Multi-Factor Momentum Trading Strategy

Overview

This repository contains a sophisticated trading strategy implemented in Python using the backtrader framework. The strategy is designed to achieve robust portfolio performance by integrating multiple modern quantitative concepts. It combines a top-down market regime filter with a bottom-up, multi-factor stock selection process, and uses a risk-parity approach for capital allocation.

The strategy is long-only and dynamically rebalances its holdings based on changes in the top-ranked stocks, ensuring it adapts to evolving market conditions.

Key Ideas and Concepts

The strategy is built upon a foundation of several key quantitative finance ideas:

1. Market Regime Filter: The strategy first assesses the overall market health using a simple but effective filter.

   • Mechanism: It checks if the market index (e.g., SPY) is trading above its 200-day Simple Moving Average (SMA).
   • Action: Long positions are only initiated or held if the market is in a bullish regime (price > 200-day SMA). If the market enters a bearish regime, all positions are exited to mitigate systemic risk.

2. Time-Series Momentum (TSMOM): This is a trend-following filter applied to each individual stock.

   • Mechanism: A stock is only considered a candidate for investment if it is trading above its own 200-day SMA.

3. Cross-Sectional Momentum: After filtering for trend, stocks are ranked against each other based on their momentum.

   • Mechanism: Momentum is calculated over multiple lookback periods (60, 120, and 252 days) and then averaged to produce a robust momentum score.

4. Frog-in-the-Pan (FIP) Score: This factor measures the quality and consistency of a stock's trend.

   • Mechanism: It calculates the ratio of positive daily returns over a 252-day period. A higher score indicates a smoother, more consistent uptrend, rather than one driven by a few large, volatile spikes.

5. Skewness Factor: To mitigate the risk of holding stocks prone to sudden crashes, a skewness factor is included.

   • Mechanism: It calculates the rolling skewness of stock returns over a 90-day window. A high negative skew suggests a stock has a "tail" of large negative returns and is thus penalized in its final score.

6. Inverse Volatility Weighting: This is a risk parity approach to portfolio construction.

   • Mechanism: Once the top N stocks are selected, portfolio capital is allocated inversely proportional to each stock's historical volatility (standard deviation over 126 days). Less volatile stocks receive a larger capital allocation, and more volatile stocks receive a smaller one, with the goal of equalizing the risk contribution of each position.

Important Formulas

• Daily Return:
  rt = (Pt / P_{t-1}) - 1

• Momentum Score (Average over Lookbacks):
  Momentum = (1/N) * Σ Momentum_lookback_i

• Frog-in-the-Pan (FIP) Score:
  FIP Score = (Number of Positive Daily Returns) / T

• Rolling Skewness:
  Skewness_t = skewness(r_{t−τ+1}, ..., r_t)

• Combined Ranking Score:
  Score = (w_m * Momentum) + (w_f * FIP) - (w_s * Skewness)

• Inverse Volatility Weighting:
  w_i = (1 / σ_i) / Σ (1 / σ_j)

Performance Summary

Over a 2-year backtest period, the strategy demonstrated strong growth in both held positions and realized trades:

• Unrealized Profits (Total Portfolio Growth): Up by 89.71%
• Realized Profits (from Closed Trades): Up by 73.93%

Usage

The strategy is implemented as a Python script using the backtrader library.

Requirements

• Python 3.x
• backtrader
• pandas
• numpy
• scipy
• yfinance

How to Run

1. Ensure all required Python libraries are installed:

   pip install backtrader pandas numpy scipy yfinance

2. Execute the main Python script from your terminal:

   python main.py

3. Upon completion, the script will generate a detailed Excel report named trading_results_final.xlsx containing multiple sheets for in-depth analysis.

Notes

• Dynamic Rebalancing: The strategy does not use a fixed rebalancing schedule. Instead, it rebalances the portfolio dynamically whenever the composition of the top N ranked stocks changes.

• Indicator Warm-up: The strategy relies on several long-term indicators (e.g., 200-day SMA, 252-day FIP). As a result, it requires a significant "warm-up" period (at least one year of data) before it can generate reliable signals and begin trading.