DGOJ, Spain Directorate General of Gambling believes that in the coming months it will conclude an imperative stage of the overhaul of the compliance and governance of Spain’s online gambling sector.
Last week, at the Gaming in Spain Conference in Madrid, DGOJ Director General Mikel Arana chastised the regulator’s detractors, standing by its track record on delivering effective protections and standards to Spain’s gambling market.
Arana decried that critics should take a long lense to appraise the changes brought on by the DGOJ under the mandate of Spain’s Ministry of Consumer Affairs, as the overhaul of Spanish gambling is a regulatory project which has evolved since 2020.
The DGOJ has been dogged by criticism of stalled and haphazard implementation of the mandates carried by the Royal Decree on Gambling Advertising and the Royal Decree on Safer Gambling Environments.
Arana rejected those criticisms, stating that the DGOJ had delivered three of four key projects, now in the testing phase ahead of their intended application over the next six months.
He underscored that these were substantive, transformative ‘first-of-a-kind’ projects, with ambitions to establish European precedents requiring extensive development and testing.
In the coming months, the adoption of central deposit limits takes priority, yet the DGOJ underscores the application of XGBoost and maching learning algorithm that can detect 48 behavioural vulnerabilities as the showpiece of its five-year overhaul of Spanish gambling.
On the sidelines experts await for the final delivery of the algorithm of detection that may prove the DGOJ’s most quixotic moment. Has Spanish gambling’s bold and idealistic knight found its key weapon to slay giants or is it just fighting windmills once more…
No independent expertise
Pedro Romero, Chief of Safer Gambling Partnerships at BetBlocker, questioned the approach taken to develop and the algorithm and its delivery onto licences.
“It was a real surprise for me and others. A project this big and complex with no tender for partners or advisors on the design of the algorithm.
“There are many independent actors who would have helped with the most problematic aspects of customer detection and intervention. I don’t know why they have chosen to build in secrecy.”
He called for more dialogue and the need for collaboration, warning that regulatory decisions could be reviewed and taken from approaches elsewhere without fully understanding the elements needed to make them work.
“They may make a decision based on what other regulators are doing, and they may do a version of what they see somewhere else, but perhaps I feel there are some pieces missing,” Romero said. “I think it’s really important to have dialogue and to listen to different points of view.”
Traffic Lights
Paula Murphy, of Mindway AI, said there had been some meetings with the DGOJ but had not contributed any input to the algorithms design or parameters. The DGOJ had launched just one public consultation on the proposed detection mechanism on 13 October, with responses only available for submission by 22 October.
Murphy’s concerns extended beyond the short window as to how operators would be expected to act on the model’s findings.
From obscure feedback, Murphy understands the proposal envisaged customer interventions to be activated at intervals, potentially once a month, followed by a classification determining whether additional customer controls should apply.
“Players either get a green light or a red light,” which Murphy views as a simplistic diagnosis as “problem gambling isn’t binary. It isn’t either on or off. It exists on a scale.”
A blunt tool of faith…
Murphy argued that operators need to understand why a customer has been flagged and how their behaviour is changing.
That means identifying escalating activity, recognising where a player retains control and assessing whether particular products contribute to more problematic gambling.
“What is the trajectory of their play? Are they escalating? Are we seeing control in certain areas? Are we seeing that this player becomes more problematic when they choose a certain product type?” she asked.
“You need the analytics to understand what you’re seeing.”
Murphy described the proposed approach as a “very, very blunt tool”, questioning whether it reflected the depth of information already available through risk detection systems.
Mindway AI had integrated the DGOJ model into its GameScanner platform, she explained, enabling operators to run it alongside its own predictive algorithms. Data returned prompted questions from its data scientists and machine learning engineers about the model’s scope and training data.
“Everybody would find it helpful if there was a bit more transparency on what kind of players the model has been trained on,” Murphy bluntly concluded as the application of the DGOJ’s algorithm cannot be a “question of keeping faith”.
