◾Sport Recommendations
Last updated
Last updated
The Sport Prediction AI feature is designed to recommend the most relevant sports events tailored to each player. By analyzing past activities, demographic information, and upcoming sports schedules, this AI-driven system predicts events that align with player interests, enhancing engagement and personalization.
To enable the recommendation service, your platform must provide:
The platform must supply information about all player sports bets, including the IDs of the sporting events (commonly referred to as "Sport Event ID" or "Sport Match ID")
The platform must provide a feed of upcoming sports events, where each event has "Sport Match ID", Sport type, league, teams, and date of event
These inputs are crucial for analyzing player betting activities and building personalized rankings of future sports events for each player.
The model setup is managed by Smartico personnel and requires the following configurations:
User Segment: Define the segment of users for whom the model will be evaluated (e.g., users with at least one sports betting activity).
Included Sports Types: Specify the types of sports to include in predictions.
Re-evaluation Periodicity: Determine how often the model should update predictions (e.g., nightly).
Time Window for Future Events: Set the range of future events to be ranked for each player (e.g., the next three days).
Top Matches to Keep: Define the number of best matches to retain for each player during prediction (e.g., top 20).
Historical Bets for Prediction: Specify the number of recent bets to use for analysis (e.g., last 50 bets).
At a specified time (e.g., 3 AM), the model identifies all players meeting the segment criteria.
It considers future sports events based on the defined sports types and time window (e.g., 3,000 events over the next three days).
For each player, the model ranks these events using their last 50 sports bets.
Only the top 20 events with the highest matching scores are retained and can be used in communication with the player
You can use "Visual widgets", that are based on the Liquid templating engine and build an HTML presentation of the top matches for each player. This widget can be placed on the website or used in a mail or popup. The widget will render dynamically the content for each player.
The "Model Performance" tab provides insights into prediction accuracy - metrics show the percentage of players for whom the model correctly predicted top matches (e.g., top 1, top 3, top 5, or top 10 matches).
Example: A value of 45% for the top 3 matches means that 45% of players placed a bet on at least one of their top 3 predicted matches.
