Jens Hilgers – The challenges in esports data

22 November 2016

Share

Last week, esports startup DOJO Madness partnered with global leader in sports data, Sportradar. We spoke to DOJO’s CEO, Jens Hilger about the partnership – you can read it here.

We also spoke to Jens about wider issues in the world of esports data. Here’s his thoughts and insight into how patches, player weightings and more effect the algorithms that his team use to provide great data insight. cpa0vvexyaakte3

ESI: Modelling esports in general must be a great challenge. Building algorithms to help predict and ultimately being able to offer in-play odds must be tough. Tell us a little bit more about it.

Of course, it’s fundamentally a challenge for us. There’s also a great opportunity there. The amount of data we have is greater than what we would have in most traditional sports. There’s a greater amount of inflection points and game making points that allow us to model outputs and feed into our algorithms. It allows, or will allow us to create greater diversity on in-play odds. Typically esports titles have a lot of exciting in-play events and objectives that are genuinely fun to follow.

ESI: How does it work when a new esports game appears? People often forget that it would be akin to a whole new sport appearing on the scene and bookmakers suddenly offering odds. Also, big gameplay patches surely fundamentally alter the way your algorithms work. Does a lot of the historical data become obsolete when a patch comes out?

You’re spot on. These games change and it’s great for esports in general. It’s an additional point of drama and excitement. When did rules in soccer or NFL last change? Simply put – they don’t. It may be great because it’s more comparable from a historical perspective but on the contrary – for the audience it’s really exciting if some element of the games change or adapt. Sometimes they want a reset. It helps to extend the lifetime and excitement around the game if certain things change. If there weren’t gameplay patches in a lot of these big titles, people would probably get bored. You might argue that they’ve changed for the better and the greater of the game – if you tweak a game sometimes then the game will feel like it’s renewed. Kind of like “this is soccer 2.0 and that’s exciting”. The challenge for us is, like you said, large chunks of historical data can suddenly become obsolete overnight with the announcement of a new patch.

That’s why I need, and have experts that understand the games deeply and essentially think in game design mechanics to evaluate which of the historic data we can still take along with us. The most stressful time at DOJO is when we have big patches and we need to see what has changed, how big an impact it will have and what it means for our predictive models. That’s where the responsibility but also the skill and deeper understanding of the game of our data scientists at DOJO comes into play.

“You really need people who are intimately knowledgeable of the game and have played them for years if not decades.”

That’s the kind of people we’ve hired and make up our team. It’s almost intuitive to them. They read the patch notes and they can tell you “the bigger picture here of this tweak is XXX”. They hit it to 99% accuracy as they know the game so intimately. Yes it’s a challenge but it can be overcome to a certain extent.

ESI: Are your algorithms purely based on game related data or is there a weighting on the players?

It’s an area that we’re already starting to have a look at. We’re trying to reflect in-game data but also “contractual” data as it also dampens any slight misinterpretations with a new patch. If we start to look at contractual data – player salary, transfer fees and the impact that individual players have then our algorithms can become more robust.

Don’t get me wrong, it probably wouldn’t be as strong as say – a recent headshot percentage – it’s still a data point for us and the weighting remains constant through patches. I think individual skill characteristics, however we manage to aggregate them are vitally important going forward. If we get it right it will help to mitigate to some extent the challenges of frequent patches and changes to the game itself.

ESI: Where do you see the future of in-play betting in esports going? Just how granular can we go with markets?

I’m not strictly a betting man, but if I would offer an in-play bet I would want it to be an interesting user experience. I have a hard time qualifying and commenting as I am not strictly from a betting background so to speak.

What I can tell you though, is that we will be able to provide a good depth of odds at the micro, granular level of in-play events. I think the most important part right now is that we offer what is being praised and hyped by commentators and hosts.

What is so important at the moment in events is the drama that commentators present on a stream, and I think we need to transfer that to a betting opportunity. One of the greatest things in esports is the retake of a bomb spot in CS:GO. The probability of that happening is a big drama point in the commentary and so that’s obviously of great value. Similarly, if someone scores an ace in a round or a big clutch play is made then these are the events that are most interesting. They’re of huge excitement to the audience and the hosts and commentators place extra emphasis on them. These are the points that we would love to offer odds on.