Ready for the Big Game: How FanDuel Boosts Conversions and Revenue at Scale with a Real-Time Recommender System for Daily Fantasy Sports

Daily fantasy sports allow sports fans to participate in the matches they are watching. The daily fantasy sports landscape is competitive and evolves fast. To see off their competitors, some operators now personalize their offerings to improve the engagement of their users and increase their revenues. This paper presents the development and implementation of a scaled and real-time recommender system at FanDuel for its Daily Fantasy Sports (DFS) users. The system aims to recommend contests to users based on their past contest entries, leveraging historical data and machine learning techniques.In this paper, we propose a scalable contest recommender model that was built on real-time data pipelines and is re-trained every hour. We identify key factors that influence user preferences and contest selection. The recommender system analyzes users’ historical data, identifies similar users based on their participation behavior, and suggests contests that align with their interests. On average, we recommend contests to 180K users daily and have recommended over 300 million contests to 2 million users since the model’s inception. Notably, we saw over 1000% lift in user conversion rates and 4.43% lift in overall entry fees, peaking at 15.6% in April.