of the node of where our organization lives, it's great because, you know, people, there's not a wall between R&D and the rest of basketball operations. But then, like, you'll have 20 guys gathered around and, like, watching film or, like, talking to the scouts. And so it's, like, how do you sit down and say, like, I have, like, this really difficult problem that, like, I have to, like, enter the matrix to solve.

So I've been trying to kind of put on, like, what's my project manager hat. Like, we do, you know, stand-ups on a weekly level and, you know, coordinate all of our processes through GitHub. But as far as, like, determining at a high level what's an emergency versus, like, what's our long-term research prerogative. And then something also that's very important to me is making sure that everybody on our team has room to learn and that that time is protected. You know, that concept of 20% time where, like, I can't really explain or sell to, say, like, my GM. Like, we're going to invest a lot of time on, you know, working on computer vision. It may or may not work. And you're not going to see, like, a direct outcome of that project. But if it does, then it's going to be extremely valuable and it's going to be a significant advantage. But, like, I can't promise that. So I try to protect, like, that 20% time of just, like, this is just, like, learning. This is just personal development. This is experimental.

And, like, yes, it will, you know, projects that we come up with and experiments that we do are kind of geared towards understanding the game of basketball better and to gain a competitive edge but are a little bit more closely aligned with, you know, that technical maintenance. So that's about, like, one day a week. And it's a little bit more flexible during our off-season push draft in the summer before training camp.

And then we kind of have to balance, like, most of my team has been in basketball for many years. So we can kind of anticipate, like, what the cadence of the season, like, what are the kind of questions can we either, like, anticipate and build out, like, what is, like, okay, like, we didn't crash the glass well enough. Like, we know, like, what the question is going to be the next day. And so, like, we have resources of here's the study that we did a year ago on this question. Here we've applied new data to it now. This is what we might recommend or this is what we would look into.

And then, like, the one-off requests, we try to be pretty, like, on top of that depending on who's asking. And I think, like, that comes more from organizational structure and having strong leadership from the top and, like, a lot of buy-in, which I feel very, very fortunate to have.

But also approachability is really important. So I would, like, I wouldn't want our department to feel inaccessible or anybody in our department to feel inaccessible even to like the new grad basketball operations assistant. Frequently we've gotten questions of, like, oh, I'm really interested in what you're doing. I can even be, like, super beginner. Like, how do I, like, I have this question with X domain. Like, how would I start to learn R? Or, like, can you show me how to query the database? Like, what does that look like? And so we try to, like, build tools to empower different levels of users, both from their role but also their comfort with technology and to make sure that there's not this hard line between R&D and basketball.

And then there's something the former director of R&D for the Milwaukee Bucks recently told me. He's like, I don't have a tech job. I have a basketball job. And I thought, oh, I've never really thought. I've thought it more like I have a tech job in a basketball team. And I think that's really more, like, a difference of philosophy of, like, how do you protect that and also how do you make sure that it's a goal, right? Like, the goal in basketball is the same as any other business, which is to succeed. Like, it's just a little bit more literal for us, like, to win.

I really like that idea of protecting 20% of your team's time. And I know that can be so hard sometimes, too. So how do you actually go about doing that in practice? Personally, I try to make sure that I have, like, literally, like, 20% of my time during the week logged on some other, like, research or development initiative that I'm interested in. It's very difficult to do during the season because we play games on weekends. We play games late at night. And it's almost, it feels a little unreasonable to say this is a 9-5 job, but also you need to show up at the arena at 6.30 or go straight from the practice facility to the arena and show up the next morning. So we're pretty relaxed on, like, you're responsible for managing how you want to see your time. And, like, there are only, like, true, like, real emergencies, and they're not going to come from our department.

The future of sports analytics

How do you see sports analytics evolve over the long term? Is it more proprietary advanced algorithms or more proprietary data or something else? Hmm. So there's this book called The MVP Machine, and it's about player development in baseball. And I think that the biggest advances in sports analytics are going to be moving in that direction. And, like, how do you empower your, quote, unquote, end users, end users, your players or even your front office to grow using database insights?

I think that in basketball, because we are relatively small compared to a data team for, say, baseball, which will have, you know, departments in the dozens, like a true startup, that we will probably remain more of, like, the interlocutors and some teams who are interested in, like, building out more models or taking on new experimental technical projects. I think that'll be more of a team prerogative rather than a trend. And, like, that advantage is going to be the same advantage that any team has in filling personnel, right? Like, it's how much money do you have to spend?

