Media Column: Tim “magic” Sevenhuysen on using statistics in League of Legends writing

(Editor’s note: This is a recurring column that covers the ins and outs of esports media and journalism. With more interest than ever in esports, there has likewise been more interest in the reporting to this industry and what goes into it. I hope to shed a light on the industry through interviews with media professionals and news analysis from the perspective of someone with six years of professional journalism experience. Anything you’d like to see? Send me an email at Vince@slingshotesports.com. Enjoy!)As League of Legends popularity has expanded, it’s logical to see the evolution of statistics and analysis.

Analytics have become common in traditional sports, and League of Legends’ growth has come with a similar numbers-based movement. The game has many different statistics — some more complex than others — that has led to many uses when writing and reporting about the game.

Much like in traditional sports reporting, using statistics correctly and with context is vital. I had the chance to talk to Tim Sevenhuysen, a freelance writer who started the statistical website Oracle’s Elixir, about how to frame League of Legends stats, what changes he’s seen in using numbers, and the importance of sample size. (Editor’s note: The interview has been trimmed and edited for clarity, and a video of the interview is below)

Vince Nairn: There’s been this growing movement of analytics writing — and we’ve seen in traditional sports, too. So I’m curious as to how you first got into League of Legends, how you came up with the idea for Oracle’s Elixir, and how did you get from that point to where you are now?

Tim Sevenhuysen: Sure. So backstory is, my very early interest in esports was through StarCraft: Brood War, when it was kind of the second half of its original run in Korea, and then I started finding some English links and started watching. So I had a background awareness of esports, and enjoyment of it playing the games myself. And then around early 2015, I discovered the match history pages of matches at lolesports.com. I had no idea they existed. So I started poking around and found there’s a whole bunch of cool numbers here. I’ve never seen anybody do anything with them. I did some web searching. “Hey is anybody doing anything with these stats?” And I didn’t find anything that I really loved. But I thought I could gather these and start aggregating it into my own database. I know how to build a basic website. I could report these in some stats tables, and I think it might fill a niche that hadn’t really been built up yet.

I have a personal kind of traditional background in research. I have a Masters degree in sociology, where I did a lot with statistical analysis and other kinds of research. So it kind of fit within my skill set to work with these numbers — and not just to be able to process these numbers and get something out of them, but to communicate them and write reports and writings, that sort of thing. So I started up the website, started posting numbers, posting interpretations and some articles. It all kind of snowballed out from there with OraclesElixir.com.

VN: How did you come up with the name?

TS: It was a classic item. It used to be a consumable item in League of Legends. I don’t remember when they took it out, Season 3 or Season 4. It basically gave you an aura, either for a couple of minutes or permanent until you die, depending on which iteration of the item. I thought it kind of played into the symbolism of “See what’s beyond that obvious” with numbers and statistics. I have a lot of vision based plays on words with different things I do on the site.

VN: As you’ve been running the site, how have you seen the use of data and analytics evolve?

TS: I think the first place I started seeing a difference being made was in journalism, actually. Writers who would write about League of Legends, whether they’re analytical writers or more storytellers. They either want something to say “Hey this player’s really good” or “They’re good at doing X,” and be able to have some numbers to support that. “Hey this support does a lot of warding. Here’s how many wards they place.” That kind of thing started to come around more and more. That was a big change, I’d say, over the first year of the site. I started seeing may more articles contain numbers within the articles, and not only being aware that the numbers are available to use, but being aware of what the numbers actually mean, the proper ways to use them — some of the improper ways to use them…the journalism community was the first community in which it started to make a difference and made some inroads. I also started seeing it happen more on air with broadcasts and the casters in different regions. One of the people who started picking up numbers from my site quite a bit was MonteCristo in the LCK. So every now and then I would provide him a little mini report with something.

And I noticed some people in other regions doing that as well. The LCS regions in the past year or so have really amped up their internal stats teams. They’re doing things in house more and more and are not relying on outside sources as much. They’re doing great work. And then also within the kind of coaching sphere, that was one of the audience groups that really took to my site as well. People who (say), “Hey, we hadn’t been pulling in these numbers before. We’d like to have pulled these numbers but we didn’t have the manpower of knowhow to do it.” And so that’s obviously harder to see because people are obviously gonna be secretive about their methods. “We’re gonna use these stats in these ways, and we’re not gonna tell you because it’s a competitive advantage.” But definitely with the networking I’ve done, conversations I’ve been able to have there, there has been a greater push for using statistics in professional analysis and coaching.

VN: What are some of the challenges that go into knowing which stats to use when reporting and such?

TS: There are two sides to that story, and I’ll go through a couple of the big challenges. One of the things I was doing, and I caught myself on a while was, people would (push ideas) and I said “Well, we can’t do that because of these reasons.” But there’s also some good sides to the League of Legends statistical environment, which is that we don’t need a team of people sitting at their TV and writing stuff down by hand or entering it into a computer to get our data, right? Because in traditional sports you either need a large team of people doing cheap labor to gather a bunch of stats by hand, or you need really sophisticated technology.

In League of Legends, the game itself is gathering most of the data, as long as Riot has programmed it to do so. So there are things that it can’t capture. So for me to be able to run my website — and I’m the only one who’s ever touched it, really, other than the design of some of the pieces of the site — but I put the data into my site. I’m the only one who has to touch it because the game is doing it all. So that’s the big advantage. Some of the biggest challenges. The core challenge in LoL stats, is every traditional sport you win or lose based on scoring points, pretty much…so when you’re doing a stat in that game and you wanna know how much does doing that help you win. Well, how much does that affect scoring points.

