Background

In 2016, 59.3 million people in the US and Canada made Fantasy Sports a $7 billion industry. Despite this popularity, winning big in Fantasy Sports is not easy. A 2015 study of Daily Fantasy Sports found that the top 1% of players took home over 90% of the payouts. Clearly, winning strategies rely on more than luck and intuition. We believe that top players utilize sports data analytics to place informed bets and increase their chance of winning.

To test and capitalize on this theory, we narrow the scope of this project to one particular type of fantasy sports league: a Daily Fantasy Basketball Roster Contest (subsequently referred to as a contest). In a contest, participants are given a fixed budget to draft a roster of active NBA players whose prices are predetermined by the betting sites. These rosters are entered into a pool with hundreds or thousands of other players' rosters, with each roster's performance determined by its players' scores in a set of predefined categories (points, rebounds, etc.). The owners of the top-performing percentage of rosters, usually 5–10%, are declared the winners, and receive large payouts relative to the entry fee.

In this form of fantasy game, we have identified two specific areas for optimization. The first is a modification of the knapsack problem. In our implementation, we seek to maximize player value while maintaining a total player cost that is less than or equal to our budget. The second is a version of the picking winners problem. Here, we seek to minimize the covariance between generated rosters, reducing our exposure to the high variability characteristic of sports leagues. A web application will wrap the optimization model, allowing users to input custom settings and generate their own rosters.

We hope to supersede the many existing services that provide similar analysis for high annual premiums. We will initially offer our service for free, and as it grows, perhaps transition to a freemium model.

Project Summary

There are two main deliverables in our project. The first is an analytics model which can make recommendations to daily fantasy sports players to maximize their expected profits. The second is a web application which integrates our model with other third-party projections and displays all information in a user-friendly interface.

Fall 2017 Semester Report

Built With

Share this project:
×

Updates