The role of AI in esports betting
By Kenneth Williams
iGB: Can you tell us a little bit about what PandaScore is?
Flavien Guillocheau: We founded the company [a little over] five years ago. When we started out, I was looking at what existed in traditional sports. It’s more mature nowadays, there are many different activities for fans to engage with esports. Five, six years ago, it was much more difficult for a fan to do anything. For example, if you were trying to place a bet, it was difficult to find a bookmaker. If you were trying to find very basic information like who’s playing tonight or the result of the match yesterday – the basic stuff that, as a fan, you want to engage with – there wasn’t much there. My idea at the time was to help businesses build this kind of basic experience. What we’re doing is providing data on esports matches to different types of businesses. We work with some professional teams, we work with the media, broadcasters and tournament organisers, fantasy, betting and so forth. Our mission is to help any business build whatever kind of experience they want for fans.
iGB: What was PandaScore’s impact on the recently upgraded esportsbook from Pixel.bet?
Flavien Guillocheau: A year-and-a-half ago, we decided to launch a product dedicated to bookmakers because we realised that the betting companies couldn’t do much with just data, raw statistics. They don’t have the esports expertise internally; they can’t build the odds themselves. So two years ago, we decided to build our own odds feeds coming from our data that we collect in real-time, after the match, etc., and transforming this data into odds. What we do with Pixel.bet and all of our customers is we provide them the odds feed pretty much live. We provide them on many esports. We cover the big three: League of Legends, CS:GO and Dota 2, plus Rocket League, Overwatch, Call of Duty, so on and so forth. More recently we added Rainbow Six.
Our relationship and involvement with Pixel.bet is on helping them offer the most complete experience possible in sports betting. One of the few things that really stands out is the diversity of bets that we can help Pixel.bet offer to their customers. For League of Legends we would offer around 70, 80 different types of bets ranging from the basic match-winner to how many tricks, how many kills, very specific esports types of bets. At the end of the day, this is what esports fans want, to have something specific to esports and not just very generic match-winner, handicap, things that esports fans might not relate to that much compared to sports markets.
The last thing we do with them is provide them with some raw statistics through a widget. It’s basically a webpage that you can integrate into your own website. These statistics help the fans and punters bet. It helps them relate and understand what is happening. This is also increasing engagement from the punters and something we really find valuable, especially in esports. Esports is more data-driven than traditional sports. It’s the DNA of gamers to compete and score points in every game they play. They have a strong relationship with it all the time.
iGB: Why is PandaScore’s AI a game-changer for esports betting odds?
Flavien Guillocheau: Since the early days, it was very difficult to imagine doing data collection manually or with an approach that isn’t scalable or automated. Very early on, we decided to build an automated process to collect data. Six years ago, it was the early days of AI in the more mainstream startup thing. What we do is we use two different parts of AI. First is computer vision, so we analyse video feeds with AI the same way a Tesla car drives and analyses what’s happening on the street. We use the same type of technology, and this allows us to create statistics.
As for the impact of AI and how it’s a game-changer, the depth AI can bring in data collection is unique. It can create a huge amount of data very quickly. Coming back to League of Legends, we collect 300 data points in half a second. It’s a lot of data. It’s very quick, very automated and very accurate. That’s the first layer of where AI can help. If you think about it, it can also be applied to traditional sports. Not just to esports, it can be applied to anything.
On the second side of AI, it’s more traditional. It’s about how to predict what’s happening. With the mass of data we create, how can you predict who’s going to win? How many points, how many kills, how many objectives, and so on and so forth. Which player is going to perform the best? It’s a lot of machine learning, which is a bit different from AI, but it’s also applying some newer models to these kinds of predictions. On this specific aspect, I think esports and the complexity of esports rely on the amount of data you can generate. It’s much more virtual: there is much more of an economy in the game, there are more items, so many things that compose the game that are very rich. If you’re treating esports with a normal sports approach, it’s going to be very difficult to predict anything properly. It changed the quality and depth of the product that you can build and the kind of content you can create for fans who want to place bets.