DataBet
DataBet
Go back
30 April 24
DataBet

Unravelling the enigma of personalised markets in esports betting

In the expanding world of betting, where bookmakers accept wagers on an array of events, the scarcity of personalised markets for players, especially within the realm of esports, stands out as an intriguing anomaly.

In conversation with SBC News, Yurii Lysenko, Lead Data Scientist at DATA.BET, reveals the potential and challenges of personalised markets, shedding light on why bookmakers hesitate to enter this uncharted territory.

This research embarks on an in-depth exploration of market customisation tailored to the preferences of bettors, exploring their popularity and unveiling the factors that shape player interest. Additionally, examining challenges prevent the integration of such bets, noting their availability in traditional sports for prominent events.

Initially, we selected two of the most popular disciplines, such as Counter-Strike and League of Legends. Following the implementation of personalised markets, the appeal to players was analysed, with a focus on the evolution of interest over time. In the first weeks, engagement was minimal (up to five per cent of all bets), but it gained accelerated growth by the third month, with the total bet volume being more than 20 per cent of all bets, as depicted in the accompanying graph.

This might have occurred because players needed time to understand and build trust in a new type of market.

The interest eventually rebounded to its prior levels, followed by an exponential surge of up to 25 per cent. This pattern suggests that offering customised markets consistently has started to gain popularity and will continuously increase player interest, allowing bettors to try to predict specific outcomes or scenarios that align with their interests, preferences, or expertise.

Such a tailored approach enhances players’ overall experience by offering options that resonate with their preferences. Despite the evident appeal, there is still a scarcity of personalised bets, which can be due to the difficulties of modeling.

One of the main challenges of modeling markets is that some disciplines follow different distributions or do not follow any well-known ones, which require custom modeling. As an example, let’s consider the distribution of the number of kills per team in Counter-Strike.

The local peaks mentioned on the graph at points 250 and 270 correspond to overtime rounds. It can also be observed that even before overtime, the distribution is not symmetric. One would think that this can easily be fixed by combining beta distribution, but this approach ceases to work closer to the end of the game, especially when considering individual players rather than teams.

In scenarios where the first team is almost guaranteed to win the game, the number of deaths of a player on the second team will be at least equal to the number of rounds they lose. For example, in a match with a score of 11-4, where the first team needs only two rounds to win, a player from the second team will die at least twice.

The actual distribution will differ from a normal or Poisson distribution because the probability of dying fewer than two times is practically zero, unlike the values of a normal distribution calculated for the expected number of deaths for a player in this situation, which would be 2.89.

Although only one specific game situation was considered here, such exceptions are very frequent. Simple things like this will contribute to every step of modeling by distributions making it inappropriate for this discipline.

Hence, the chosen approach for modelling is using the Markov chain method, which allows custom processing of each state in the game and modelling the next state from round to round.

It also takes into consideration Bayes statistics when a player’s actual performance deviates from the expected value and the player’s characteristics need to be adjusted.

Above are the examples and encountered problems of modelling personal markets provided for Counter-Strike. In other disciplines where there is no fixed score, such as Dota 2 or League of Legends, normal distributions are pretty suitable, except in situations near the end of the game. Bear in mind that for the simplicity of this article, an analysis of situations where a player changes their role in the game or when their coordinates directly influence their expected value is skipped.

Building models requires a large amount of statistics and a non-standard modelling approach. Manually offering these markets is almost impossible, as odds on player markets can change abruptly based on player behaviour and in-game position, creating additional challenges for bookmakers.

Despite the difficulties of modelling these markets, the result easily justifies the effort spent. Operators offering tailored markets cultivate customer loyalty, reducing churn rates and strengthening their market position.

For instance, in Counter-Strike, individual markets were within the top five by volume, and in League of Legends, the top three within the first two months of regular offerings. Providing such markets without official real-time data is almost impossible, which explains why so few bookmakers have such offerings on the market. However, as technology advances, strategic collaborations between bookmakers and providers equipped with relevant tools become crucial for harnessing the potential of personalised markets.

In conclusion, with the esports betting landscape further evolution, bookmakers must adapt and innovate to meet the demands of experienced bettors. By embracing personalised markets and fostering partnerships, operators can elevate the betting experience, positioning themselves as leaders in the industry.

Back to all News