Dennis “Barroi” Matz from Winston’s Lab sat down with Esports Heaven to discuss his part in creating Winston’s Lab and his thoughts on certain statistics. Added in is a fun word association test just for some laughs.

To start us off, let us begin with where you started off in esports. Talk to me where and when abouts did you really start to enjoy competitive gaming?

I probably started enjoying [esports] during Season 2 of League of Legends. Well, I watched League of Legends since Season 1, but I really started enjoying the Korean scene once it really started to kick off in Season 2. From then on out, I was watching pretty much every Korean match there was. 2015 was when fantasy esports really started getting big with Vulcan, Alpha Draft, and all of those sites. Thats where I started making some money because, fantasy esports is easy if you know your way around numbers. [laughs] I’ve always been into mathematics.

Could you talk more about why you decided to create Winston’s Lab?

Like I said, I was super into mathematics and statistics, I always have been. So, once I moved on from League of Legends, I noticed that Overwatch was getting pretty big and didn’t really have a hub for statistics. I found a cool guy who made a cool program and thus Winston’s Lab was created. [laugh]

Now, a topic that is thrown around a lot is the “moneyball” aspect of esports. The idea that you could create a winning team solely based off of statistical analysis and numbers. Do you think that could be possible in Overwatch?

I think Overwatch is probably one of the hardest for that. First of all, I don’t think that is really possible, in any sport. If you want to fully evaluate and understand a game or a sport, you’d want to make it as easy as possible. Take a game like Baseball: you hit the ball, you run, you catch the ball, you throw the ball, etc. [laughs] Overwatch has too many aspects to it. You have 25 heroes to pick from, then you have a team of 6 players, you have a possibility of 25^6 different combinations of heroes you can play, more or less. Then you have to start considering different abilities and maps, it just gets out of hand. In theory, if you want to understand something just off the statistics alone, you want something that's a bit easier and Overwatch has too many variables.

Most recently Winston’s Lab added an “ELO” style rating system that judges players on a match by match basis. Could you talk a bit about the reasoning behind why you did that and what you're hoping to find?

My primary motivator was that I, myself, tried to look at players for some of the reports we write at Winston’s Lab and there was around 20 categories of numbers that we would take into consideration. With all of the different metric spanning from how important is it that you get kills to how important not dieing in this scenario is, it tended to get a little mentally stressful. I just wanted to have one number that showed you a rough idea of how the player played on a certain map. Obviously we can only work with what we have available to us, so we try to be as accurate as we can be with the statistics we have to work with. That’s why I went and made those ratings, really it was for my own selfish pleasure. [laughs]

Something that is more of a topical point in Overwatch currently is the ongoing discussion around SR is derived based on statistics? Why do you think Blizzard chose to weight your SR gains or losses the way they do?

Plain win and loss ratios, also you could look at win rating, tend to need a lot of games for that to really calculate your proper SR rating. You really need hundreds or thousands of games for your win rating to be the ultimate factor. If you can consider damage and healing in a reasonable way, you can use those to decrease the number of games players need to go through until they really need to see how good a player is. That's why I think Blizzard made those performance ratings, which I don’t think are a bad idea. The real question is if they were implemented the correct way or if they are a little too strong in certain areas. In general, I think the idea was good. It’s hard to say if the implementation was bad or good, because we don’t have access to the algorithm. If I could see that I could obviously tell you, but it's hard to say because we don’t know the benchmarks that Blizzard is using.

Something that has been clouding the general consensus around statistics is this stigma that they promote bad behavior. In your mind why do you think that is and do stats cause toxicity?

In a way, it is easy for players on the ladder to place the blame on someone if they had the statistics to know that this player was in the bottom 25% of all players. It is very easy for someone to take statistics and use them in to flame someone who isn’t performing very well. Overall, that is a good question. ‘Why do people think statistics are bad?’ I think for players it’s moreso that they believe that statistics are too plain to explain the nuance of the game and I agree with that. Using only statistics to judge someone is probably not ideal. I think the community take the pro-players opinions and extrapolates them to the extreme. It seems like players think that the stats are not in a good position to be a direct indicator of what is arbitrarily good and bad.

Let me throw this curve ball at you. Do you think it is possible for a team to win a professional game with worse stats, on average?

It’s slightly possible. It’s very fringe, but let's say you win a best of three, 2-1. The first game of the series you win in very dominant fashion. Game two, your team ends up losing a very long and drawn out map. Let’s say for the sake of simplicity it was on Temple of Anubis and your team could not capture Point B and the enemy team quickly takes both points. Then you got to the third map and you quickly dominate them again. Because the second game was longer, it gives the enemy some room to pad their stats, thus making your team's stats look much worse than they actually were. There are ways to manipulate stats, but it’s very fringe and unlikely for these things to actually happen. It always depends on how much context you add to statistics; that is pretty much always the case. If you have people saying that the stats are bad, then you probably did not add enough context to support them. You always have to take into consideration the map, the team composition, etc. For Overwatch you need a lot of context because you have so many variables to the game.

Do you think there are intangibles in Overwatch that we cannot account for? Do you think, in the near future, we could have the technology to attempt to account for them?

