Featurespace report: Secondary analysis of machines data


Executive summary :A consortium of NatCen Social Research, Featurespace, Geofutures and RTI International conducted the Responsible Gambling Trust’s (RGT) Machines Research Programme in 2014. A number of useful datasets were created for this project and Featurespace was asked to answer the following research questions in relation to this data:

- RQ5: Can the range of linked data set variables be examined through a process of ‘reverse engineering’ to explore whether any other variables might play a useful role within the development of algorithms?
- RQ7: What are the differences in demographics between B2/B3 players? What else can we learn about players’ transitions between B2 and B3 content?
- RQ8: What further descriptive data can be extracted about the £100 stake?
- RQ9: What are the differences in behaviour when players are spending wins vs loading their own new money into the machine?

In this report we have answered each of these questions and presented the results as four independent chapters. The research has been conducted from what the data has told us from the 4,000 players who were surveyed as part of the 2014 research project. These results are not intended to inform a general understanding of the extent of problem gambling against different factors of gaming machine activity as the survey population biases have not been factored into the presented figures.

However, from the research we have been able to identify the following key findings:
- The accuracy of the problem gambler identification model developed in the 2014 research has been improved. This was principally achieved by including a new marker of harm that measured the diversity on money loaded and money spent by the player.
- The most distinct identifiers of problem gamblers are their chaotic behaviours and the fact that on average they are more successful when playing (they win more often, have higher return rates and more often have winnings to spend).
- Transitions between B2 and B3 bets are not useful when it comes to differentiating between problem and non - problem gamblers.
- Players who place £100 bets are distributed uniformly across problem and non - problem gamblers within the surveyed data. However, player s with 100 or more £100 stakes are more likely to be problem gamblers within the data set.
- A typical £100 stake scenario is one where players place the maximum bet several times during a session, it is rarely an isolated single event. £100 stakes happen very rarely at initial stages of sessions and become more common at later stages.
- Variable and intensive activity at early stages of sessions often leads to £100 being staked later.
- When it comes to the differences between playing with winnings and with the player’s own money, in the former case players tend to bet higher amounts of money and withdraw money more often. In the latter case on the other hand, players are loading money more often and spending more as a percentage of the balance.

Problem with this document? Please report it to us.