Case Studies


Back Case Studies

Client: Online Sports Betting
Industry: Gambling
Length: 6+ months (ongoing)
Goal: Utilize a vast amount of user data, including past betting behaviour, in order to build a system that could generate recommended tickets for users.
Tech: Python, Spark


Recommending items to users is an important aspect of business, since it can improve user experience by tailoring site content and suggesting items that match the user’s taste. In case of online betting, this may increase user activity and lead to more bets. Our goal is to build a recommender that could suggest users to bet on games based on their past betting behaviour.


The project would be divided into multiple stages:

  • Phase 1: Create a prototype model for recommending bets for football.
  • Phase 2: Model improvement by using additional techniques, and extension to cover different sports. This phase is optional.
  • Phase 3: Integrate the model within the online platform.
  • Phase 4: Analysis of user reactions to given recommendations.
  • Phase 5: Improvement of recommender based on user reactions.


Historical betting dataset was used for modeling. The solution is based on classification - for each possible bet (pair user - match), we want to predict the probability that the user will bet on the match with a given outcome. After that, bets with the highest probability would be used to form a recommended ticket. Input for the classification model is a set of features capturing past betting trends. Technologies used include Spark in order to deal with large amount of historical data.


Recommender engine which uses historical data to recommend ticket to player. It works on two types of tickets, live bets and standard bets. Ticket with standard bets is preprocessed since it is daily offer tailored for particular user. Ticket with live bets is ever changing and works in real time based on preferences of particular user. Decision which ticket to use is done in real time based on players preferences.

Previous post Next post