FuerteFund

Inspiration

According to the social security administration, only 33% of Hispanic-Latino ages 55 to 75 retired, compared to 71% of the total population. This is in part due to a variety of financial and cultural factors, including "more than two-thirds of all Hispanic workers not using workplace retirement-savings accounts like 401(k)s" (Boston College).

What it does

Our project aims to help minimize this problem by providing a smarter and easier way to invest long-term. We introduce Fuerte Fund, a fund geared towards retirement savings. It uses machine learning and alternative data to maximize monetary growth and minimize risk. We analyze a dataset of senator’s trade filings, and relevant senator information to cherry pick the best trades, based on how well our ML model predicts its ability to beat the S&P500 over a 90 day period. The trading strategy then naively buys and sells the stock from the day after the filing date to 90 days later. This algorithm can be used to make money on the market to generate retirement savings. In a backtest spanning a total of 2.9 years, we were able to achieve a Cumulative Annual Growth rate of 15.69%. Compared to SPY, an ETF that tracks the S&P500, which performed at 10.02%. Here are the data visualizations, which show cumulative return, drawdowns, and a distribution of returns over the backtesting period.

How we built it

We trained an XGBoost model over 8 years of senator trading data and backtested over the following 2.9 years.

Challenges we ran into

Our trading strategy started with making a loss consistently despite being trained on the senator’s trade data. As we tweaked and pruned our feature set the CAGR plateaued around 6%, almost half the SPY benchmark we wanted to overcome.

Accomplishments that we're proud of

It wasn’t until we started weighting our trade amounts by the model’s confidence metric as well as the inverse volatility that our CAGR shot up to 15% on the validation backtest.

What's next for Fuerto Fund

In the future, we plan to use this ML classifier as an element of a more holistic approach, leveraging Deep Reinforcement Learning algorithms such as Soft Actor Critic to manage our portfolio and develop more complex trading strategies.