HIKE: Fantasy Sports Assistant

---

Over fifty million people in the United States play fantasy sports every year. Yet, they are not able to harness the beautiful power of AI. That is, until now.

With HIKE, an AI-powered chatbot for fantasy sports advice, fantasy sports players can receive educated, accurate advice about anything from trades to player statistics. HIKE is not restricted to the domain of one sport—this chatbot will take you to the top of your league in both fantasy football and basketball.

Why HIKE?

What sets HIKE apart from other options on the market? Well, other fantasy trade analyzers are strictly data-driven, limited to trade analysis, and have outdated data. When determining whether you should trade a player, you can’t simply rely on numbers, as there are other factors to consider that establish whether a player has good trade value. Other fantasy trade analyzers on the market lack sentiment and they often have outdated data, causing them to occasionally recommend detrimental trades. Furthermore, they only give information about trades, so they do not give you information on how a player is growing and declining, or how injuries come into play.

However, HIKE has conversational guidance, real-time data updating, and multi-sport capability. The conversational nature of this chatbot makes it more user friendly compared to the other options on the market, as it allows users to receive personalized advice for their fantasy teams, rather than only receiving an output of numbers or a one word response like "yes" or "no." Additionally, this AI-powered fantasy sports chatbot is the first of its kind to use both score projection and injury analysis, as no other fantasy trade analyzers on the market utilize both of these options.

Development

To retrieve data for the model, the project members collected data from NBA.com and Kaggle datasets for the NFL and made a Python script that converted all of this data into CSV files using pandas, a software library for Python. The same CSV files were put through JavaScript Object Notation (JSON) to help with the sentiment analysis aspect of the chatbot. They also used statsmodels—a Python library that allows users to perform advanced statistical testing and modeling—to implement time series analysis, which helped forecast the growth, trends, and tendencies of the various players in the NFL and NBA.

They used a large language model (LLMs) and utilized LangChain, which is a framework built around LLMs that makes it easy to connect other data sources to LLMs. LangChain allowed all of the data that was retrieved and processed to be sent to the models along with the question that the user asked to HIKE to get an answer to the question. As for the pipeline optimization, they used Retrieval-Augmented Generation (RAG), where the chat history, question, and data were sent through the LLM in the form of vectors to eventually get the accurate, helpful answer to the user’s question. To ensure fast deployment, they used Streamlit for the front-end, which eliminated the need for communication between the front-end and back-end.

Areas of Improvement

Despite clinching first place at the Aggie Sports Analytics Winter Quarter Case Competition, the team recognizes that there is still room for improvement and remains committed to refining HIKE. In the future, they look forward to expanding the scope of their chatbot to more sports, such as soccer and baseball, and to collaborate with big corporations to gain more outreach for their project.