How Esports Stats are Tracked | Stats and Stories at JSM / by Stats Stories

BrianMacdonaldHeadShot.jpg

Brian McDonald is currently the Director of Sports Analytics in the Stats & Information Group at ESPN. He was previously the Director of Hockey Analytics with the Florida Panthers Hockey Club, an Associate Professor in the Department of Mathematical Sciences at West Point, an Adjunct Professor in the Department of Management Science at the University of Miami, and an Adjunct Professor in Sports Analytics in the College of Business at Florida Atlantic University. He received a Bachelor of Science in Electrical Engineering from Lafayette College, Easton, PA, and a Master of Arts and a Ph.D. in Mathematics from Johns Hopkins University, Baltimore, MD.

+ Full Transcript

Tarran: Hi, it’s Brian Tarran from Significance Magazine and I’m here at JSM 2019 in Denver, Colorado with Brian McDonald, Director of Sports Analytics at ESPN, hi Brian?

McDonald: Hi, how are you?

Tarran: Very good, thank you. Welcome to JSM and today we’re going to be talking about e-sports data, which is something that I’m sort of interested in from a personal perspective because I play a lot of games, but not competitively because I’m pretty rubbish at them. So, tell me, e-sports data, how long have you been involved and interested in this area?

McDonald: I've only been working with e-sports data for about a year or so. I just kind of got interested in it because of the booming popularity of e-sports. I’ve done a lot of sports analytics work, so I’ve worked with sports data a lot. E-sports seemed very similar to that just because there’s a lot of team games- like five-on-five type team games, so that has some similarities with hockey and basketball, it seemed. But it also seemed different enough so that it wouldn’t be exactly the same as those. So, I thought it would be interesting to start studying those things, and with a quick search of the web, I didn’t really see a lot of the analysis that has been done in sports. I haven’t really seen it done in e-sports.

Tarran: That’s quite interesting because you’d imagine it being all based on computers, that you’d have access to the data and that would almost be built into the sport from the ground up.

McDonald: Yeah, I think the data- definitely, there are benefits of the data. In theory, it should be pretty much perfect data, in theory. So that’s one of the benefits. But it seemed like a lot of time was spent on other things other than analyzing that data. Some folks spend time creating AI robots to play the game very well, and things like that, but as far as analyzing player performance, which is one of the things that is done a lot in sports analytics, it didn’t seem like there was a whole lot done in that area.

Tarran: So can you- I mean you spoke a bit about being interested in sports analytics before, but can you pinpoint the moment that you thought that this e-sports data might be worth digging into? Is there anything that catalyzed that interest?

McDonald: I don’t know that I can remember the point where that happened. I think it was just I saw some statistics on what the revenue was or what the viewership was for these things, and I also had a couple of people who I work with who would watch Fortnite on Twitch, like during the workday, for example, so that might have helped catalyze it. They were a younger group of colleagues and it was obvious that it was much more interesting to those folks and so e-sports must have a pretty good future, as those folks grow older. I think viewership and revenue will likely grow.

Tarran: So where did you start when you wanted to start digging into this data? Were there ready-made sources of information online that you could download and analyze? Or did you have to approach companies about getting access to their databases?

McDonald: Yeah, there were a couple of sources online that we heard about through a friend, where they have done a lot of the polling of the data from the company that makes the game, and they’ve sort of cleaned it and put it in a pretty useable format. And there’s also an R-package that helped with pulling some of that data so I think we got data from two different sources, the R-package and scraping, like web scraping.

Tarran: What sort of data are we talking about here? What are you dealing with? What sort of data are you looking at?

McDonald: The basic data that we use is basically game-by-game data for players. And it’s- the data just tells you how many kills, assists, deaths. The one game that we’ve been focusing on is Defense of the Ancients 2 (DOTA 2). So they have gold per minute is one of the big things with that game and so it has those things, and then you can tell what team they’re on and whether or not that team won or lost the game, so those are the main data points that we’ve been using.

Tarran: And what are the questions that you’re looking through with this data? What are the areas that you’re focusing on in particular?

