Quantitative Trading Strategy - Low Volatility Factor

FINS5546 asked us to build and evaluate a real trading strategy from scratch. Our team - Well Done - tested the low-volatility anomaly on NASDAQ's top 50 by market cap: the documented tendency for low-volatility stocks to outperform high-volatility ones on a risk-adjusted basis.

What I built: a full Python pipeline with an ETL module for cleaning raw price data, a characteristics module for computing monthly volatility signals, and a portfolio module handling the sort, weighting, and rebalance logic. Long the bottom volatility tercile, short the top, equal-weighted, rebalanced monthly.

What I learned: quantitative strategy construction is as much about the plumbing as the signal. The data cleaning decisions, the rebalancing logic, the outlier treatment - these implementation choices have as much impact on the backtest as the signal itself. Seeing how dramatically performance changed with and without IQR/Z-score outlier treatment was a reality check on how fragile a supposed "edge" can be.

The hypothesis testing component gave me a framework I've used in every analytical project since: run a formal test, set a null, and be honest about whether your result is signal or noise. The question is always the same - is this real, or is it a product of the specific data I happened to use?

Key knowledge gained: a strategy that looks clean in theory can behave very differently once you add real data problems. The implementation is the strategy.