There are still some intangibles. For example, to calculate the impact of the kills you had in a certain game or match. To grade how impactful each of your kills was, you would need to know which heroes you killed and if they had ultimate charged. I think that is possible to do, it is just a lot of code and a lot of thinking. Especially when it comes to thinking about the impact of which hero you killed. Or we could look at how impactful opening up a flank is. That is something that is completely intangible. The idea of potential positioning is probably completely intangible, so you couldn’t add that to an “impact rating.” The concept is too abstract to really put a numerical value on it. In theory technology could put a numerical value on it, but I don’t think it will happen. It would require the API to be incredibly more detailed than it probably will be. I hope at one point we can consider positioning into the equation. That is one of the most common topics that coaches and I talk about. They would love the ability to look at heatmaps to see where teams position and how often they take that position.

Something that kind of popped into my head as you were saying that was, wouldn’t this become infinitely easier if we had replays? I am sure that Blizzard is currently working on them, as we speak, but if and when we get the ability to record replays wouldn’t we be able to see heatmaps and positioning more clearly?

No, I think that something that is way more important to statistics alone is the API. Replays help to open up more ways to refine the infamous “eye test.” I think replays will be key in helping to really push analysts ability to read good and bad play, but as along as replays don’t have any more or less information than the games itself, statistically speaking, then I don’t think replay will get us on a new level just solely statistically. Replays are an important key in a different way, I feel.

You’ve and the team at Winston’s Lab have been doing quite a bit of research and analysis of the current pick rates of heroes. So, are we in a “stuck” metagame? Statistically speaking, do you think it’s time for change or is there more tools that could be explored?

Statistically speaking this metagame is a mess. I wouldn’t say that is good or bad. If anything it would be good for us to look at it statistically speaking because there is not a lot of variety. It is however, the least diverse metagame we’ve ever had. I go more indepth with this idea in an article I wrote for Winston’s Lab where we took a look at the cumulative pick rate for the top ten team compositions and how often they are played. Not only do we have four heroes that are picked a majority of the time, but in 3 out of 4 matches we basically see the same 10 team compositions and only those. Whereas early this year we had much more diversity in how the game was being approached. This metagame, is by far in the way, the least versatile metagame. Take that however you will, it isn’t good or bad, it just is. [laughs]

As for statistics, is there anything that Blizzard could do that could make analysis easier? How pivotal would the Blizzard API be in statistical analysis?

They could do a lot to make things easier. The more things the API tracks, the better. With access to the API we could automate stuff, which makes things a lot easier. If the API has certain things, than we can start to look at different and new metrics, which is always good. I am just hoping the API is as detailed as it could be. I could do so much stuff with it, it is unimaginable the number of new metrics and new statistics that could be derived from it. It is just too much to think about. [laughs]

Something that I thought would be interesting would be some simple word association. I am going to say a word, related to Overwatch and esports in general, and you tell me the first thing that comes to your mind. Ready?

Ahaha. Sure, let's go.

Tracer.

The first thing that comes to my mind is the stupid butt controversy, but for more esports related things I guess the pick rate? That’s boring though [laughs].

Defense.

Reinhardt and Bastion. It’s not super exciting, but that what first comes to mind.

Winston.

Peanut butter. Another thing as well, Winston has the best death animations in the game. I love how Winston dies.

Zenyatta.

Balls. Big, painful balls. I love Zenyatta by the way, best hero to play. :D

KT Rolster.

I like KT Rolster, they were the first team I bought merchandise from in esports. When I think of KT I always picture the big SKT vs KT matches and how KT loses every single time. So, I guess the first thing that comes to mind would be; losers, but in a good way. [laughs]

Looking at the upcoming final for APEX Season 3, what are your predictions on the third match between KongDoo Panthera and Lunatic-Hai? Who takes home the gold?

One months ago, I said that the match between Lunatic-Hai and Afreeca Freecs Blue would decide the tournament. So, I am sticking with my guns and saying that Lunatic-Hai will win convincingly. I don’t think it will be close, either. If KongDoo wins this, I think it would be really exciting and interesting statistically. Winston’s Lab has this PTK (percentage of team’s kills) model to start to see who the impact players are. More often than not, the team who wins is the team who is “carried” by the tanks and supports. It is way more important to have an above average tank or support player, than it is to have an above average DPS player. Obviously you don’t want bad players, but if you had to pick one or the other, I think tanks and supports are way more important. KongDoo is an outlier to that rule, where Lunatic-Hai is the direct example. If KongDoo wins, it will be very exciting, but I don’t think they can.

And to end I’d like to ask one last question; why? Why Overwatch? What about Overwatch sold you on it as a competitive environment? What about it do you enjoy? Why do you enjoy watching it? What about it do you love?

That is a good question. I guess you could say I don’t really ‘love’ it. I enjoy the numbers behind it. Overwatch has a lot of problems. I just really enjoy looking at numbers really. Overwatch was my first chance to really dive deep into an esport and see how the numbers work. Metaphorically speaking, Overwatch might not be the fastest boat, it might not be the most beautiful boat, but it’s my boat.

Joseph “Volamel” Franco has followed esports since the MLG’s of 2006. He started out primarily following Starcraft 2, Halo 3, and Super Smash Bros. Melee. He has transitioned from viewer to journalist and writes freelance primarily about Overwatch and League of Legends. If you would like to know more or follow his thoughts on esports you can follow him at @Volamel.

Images courtesy of Blizzard Entertainment and Winston’s Lab.