McDonald: I think the most important thing that we were looking at was- we wanted to come up with an evaluation of player performance. It’s a common thing done in sports analytics, where you have these team games and each payer has an individual contribution, but looking at the most basic stats won’t tell the whole story because the – what we would call “box-score” type stats- that’s what we call them in regular sports, so just looking at kills, deaths, gold, something like that- it’s not going to tell you the full story because those are highly dependent on the player’s role. They’re also highly dependent on how good the player’s teammates are. And whether or not they lose the game is also highly dependent on who the player’s teammates are. So we wanted to come up with a metric for player performance that accounts for the player’s teammates and their opponents.

Tarran: And what is that metric that you’re using?

McDonald: So it’s a regression-based metric. We’re calling it Adjusted Plus Minus- that’s what it’s been called. For example, I've done it in hockey before with a few other folks. I think it was originally done in basketball, but the idea is the plus minus statistic in basketball or hockey is you know, you get a plus one for every point that’s scored when you’re in the game, and a minus one for every point that the opponent scores when you’re in the game. So that’s what plus minus is. The adjusted plus minus is just referring to the fact that you kind of have a plus minus type statistic but you’re adjusting for the player’s teammates and opponents.

Tarran: The metric that you’ve got, the data that you have, are you starting to see e-sports teams interested in it in the way that baseball, basketball, hockey teams are really investing in data now?

McDonald: Yeah, we hope to get interested in it. We haven’t published it yet, so we’re giving a presentation here at JSM. We have a paper that’s close to being a final draft that we’ll be submitting sometime soon. But hopefully, after that, we’ll see some teams get interested. I think some of the leagues are coming up with their own advanced stats. I think the Overwatch League released something a couple of days ago that was sort of a more advanced player type metric. It didn’t seem to be the same kind of thing as what we’re doing. It didn’t seem to be a plus minus thing, but it seems like an improvement over what’s out there. So leagues are definitely interested in providing these metrics and I imagine teams would be interested as well. So that’s kind of the hope.

Tarran: The U.K.’s case is going to be the same in terms of trying to optimize the strength of the team. Make sure that you’ve got a good number of attacking players and defending players and whatever it might be.

McDonald: Yeah exactly, that kind of thing, and then also eventually, hopefully, we’ll be able to model what sort of players play well with which other players. Or what characters or heroes that the players used mesh well with other ones. And so I think there's a lot of things that people know by experience from playing the game, but maybe there’s something new that we uncover or maybe we quantify by “how much does this actually matter?” We know it matters, but how much does it matter? Does it matter a little or does it matter a whole lot? So those are the kinds of things. The chemistry between players, team chemistry and things like that. Pretty much things that you might ask about a regular sport we ask the same kinds of questions with e-sports.

Tarran: Well, of course, the interesting thing you’ve got with e-sports is that you’ve got the players and then you’ve got their avatars, haven’t you? So you’re kind of dealing with almost two characters, or two individuals because they have to –well that player, the human player performance might vary depending on which character they choose, which role they decide to play?

McDonald: Yeah, exactly it’s – that adds an interesting element to e-sports, that’s not really there in sports. I mean it’s sort of there in that there are some things about an athlete that they can’t really do- by training they can’t change this attribute about themselves. So, height, for example, can’t really change your height. But something like strength you could change. So these avatars, the heroes, they have these attributes and so that’s almost kind of like the physical body that an athlete has. So it’s an area that is one of the most interesting things that we hope to look at is just the role that heroes play. And it’s pretty different from the kinds of things that we would see in sports analytics.

Tarran: Are there other differences between your experience with traditional sports and e-sports?

