LoL: Introducing the iTero Drafting Coach, in partnership with Overwolf
Over the years whilst watching League of Legends esports, there’s one question that seems to remain forever debated: who won draft?
As I was working in Artificial Intelligence (AI) I knew that there would be techniques that could at least bring us closer to an answer. This theory kickstarted my move into gaming/esports and ultimately led me to quitting my job to work full-time on my start-up; iTero Gaming.
My seemingly simple challenge was to be able to predict the result of a professional game better than anyone else once draft had been locked in. To this end, I was fairly successful — although it took a lot more work than I ever expected. The average betting company came in around 62% accuracy and I’d average somewhere around 67%. A decent Alpha (not referring to myself, Alpha meaning the advantage one has over the market).
I couldn’t have reached this level of not just accuracy, but clarity, in the model if it wasn’t for time spent with professional coaches. They question every aspect of the model. I can’t just say “you lost draft here”, I have to explain why. By trying to understand a result, you really get into the depths of the model. Drilling into every minor detail until I was confident in the results it provided.
At this point, I was left with a choice of how to drive iTero Gaming forward. Esports in all it’s beauty still has many flaws. A predominant issue I came across is that the credit (and thus compensation) for analysts is inequitable. In essence, the amount of effort/time/skill required to build these drafting models was unlikely to be worth it, at least for now.
However, as one door closes...
You see, although the professional players in a competitive game will play differently to your average casual gamer — the fundamental principles remain the same.
For example, in my research I found that professional players have a better chance of winning the more games they’ve played on their chosen Champion. I can therefore take this feature (“Games Played”) and test it on the data from average players. Undoubtedly, the relationship is different but there is still a relationship to work with! We just need to adjust the influence it’s having on the prediction. All a cakewalk for modern AI techniques.
So, I repurposed the models to be able to take an average game and predict the result based on the draft. Then, it’s just a case of brute force searching all options (around 160 Champions at the time of writing) and telling the player which ones increase their chance of winning the game the most.
Importantly, I focused on making these models explainable. That means, we aren’t just telling you who to play — we’re telling you why. That way, players can understand the decisions their making.
The problem I had though, was that it was a pain typing in all the information each game. I also found a lot of value in what the model was telling me (for instance, is this a good choice because I beat my lane early — or is it because I out-scale them and just need to survive the first 20 minutes? I got frustrated having to tab in and out of the game constantly.
That is where Overwolf come in. They are the market leaders in integrating apps for games, they just need people to build good apps. iTero is looking to be the market leader in helping competitive gamers improve, we just need a better way to integrate with the games. It’s the perfect match. From here, we started our partnership.
Fast forward three months and I’m incredibly excited to introduce the iTero Drafting Coach. Simply download the app, jump into a game and let the coach work out the best Champions for the job. You can also use the old-school approach of typing the draft in, over on the iTero site.
Here’s a video showing off some features:
The important thing for me, is that it’s proven. This isn’t some hocus-pocus dice roll. If you pick one of the recommended Champions you’re more likely to win the game then choosing any other. It’s not guaranteed, of course, but climbing in LoL is a game of inches. If you’ve been stuck at a 50% win rate all season and you follow the recommendations, the chances are that your win rate will increase and you’ll start to climb.
You can download the app now by clicking the image below. I’d also love to hear your feedback, which you can send on the iTero discord.