backtest.ai

LLM Reasoning for Pepsico franchise company on stock split

Inspiration

Everyone loves money and some find grow their wealth using capital markets either for short or long duration. But identifying the particular stock for short term or long term gains is very risky and cumbersome. In Peter Lynch's One up on the Wall street he highlights that one should prefer the companies that constantly buy back their shares. So i thought to myself that what if we have a tool which calculates the buyback/bonus/split returns of the company in short-long term view and provides llm reasoning on the trade/investment idea.

For now this tool covers the listed companies in india only. But have future plans to expand this to all capital markets.

What It Does

backtesting.ai provides insights into stock performance based on corporate actions such as buybacks, bonuses, and splits. It calculates returns over short and long-term periods and offers reasoning for investment decisions through natural language processing.

How we built it

India's stock market corporate actions data is not readily available, so i have written a script that will extract the corporate actions for given company and stream it to redpanda connector and produce the messages and on the consumer side, we write the messages to supabase hosted postgres database.

Python app uses the llm to identify the stock ticker and corporate action with proper handling. Once identified, it generates returns on short & long-term time frames. These return datas are given to llm to provide reasoning on the hypothesis testing of taking particular trade make sense based on the corporate action.

Challenges We Ran Into

Lack of readily available corporate actions data for Indian stocks.
Implementing a robust data streaming solution with Redpanda.
Ensuring accurate identification and handling of stock tickers and corporate actions.

Accomplishments That We're Proud Of

Successfully integrated Redpanda for real-time data streaming.
Developed a functional tool that analyzes corporate actions and provides actionable insights.
Created a natural language backtesting platform that facilitates hypothesis testing for investment strategies.

What We Learned

I gained valuable experience in data streaming with Redpanda, which is crucial for managing stock market data. Additionally, I learned about building a natural language backtesting platform capable of testing corporate actions on specific companies.

What's Next for backtesting.ai

Future plans include expanding backtesting features across various areas of the capital markets, enhancing the tool's capabilities, and broadening its reach to more companies and markets globally.