McDonald: Yes, I think there’s definitely a lot of similarities. I think the big difference is- we kind of just talked about the heroes- I think age might be another big difference. Not really sure. So, typically in sports, it’s much more difficult to project a player’s performance if the player is young. So it’s much more difficult to project a 16-year-old than a 21-year-old, for example, because of the growth- the physical growth and development that they undergo during those years. I guess it’s not totally clear yet whether the fact that e-sports athletes are much younger, it’s not clear yet whether that will make it more difficult to project their future performance, or whether just the nature of e-sports is such that it’s actually easier to project future performance for some reason. Maybe the kinds of skills or physical growth maybe doesn’t matter as much- the physical development. And so maybe players are roughly the same when they’re 16 versus 21. So it’s another unanswered question that would be pretty interesting to try to tackle.

Tarran: And are there other challenges that you foresee on the way to getting e-sports analytics established within the e-sports community? Getting that investment that we see elsewhere?

McDonald: Yeah, I don’t know. I don’t have a good feel for what the hurdles would be. Part of me thinks that you’ll have the same hurdles as you do in sports. Part of me thinks that the community might be more familiar with or more accepting of technological advances or just making use of data, just because of the nature of- you know the folks that worked on the game and created the game you know, a lot of computer scientists, things like that- so, in that sense maybe it would be more quickly accepted than it is in more traditional sports.

Tarran: I guess one of the interesting aspects is that by and large, traditional sports don’t change very often, whereas- when I say don’t change, I mean the rules just stay fixed. E-sports, there’s an expectation that games will be updated, new maps, rebalancing of weapons and character skills and things like that. I guess that adds another layer to consider?

McDonald: Yeah, that’s another interesting part of this. One of the- I should have mentioned it when you asked about the differences, but that’s another one of the big differences that’ll be interesting to look into. But I think the Overwatch League metric that I mentioned before takes into account patches. So different ratings for players, depending on which patch- game patches, and so I think that’s a really- I’m not totally sure the best way- I don’t know right now how we’re going to go about dealing with that. But it’s definitely something that should be dealt with because it does change the game mid-season. I mean, regular sports might have rule changes in the off-season, or enforcement rules- sort of change from regular seasons to the playoffs sometimes, depending on the sport, but there's nothing where in the middle of the sport- in the middle of the regular season, that a rule change that affects the player's abilities-

Tarran: Where your kick becomes less powered or anything like that.

McDonald: Right.

Tarran: So with your role at ESPN, will you be working specifically on e-sports data? Because I know they show a lot of the e-sports competitions now don’t they?

McDonald: Yeah, I think at some point I’ll be working more on e-sports. I think there's more of a focus on football, college and pro football, and college and pro basketball. You know I’ll be doing some hockey there as well. I think eventually I’ll do some e-sports as well. Especially if it’s something that ESPN starts covering more and more just as the popularity of e-sports grows. But for now the most popular things on ESPN are football and basketball, so that’s where a lot of the focus will be.

Tarran: But I can definitely see, from my own experience of playing games and watching these online competitions, you can imagine the data becoming a much richer part of the fan experience than say, it would be- you know you’ve got your football fans and baseball fans that are into statistical analysis in that they follow the numbers, but I think that maybe video gamers is a naturally predisposed to kill ratios, kill-death ratios, that sort of thing, that they’re going to be really hot on that, I think.

McDonald: Yeah, I think so. That was sort of one of the hopes when we started this, that the kinds of things that we’re working on would be adopted and maybe with less hurdles than are in the traditional physical sports. So hopefully yeah, folks are more predisposed to being interested in this sort of thing.

Tarran: Well as they say on TV watch this face. Thank you very much, Brian, it’s great to talk to you today, and I hope you enjoy the rest of the JSM.

McDonald: Great talking with you too, thank you.

Brian Tarran: My name is Brian Tarran and I’m the editor of Significance Magazine. Find us online at significancemagazine.com. For this special JSM series of podcasts we’re collaborating with Stats and Stories. Stats and Stories is a partnership between Miami University’s Departments of Statics, and Media, Journalism and Film and the American Statistical Association. Follow us on Twitter, Apple podcasts or other places where you can find podcasts. If you’d like to share your thoughts on our program send your email to Statsandstories@miamioh.edu, or check us out at Statsandstories.net. And be sure to listen for future editions of Stats and Stories, where we explore the statistics behind the stories and the stories behind the statistics.