BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:What can gambling machine data tell us about betting behaviour? - 
 David Excell\, Featurespace
DTSTART:20170202T184500Z
DTEND:20170202T210000Z
UID:TALK64274@talks.cam.ac.uk
CONTACT:Peter Watson
DESCRIPTION:Bookmakers’ betting machines continually attract political a
 nd media attention over their impact on problem gambling. Featurespace and
  NatCen partnered with the Responsible Gambling Trust to investigate harmf
 ul patterns of play on gaming machines\, and draw implications for interve
 ntion.\nI’ll present the methods and results of this ground-breaking inv
 estigation\, linking industry-held data from the five largest UK bookmaker
 s with surveys of loyalty card customers (which measured customers’ Prob
 lem Gambling Severity Index score as a proxy for harmful play)\, integrati
 ng research methods for 10 billion individual gaming machine events. The d
 ata set included 6.7 billion bets and 333\,000 customers.\nWe harnessed th
 is huge data set to model actual gaming play\, measure theoretical markers
  of harm (e.g. faster gaming)\, survey loyalty card customers (matching 4\
 ,001 responses with transactional data) and explore consumer interventions
 .\nWe produced two predictive models\, exploring the statistical relations
 hips between the data and the customer surveys. The results showed it is p
 ossible to distinguish between problem gamblers and non-problem gamblers i
 n industry data:\n Player model: behavioural analyses in loyalty card h
 older data – 66% improvement over the current baseline model.\n Sessi
 on model: proxy measurements for anonymous players rather than individual 
 players –550% improvement in accuracy of detecting problem gamblers over
  the industry standard.\nThe research demonstrated that a combination of v
 ariables are needed to identify problem gamblers\, in contrast to proposed
  policy suggestions of regulating individual parameters (e.g. stake size).
 \nI’ll assess approach limitations\, including data skewedness\, and exp
 lore the challenges of incorporating big data into social scientific inves
 tigations.
LOCATION:Coslett Building\, Anglia Ruskin University\, Cambridge
END:VEVENT
END:VCALENDAR