But I do think that vendors and, like, the ability to not necessarily deliver modeling and insights to a team, but doing a lot of the processing power that would be expensive for us and taking care of the sort of things that maybe a general manager doesn't, like, can't see the immediate impact. So, like, my example of, like, I would love for my 20% time project to be working on computer vision projects, but is that if there were a vendor that could provide that, that might be a more efficient use of both our time and resources.

Player data, privacy, and consent

Sure. So, the NBA Players Association has very strict regulations for how and what kind of data we can collect. So, the inferences that we make from the data sets that we subscribe to or collect ourselves like that, that is, that's totally fine. But for instance, like you can't put a tracking chip in a ball. You can't have a player wear wearables in-game. They have to give consent to do that and then only within a particular context.

Yeah. We have, like, a very robust and excellent player performance staff that does address, like, more of the, like, physiological and athletic questions, like that data. But, you know, I was saying, like, a lot of this requires player consent and, like, player willingness to opt in because you can't just say, like, well, like, take the example of sleep. Like, sleep is such a huge concern in the NBA because we play 82 regular season games. And so the travel distance, like if you're playing games every other night or back to back and you're flying from New York to San Francisco to San Antonio, that is such a huge strain on your ability to perform at an elite level. How do you monitor, like, okay, like, maybe you're not going to, maybe you need to rest or maybe, like, you shouldn't play at, you know, at an extremely hard level during practice, things like that.

And I think that to previous question's point of, like, what's the next direction in sports analytics? It is this kind of, like, behavioral question of, like, one, how do you get players to opt in? Personally, I think that it has to be, like, it can't be surveillance analytics. It has to be, what do I, like, what am I going to get out of this? And what do I want the team to know? And it becomes kind of tricky because, like, you want to optimize as a player, how am I going to perform, right? But you also need to be very careful of exposing, like, certain data to your team because it can affect where, you know, how much your next contract might be or if they know something about, you know, an injury I have, are they going to try and trade me? Like, how much leverage do I have?

Team size and hiring

I see, Mark, you asked a question as well around managing requests. Would you want to jump in? Yeah, I know you mentioned you have, like, a super or pretty small team, and this is tangentially related to Mike's question about balancing priorities. I guess, when do you, like, what's the cutoff between we have, you know, like, what's the cutoff between we have all these different priorities, and some of them we can push back, and some of them, you know, we'll just communicate that these are lesser versus, hey, we need to accomplish all of this stuff, and we need to either expand the team or, you know, offload something to a different team.

We always need at least one extra person, I think, both on the engineering side and on the data side. But a model, I think, that is maybe more successful for us is we almost have technical translators embedded in certain departments within basketball. So, we have a director of coaching analytics who's the main liaison with the coaching staff, and he sits on the bench at every game and provides in-game analysis, but he also does, you know, modeling specifically related to gameplay. And I think a reasonable question is, like, should that be one role? Should it be, you know, distinct roles? How do you kind of merge a team if it were distinct?

And then, like, what about departments that don't have that embedded technical person? So, like, our cap and strategy group is two people, but when I was in Philadelphia, cap and strategy was technically part of my department. We were officially, when I started, before we converted to research and development, we were strategy and analytics.

Yeah, it's always tough because we're very lean, and it's almost like there are too many hats for for one head, but at the end of the day, like, you're essentially a consulting group embedded with a team, and you have to respond to that more so than, you know, technical prerogatives that you might have as a team and kind of negotiate, like, I think that the payoff, if we invest in this research direction or if we build out this tool, then we'll save both our time as well as our stakeholders' time. So, not a satisfying answer, I think.

Yeah, I don't think that it's solved anywhere, and it's maybe not solved even by more people. My group in Philadelphia, we used to claim that we were 10, but that also included outside consultants who worked in academia and at Google Brain. And, you know, the work that our consultants did certainly dovetailed with the work that we did, but it wasn't, they're kind of insulated from the demands that those of us who are directly embedded with the team and have front office roles or coaching roles would have to support.

Getting into sports analytics

Very, people say that it's very competitive, but I think that it ends up being very lucky. And as I was talking about earlier, it's supposed to be one of the more equitable ways to get into a front office or to work for a team because you're supposed to be evaluated on your skill set and not whether you've played professionally before. I certainly haven't played in the NBA before. I've never played basketball even at a highly competitive level.

I think that there are kind of like some secret tricks that people on the inside know. And everyone usually says that you have to build a portfolio with basketball related projects. You should have kind of an intuitive sense of like what might be useful for communicating data to a coach or or to a front office within a specific domain like salary cap or scouting.

