Future Anthem: five data-driven recommendations every operator and studio should actionBy William - 20 January 2022
Pan-industry initiative, Safer Gambling Week, saw a landmark collaboration this year as research found there to be no link between game features and problem gambling.
Dr Edmond Mitchell, head of data at pioneering AI and game data science specialist, Future Anthem, explains how the research was conducted and the key takeaways for operators and studios.
Take the merest of glances at the public statements of the CEOs of the biggest firms in our industry and it is clear: no subject is more important than responsible gambling.
It is woven into our DNA at Future Anthem. We are committed to enabling operators and studios to provide an enjoyable and sustainable experience and this year’s Safer Gambling Week gave us the opportunity to make a statement with our groundbreaking partnership with operating group giant Gamesys.
After analysing more than 36 million gaming sessions by 800,000 players, totalling two billion spins across five million hours of gameplay on 368 games, we found no statistically significant correlation between player markers of harm and three structural game characteristics: volatility, return to player (RTP), and hit rate.
Furthermore, there was no evidence that specific risk indicators, such as staking up, were linked to specific games within the category. The research was driven by Future Anthem’s Safer Play product, which uses state of the art AI to proactively detect in-session wagering behaviours that indicate players may be exhibiting markers of harm, creating a score for every game session for every player.
Its machine learning models assess more than 80 separate metrics that correspond to markers of harm which have been honed across billions of spins to cluster game sessions together according to their level of risk.
We believe this research is the first of its kind as most other research in this area understandably focuses on the player themselves. In addition to its headline findings, the research spotlighted a number of insights – here are our five data-driven, responsible gaming tips every operator and studio should follow:
- Be extra vigilant monitoring early morning sessions.
It often assumed that overnight play may be more prone to risk, however little research has been published in this area. Our modelling found sessions from 12 to 6am were 36 per cent riskier than other times of the day.
- Be aware that once a player has one high-risk session, they are seven times more likely to have another high-risk session compared to a player who has just had a safe session.
Our research looked at how players transition between the risk sessions and the probability of a player moving and staying on a risk label. In more detail, more than nine per cent of high monitor players continued to be in this ‘monitor cluster’ in their next session with 1.1 per cent going onto the high-risk cluster in their next session, and nearly four per cent high-risk players remaining in the high-risk cluster in their next session.
While there were some games that we’re more likely to have player transitioning from one risk category to another, there was no pattern or shared characteristic among them that correlated with risky play.
- There are specific games which, overall, tend to have a higher prevalence of risk occur on them – be aware of these games and try to understand why this is the case.
Despite no correlation found between markers and harm and game components, we discovered examples of specific games that had a larger probability of risk occurring. While this could be due to a multitude of reasons such as promotional activity, acquisition offers, risk scoring games in this manner could be used to recommend games that are less likely to display markers of harm to players who are showing early signs of risky play.
- Monitor players who drastically change playing behaviour i.e. large increase in usual max stake or sessions much longer than average.
While it may initially seem counterintuitive that a game measure such as volatility is not correlated with increased player risk, it conforms to the body of research on markers of harm, which conclude that player behaviours and players exhibiting markers of harm are related to factors about that individual.
It is therefore always best practice to monitor players who suddenly change their gameplay behaviour such as large swings in average stake or sessions that drastically exceed their average session times.
- Send risk scores back to game studios to investigate what features may drive risky play.
Improve customer experience by risk scoring your players and sharing this information with your studio partners. This will give you a view of which of your players are at risk and those regularly displaying high levels of risk. This data-driven approach will enable you to find out:
• How many risky sessions occur across your portfolio and how this changes over time;
• What a ‘risky session’ looks like with staking behaviour graphed for easy review;
• ‘Explainability’ of what markers of harm occur within these risky sessions;
• How many and which one hour-plus sessions are displaying markers of harm;
• The times and days when risky sessions occur;
• Risk prevalence by game.
The results and full report are available for the industry to view here.