Amplifier AI personalises bet recommendations by analysing player behaviour, preferences and betting patterns to deliver relevant and timely recommendations.
Amplifier AI generates personalised bet recommendations by combining two key data sources: a players preferences and a structured breakdown of upcoming betting events. These inputs are scored, ranked and matched to deliver relevant and real-time recommendations.
Modelling User Preferences
To understand what each player cares about most, the system analyses their historical betting activity and builds a profile based on preferences across:
- Sports
- Competitions
- Markets
- Teams, Players or Athletes
These preferences are dynamic and time-sensitive, with more recent activity given greater weight. This ensures recommendations remain aligned with each players current interests and betting habits.
Breaking Down and Scoring Events
Each upcoming betting event is also broken down into a consistent structure:
- Sport → Competition → Event → Market → Selection
- Example: Football → Premier League → Man Utd vs Man City → 90 Minutes → Man Utd
The system then calculates individual scores for each component using a combination of frequency (how often the player has bet on similar components) and recency (how recently they did so). This approach ensures that the system promotes events where the player has shown interest, while filtering out less relevant or expired events. It also automatically cleans up outdated scores.
These scores are then used to personalise content; such as promotional messages, campaigns or lobby layouts – based on a players specific interests. They also form the basis for our next feature: single bet recommendations.
Single Bet Recommendations
This feature operates in three stages:
1. Event Curation and Scoring
We begin by curating a set of upcoming events available for betting.
The sports betting market follows a long-tail distribution, where a few events attract the majority of bets. To maintain quality and performance, we filter out the events with very low traffic from the lower end of the distribution. Additionally, we boost the score of high-volume events to reflect popularity.
The result is a curated and scored set of future betting opportunities.
2. Analysing Bet Slip Structure
Simply ranking events by score can lead to overrepresentation of a players top sport or league. For example, a football-focused player might be shown only Premier League events, ignoring their other interests.
To prevent this, we analyse the players historical bet slip composition across various timeframes. From this, we infer an ideal bet slip structure, which reflects how the player typically distributes bets across sports, competitions and markets. This ensures recommendations maintain a balance of familiarity and variety.
3. Ranking and Filling the Ideal Bet Slip
Using both the component scores and the ideal bet slip structure, we rank the curated events for each player. We then fill the ideal slip with the top-matching events.
This process generates a personalised list of the top N recommended competitions, events, markets and selections per player, per sport. By treating each sport separately, the system supports both global betting preferences and sport-specific recommendations.
We also calculate a probability distribution of how each player spreads their bets across sports. This helps maintain sport diversity in broader views, such as homepages or landing pages.
Summary
Our recommendation engine delivers personalised betting experiences across every level, from sport and competition to market and selection, tailored to each player. By balancing personal preferences, event popularity and variety, it creates a smarter and more engaging experience for everyone.