But I personally didn't have a basketball portfolio. I didn't even really think, oh, like I have to work in basketball. So I think that my experience is probably different and maybe unique. It's very difficult to get a role for a front office because they just don't come up that often. And when people are hiring, I think it's kind of like the greater trend in the data science and software engineering industries now where everyone would love a senior developer or a senior data scientist, but nobody wants to hire an entry level person. And then how do you get an entry level job if that job's not advertised? Then it's what's your network? And it's a lot easier to build your network if you're already associated with basketball or you have a lot of time to work on projects in a visible way for free.

And I think that that kind of shadow requirement is specifically disadvantaging women and people of color because we have different responsibilities. We have different demands on our time. And oftentimes we haven't been told that that's one of the blueprints to get a job in sports. And I just think that we as an industry need to be doing a better job of making sure that like that requirement, if it has to be a requirement, that we provide kind of a development pathway that's rewarded either through a hackathon. The NBA used to have one or through women in sports data, which will have cash prizes, but to specifically target and develop those communities to mentor them and tell them specifically, this is what we want in basketball. This is what we want in football.

But I don't know because teams are not going to grow in any significant sense. If you have an R and D department that's larger than the rest of basketball operations, then the people who keep the books are going to be like, well, why don't we just hire an outside organization to do this that sells the same information or does the same thing? They're not going to understand the business case for a large department. So I don't think that positions, it's not going to be a growth industry, but you can use those same skill sets to get a job for a second spectrum or for a Hawkeye or even in like traditional big tech that would love to provide services for sports.

Showing the difference you're making

Hi, everyone. So that's always my constant question, right? So clearly, when you're doing a project, you're always asking yourself, am I making a difference or am I going to measure success, right? I mean, it's evident that more organizations, even here in Jamaica, actually getting more data science departments, hiring more analysts, it makes sense, right? But how do we justify our existence qualitatively and quantitatively? So in a sport like basketball that I like as well, is how do you know that you're making a difference in terms of the bottom line at the end of the day, if you understand what I'm asking?

Totally. I think that it's one of the joys and terrors of being in a data-intensive role for a team is that you can't point to, oh, we did X or we recommended Y, and now we have an NBA championship. But it really has to be, does analytics, does data and investment in data at a league level indicate that you're going to have sustained success?

I think that it's one of the joys and terrors of being in a data-intensive role for a team is that you can't point to, oh, we did X or we recommended Y, and now we have an NBA championship.

And I think that the investment in analytics departments and the success of the teams that have been able to do so is kind of indicative of how teams are evaluating the values that data provide. But I do think that that's really nice to say, but we also need to hold ourselves accountable of like, we're using all this time and resources. Does it contribute to winning? Does it help us make better decisions? And I think that it's almost a qualitative assessment of at certain pressure points, did we create an environment where it was easier to negotiate a trade or pull the trigger on drafting a certain player over another? Or even to the level of preparing for what happens on court, did we create an environment of, all right, we knew these tendencies, we could project this expected value. Were we able to support that in a way that it was actionable and retainable and implemented either at a player level or coaching level?

Being visible and making friends

I wouldn't say that I'm a very active or fluent Twitter user, but I got some advice from my friend, Kathy Evans, who is vice president of research and development for the Washington Wizards. And also my friend, Brittany Donaldson, who is the director of research and my friend, Brittany Donaldson, who's former assistant coach for the Toronto Raptors and data analysts. And both of them are much, much better at social media than I am. But they told me, you know, like, you're one of the only women in a technical role in the NBA, and 50% of NBA teams are hiring for software developers. And like, nobody knows what that looks like, or that you can even do that as a job. And so I thought, oh, well, it just like, through being visible, like, that might help.

Then I obviously would love to do that. And then it just kind of made everything else as far as the community building so much easier, because I could see a response, and I met so many people, and it was easier to develop a connection, because I could reach out to somebody, or they could reach out to me, and we could have something to talk about, whether it's like a thread that they wrote, or an invitation to a community. So yeah, at a practical level, just being visible and making friends.

at a practical level, just being visible and making friends.

Awesome. Well, thank you so much for jumping on and sharing your experience with us. Really appreciate it. And thank you all for all the great questions too. Of course. Yeah, thank you for having me. And yeah, if you have any questions, I'll do my best to answer them. You can always say hi on Twitter. I will never give you an opinion on a player, because I cannot afford the tampering fine, but I'll do my best.