The New York Times reports that DraftKings built a 2023 machine-learning model to score customers by expected losses after receiving promotional bets. Former employees say a parallel effort to develop risk scores for problem gambling was shelved, despite the infrastructure existing to support both approaches. DraftKings denies unfair targeting, noting it monitors risky behavior and cites a 13% boost to promotion-driven margins in 2025 and extensive AI-driven personalization. The story raises regulatory and ethical questions about whether promotions might disproportionately reach the most vulnerable players.
NYT Report: DraftKings Used Machine Learning to Direct Bonus Bets Toward Customers Predicted To Lose Most
Image: Google Play
A New York Times investigation found that DraftKings built a machine-learning model in 2023 that ranked customers by how much money they were expected to lose after receiving a free bet or promotional credit. According to former employees and the report, the same data architecture could have been used to flag customers at risk of gambling harm — but a parallel project to create protective risk scores was reportedly shelved.
How the Model Worked
The model reportedly combined variables such as play frequency, account balances, loss-to-wager ratios and a separate projection of whether a customer was likely to stop gambling. A higher score indicated a customer was expected to generate more revenue per promotional dollar spent. Jayden Butts, a former DraftKings data analyst, told the Times the team looked for "traits and features that we can target that indicate a good investment."
"The best investment would be a problem gambler," the report said encapsulates the implication drawn by some former staffers.
Company Response and Context
DraftKings told the Times it "rejects any implication" that its marketing unfairly targets customers and said promotions are directed to users who show sustained, engaged platform use rather than people ranked by predicted losses. The company also said data science and analytics improved promotion-driven sportsbook margins by 13% in 2025 and that AI helped personalize hundreds of millions of dollars in promotional spending.
DraftKings's chief responsible gaming officer, Lori Kalani, told the Times the company monitors customers for risky behavior and that it declined to deploy a risk-prediction tool because it was not yet sufficiently evidence-based. Publicly, DraftKings highlights responsible-gaming initiatives such as a collaboration with Mindway AI's Gamalyze, expanded customer education resources, and a partnership with IC360 for integrity and compliance monitoring.
Implications and Questions
Former employees said the technical infrastructure existed to score customers both for revenue potential and for gambling risk, but only the revenue-focused model was completed. Observers have raised alarms that covert digital profiling without clear notice or consent is becoming more common across industries.
Whether audiences labeled "engaged users" and those predicted to lose the most are effectively the same group depends on how the model weights its variables — a question the Times report does not conclusively resolve. If regulators and lawmakers verify the reporting, they would have a clearer, data-specific basis to examine marketing practices and protections for vulnerable customers. A free bet in your inbox might, in some cases, reflect a data-driven judgment about your propensity to lose rather than a simple loyalty reward.
What Remains Unclear
- How often the model was used to direct specific promotions and to what extent it influenced campaign decisions.
- Whether the company's public responsible-gaming tools sufficiently address the gap identified by former employees.
- What internal evidence led DraftKings to shelve the risk-prediction project, beyond concerns about available evidence.
Help us improve.
