Gaming

Dojo Madness wants to make you better at Dota

An ESL founder is trying to bring analytics to eSports – read the interview here.

Written by Ben Sillis

9 min readPublished on 02.08.2016 · 10:09 UTC

Dojo Madness wants to make you better at Dota.
© Dojo Madness

Every top eSports team has coaches to analyse their players performance, but for everyone else, such advice is difficult to obtain. Sure, MOBAs like Dota 2 and League of Legends have thriving, helpful online communities happy to talk shop and strategy, but they're not able to pore over your replays to pinpoint where exactly you went right in a match – or wrong.

Enter Dojo Madness. Not the iPhone game – the Berlin start-up of the same name. The team of 20, which just raised €2.5m in venture capital funding, is the brainchild of ESL founder Jens Hilgers, and its first two products, smartphone app LOLSumo and website Bruce.GG, are all about helping you as you play and serving up analysis specific to your game as it happens. To find out more about the company's ambitions to be the virtual coach you always wanted, we caught up with Hilgers for a chat. Read on to see how his ideas could make you a better gamer.

Tell us how Dojo Madness started. What made you all quit your jobs to found a start-up?

We got this company started as a bunch of people that knew each others for years already, with the vision of building tools that help gamers master their game. The vision was based on the issues we faced in our day-to-day gaming experience. We thought, 'we can't be the only ones having these problems. Let's build these tools that help solve our problems and very likely the ones that others are facing too'.

Your first product, LOLSumo, is a second screen app for League of Legends. How did that come about and how does it work?

It advises you in real time while you play your match of League of Legends. The app is reading the game data and gives you a good build that you should use, the hero set-up that you should have in your game and the skill priorities that you should look for whilst giving you some contextual information, such as opponent team set-up. This is the first product we've been working on and it has been live on Android for two months. I think its very useful, it makes your play easier. Personally, whenever I go back to League I have been opening sites like Mobafire, where you get a guide of what hero build you should go for.

I've played A LOT of League and Dota very passionately – since 2005, so very early on i guess – never professionally though. At all times playing i wanted context and guidance to make goodstrategic choices in the game: How to skill that hero, how to play him best, what items to buy, the list goes on and onn.

I couldn't keep up learning fast enough with the time I had available though. I wanted to have good advice: Show me a good build, show me a great strategy. I'd rather have somebody who gives me a good recommendation, than wasting 45 minutes [the typical length of a LoL or Dota match] because specifically in MOBA games you can really throw an entire game with just single mistake – it's so unforgiving.

This app helps you out with those problems, but it doesn't need you to switch or alt-tab to my browser to look up a guide or forum, which is what I always did before, missing half a minute of the game and making everyone angry at me for not paying attention. Now it sits right there next to your monitor.

Your second product isn’t an app so much as a platform. Tell us about how Bruce.GG works.

Bruce.GG tackles the problem from a different perspective. It's much more complex; a video platform that analyses your personal gameplay, and provides you with highlight videos of your matches to show you what you've been doing right and wrong. It does that once you've finished the actual match, so it's not a real time coach like the LOLSumo app.

We felt that once you have played a Dota 2 match, if you lost, you probably didn't always realise why you failed, or what was the deciding moment. You have relatively limited learning. Mostly when you lose all your friends are fighting with you and each other about who was to blame for losing. That in itself is often a fun experience, but the problem is nobody ever has the evidence on the table. Now you can see: you've not used your Ultimate right, for example.

The target that we have is providing little highlight videos of 20 to 30 seconds that allow you to learn relatively effectively compared to watching Twitch. The platform retrieves the match data, understands what the deciding moments were and then renders the videos in the cloud for you to watch or on your desktop, tablet or smartphone.

How exactly will Bruce.GG pick those moments out for you?

First we train our system, and afterwards we let the users continue to train it, and it evolves via machine learning. The feedback of the users watching these videos and interacting with it provide us with data points we use to determine the relevancy of what we have identified so that our system can get smarter and learn.

There are different components to Bruce.GG that will go live over the next few months. The machine learning part is coming in the window of the next two to three months, on a relatively simple basis. We're also developing tools that allow you to commit analytic feedbacks. Users will be able to set position markers and be able to say, "That guy was out of position." Users will be able to give pretty sophisticated feedback. In a way, this is enabling them to become coaches and analysts themselves because they have tools like an NFL coach would have.

The Dojo Madness team (Hilgers centre bottom row).
© Dojo Madness

Bruce.GG ultimately sounds like replay analytics in the same way you have for half-time in sports. Is it your plan to bring this technology to professional eSports?

There's three different applications of that technology, not all with the same priority. Our first intention is to use the technology to help players to master their play. The second application is generally identifying highlight scenes, not only of individual players but of professional players in high level tournaments, and surface those videos to the general Dota or LoL crowd and make them available, just like any sports broadcast where you get your highlight videos right away.

The third application is the professional application: we are providing tools here based on our technology that help analysts and professional TV producers to interact with the data that is underlying in these videos – positional data, cool-down and item data, and we can surface this to the users and the professionals and let them interact with it.

What’s your business model? Which of these will drive revenue?

I think in the first place we're looking at coaching individual players. We believe the broadest audience that can make use of our technology is individual players, and we believe that we can deliver such value to them that eventually they would be willing to chip in for a premium account that gives additional features on our platform. We believe there is a sufficiently big audience in League of Legends, Dota 2, Heroes of the Storm and Counter-Strike, which are all games we want to support in the future.

Why Dota 2 first for Bruce.GG?

Dota 2 has provided a very reliable and stable data API for quite some time already. It felt like that was the best foundation to start on with Bruce.GG.

Most of the games you mentioned are MOBAs, but CSGO is a team shooter, a very different genre. Will the tech still work? It's a super exciting challenge for us. The big difficulty being that we are RTS [realtime strategy] players in the first place. It's obviously easier to design something based on what we play every day. As a result we are focused on recruiting more talent with an FPS background and experience in Counter-Strike. We've got some already but we need more to build a decent product. From the research we did, we feel there is plenty of opportunity to help Counter-Strike players in the game, similar feedback we got from pro and casual players alike. We feel [the advice we can give] is probably not as as deep and sophisticated as in League or Dota 2, but we feel that there are some things that are super helpful that nobody has done yet which we should push for.

The meta in MOBA games changes frequently, with every patch update in fact. How will Bruce.GG factor that in?

That's where the machine learning comes in because we can't always reteach the entire system – that's work that is unaffordable at that scale. The machine learning allows us to save on some of that work, but we've got to keep our eyes open. The game team that's responsible needs to look at the patches as closely as the players should look at them, and identify potential changes in the most important scenes to prevent the platform from failing.

Doesn’t that also mean that the advice Bruce.GG gives is going to stop being quiet as helpful every couple of months?

It's expected. I think it's a challenge but I dont think it's something that will break the product. We will learn how to deal with it over time – even during development now we've seen a patch. We feel comfortable that it's working.

Are you talking about timings for other games to arrive on Bruce.GG?

No details here. Big picture: we are looking for the next game to potentially surface in Q4 [October to December] this year. Right now we are putting all our resources into Bruce.GG's beta to make sure it works well with Dota 2.

If we're unable to get there at all – this is a start-up, a bold idea – we might just fail with our vision and idea at all right? We might well have to say say, "Well, we've tried everything in the box to make this technology and product work and it turns out nobody's interested in it and it does not provide the value we expected it to do." This is what we are extremely honest about with ourselves.

We're trying something bold, of great technical complexity and challenge and ultimately it needs to be providing significant value to esports players. Thats what we work hard for and believe we will succeed in.