Inspiration
The inspiration for StatPilot came from a need to provide users with a more data-driven approach to decision-making. By utilizing player stats and historical data, we aim to predict a player's likelihood of exceeding or falling short of a specific stat line in a given game.
What it does
StatPilot is a web-based application that allows users to input a player's name, opposing team, stat category (points, rebounds, assists), and a proposed stat line. It then calculates the probability of the player surpassing or falling short of that stat line based on historical performance data, helping users make more informed betting decisions.
How we built it
We manually collected game log data from ESPN.com and exported it into a CSV file for processing. Using Python and the pandas library, we organized and cleaned the dataset, ensuring proper formatting and handling missing values. A linear regression model was trained on this data to predict the player's stats for an upcoming game. We used normal distribution to estimate the probability of going over or under the proposed stat line with weights that adjusts for three features.
Challenges we ran into
One of the challenges we ran into was when we had to change our entire project goal. Initially, we had the idea of using a single player's stat to predict if his team is likely to win or lose. But after finding out that the usage of a single player doesn't do the intended function accurately, we decided to shift our focus on what we can do with other players stats. From there, we came up with our presented project goal.
Accomplishments that we're proud of
We successfully built a functional web tool that can predict probabilities for different stat outcomes based on historical player performance data. We are especially proud of the accuracy of the linear regression model and how it allows bettors to use real-time player data to make informed decisions.
What we learned
We learned valuable insights into data cleaning, working with large datasets, and training machine learning models for sports analytics. Additionally, we gained experience in building web-based applications and integrating statistical models into a user-friendly interface.
What's next for StatPilot
Next steps include refining the model with more features like player health status, game context, and historical performance trends. We also plan to enhance the web tool with a more robust user interface and real-time data updates to increase prediction accuracy.
Log in or sign up for Devpost to join the conversation.