There’s a really good stat in football, for example, were you say OK, you started this down on this yard line. How likely are you to score from there? And this play brought you to this yard and now it’s this down: How likely are you to score from there? And the change in percentage, that can be measured numerically. You cannot do that in League of Legends because there is no fundamental scoring system. Either you destroy the Nexus or you don’t. So any stat we gather is based on “Well, having more vision is good” because we know it’s good. But we can’t measure how good it is to have more vision. Or getting a kill. You can’t relate this kill to that win rate. Win rate is the only thing you can measure in a meaningful way in League of Legends, honestly. So when people talk about “Wins Above Replacement” or that kind of idea in LoL or these kinds of more mainstream (stats) that have really been embraced in traditional sports, I don’t think it’s possible because you can’t tie things back to a central scoring system.

VN: How do you kind of weed through what is more or less important? Everybody has different ideas about what stats they think are useful and how they tell a story. How do you kind of weed through that to see what is useful?

TS: It’s very contextual and it relies a lot on the expertise of the person speaking or writing about it. There’s a lot of opinion involved — how important is getting towers or getting Baron, how important is damage and all the complexities of how damage is measured. We won’t go into that. Damage is the messiest statistic in LoL — but there’s a lot of need for the person speaking to understand what they’re talking about. There cannot be a purely numerical analyst in League of Legends because that person cannot generate meaningful value because they won’t even recognize their subjective bias in what they think is important and what isn’t…I think it just comes down a lot to expertise and game knowledge and to be able to adapt and react quickly as the game changes as well.

VN: One stat a lot of people seem to use or at least think is simple is K/D/A. On the surface it seems useful, but how does it go deeper than that?

TS: K/D/A is great for a lot of reasons. One is we’re very familiar with it. It’s been around since the start of MOBAs. It’s very intuitive to understand. Did you do more killing or did you do more dying? It’s pretty straightforward. And it’s easy to interpret and fairly stable. When you look at a group of players — and you have to look at the same position. That’s one of those pieces of context you have to be aware of — it’s obviously not a perfect representation of how each player has contributed. It’s usually not bad. If you have sort of a subjective idea of “Who are the best mid laners?” KDA will usually reflect that pretty well. There are some exceptions. For this year, Hai in North America is a pretty big exception. He has a pretty low key KDA because he dies a lot. But he is very valuable to the team. The reason he dies a lot is he’s making a lot of plays and throws his body at the problems.

VN: How much have a change have you seen in the numbers for players since the LCS went to best-of-threes (and the one split of best-of-twos in Europe)? How much more can you do with the extra data?

TS: There are two ways the change from single game to series play has made a difference. One of them is obviously sample size…before, we would get 18 games in a split. You’re just starting to get to the point where you can make meaningful conclusions based on numbers with 18 games. Now we’re seeing two and a half times that, which is great, especially if you want to be able to track trends. You can actually talk about it in a meaningful way. Increased sample size absolutely helps the ability to do analysis and statistical analysis. The way it’s changed the competition I think is it’s reduced the volatility in the results.

When you get a team like Fnatic going 18-0 in summer 2015, an undefeated split. If they can do that in a best-of-three or best-of-two environment, that’s a wholly different thing that doing it in an 18-game season. In that split, Fantic had two games where they fell behind by a lot early and made a massive comeback and still won. In double the number of games, they might not have made those two comebacks, and now you have a couple losses in there. There’s much less chance for these crazy high win numbers or high loss numbers to come through and make it seem like the gaps between these teams is much bigger.

On The Radar

(A rundown of important stories in the esports media realm and analyzing what they all mean.)

If you missed it (and in 8,000 words of interviews, you probably did), Richard Lewis talked about some of the challenges of adjusting to live TV broadcasts that were of particular interest.

“You’ve gotta be able to, when you’re the host, you’ve gotta be able to be talking, and what comes out of your mouth is effectively filler,” he said. “It’s buying time. Your brain isn’t linked to your mouth. It’s shit. Your brain is actually listening to the producers. “What we’ve got coming up for you,” is actually just filler. It’s a sentence to buy me time to hear what is fed to me in my ear, and what comes after that filler sentence is what’s important.”

Aside from the hilarious sound bite, Lewis brings a fascinating point. Live TV is so experimental for esports that even veteran casters of Twitch broadcasts can get thrown for a loop with just how different TV can be from a hosting standpoint. Lewis said there are as many as three different people in his ear, and the polish required for a TV broadcast is a definite adjustment compared to the more relaxed manner of most esports broadcasts.

Lewis also talked about how much more comfortable he is now than when he started working for Turner’s ELEAGUE a year ago. That makes sense when thinking of Turner’s decision to keep Lewis as the host for the Street Fighter V Invitational, a scene over which Lewis has less of a grip than Counter-Strike, where he has been a staple for more than a decade. What’s easier: Asking a journalist who grasps the technical part of being a host to learn a new game? Or bringing in an FGC expert to start where Lewis did a year ago?

Slingshot Editor-In-Chief. Former newspaper reporter from Cleveland, Ohio, who appreciates clean copy and good Counter-Strike.

Facebook Comments