AI-Powered Personalization Engine for Sports Betting

Problem

Our client, a leading iGaming operator, offers a vast and ever-changing landscape of sports events and games for players to bet on. While this extensive offering is a major feature, it also presents a significant challenge: players can become overwhelmed by the sheer number of choices, leading to decision fatigue. They struggled to quickly find events and games that matched their interests, which negatively impacted the user experience.

The client needed a way to cut through the noise and deliver a personalized, streamlined journey for each player.

Specific Pain Points:

• Players found it time-consuming to manually browse thousands of events to create a betting ticket.
• The vast selection paradoxically made it harder for users to engage, as they didn’t know where to start.
• New or infrequent players had no clear starting point, making it difficult to convert them into active users.
• The platform needed a way to proactively showcase relevant betting opportunities to improve engagement and satisfaction.

Solution

Description of the Solution:

We developed a sophisticated Bet Recommender system that delivers personalized game and event suggestions directly to the player. The system analyzes individual player data to understand their unique preferences, betting history, and behavior. Based on this analysis, it recommends events, specific games, and even entire pre-filled betting tickets with multiple selections.

This approach transforms the user experience by making ticket creation a seamless and engaging process. For new players or those who haven’t been active recently, our system addresses the “cold start” problem by generating recommendations based on the betting patterns of similar, more established players. This ensures that every user, regardless of their history, receives relevant and engaging suggestions from the moment they log in.

The result is a dynamic, personalized betting experience that feels curated for each individual, making it faster and easier than ever for players to go from Browse to placing a bet.

Unique Value Proposition:

The Bet Recommender doesn’t just suggest single bets; it intelligently constructs entire betting tickets tailored to individual habits. By solving the “cold start” problem, it guarantees a personalized experience for 100% of the user base, dramatically streamlining the betting process and enhancing player engagement from day one.

Results

Key Metrics:

• 80% of personalized recommendations are matching with what players ultimately decide to play.
• Ticket creation time has been cut by 40% to 80%, depending on the features used by the player.
• Significant improvement in overall player experience and platform engagement.

Smart Tip

When designing a recommendation system, make sure you identify as many key parameters as possible that humans use when making decisions. The closer your model can get to mirroring human logic, the more accurate and trusted its recommendations will be.

Smart Fact

The recommendation engine calculates a new set of personalized recommendations 12 times a day for over 400,000 players, ensuring suggestions are always fresh and aligned with the current live sports betting offer.

About the Clients

The client is a leading iGames operator in the Balkan region, with a significant and expanding footprint in Africa and South America. They are focused on leveraging cutting-edge technology to deliver a superior and engaging platform for their growing international user base.

Technologies Used

• Databricks: Used as the core platform for data engineering and collaborative data science, allowing for massive-scale data processing.
• XGBoost: The underlying machine learning framework used to train the predictive models that power the recommendation engine.
• PySpark / Spark: Enables scalable and efficient processing of large datasets for feature engineering and batch inference in the recommendation pipeline.