S&P 500 Momentum Stock Screener

A Python-based stock screening and backtesting engine that identifies high-performance momentum stocks within the S&P 500 universe.

Features

• Momentum Screening: Identifies stocks that outperformed the S&P 500 in each of the past 5 years
• Weighted Ranking: Scores stocks using a weighted momentum formula (40% most recent year, 30% each for prior two years)
• Survivorship Bias Handling: Uses historical S&P 500 constituent data to avoid look-ahead bias
• Efficient Data Fetching: Multithreaded batch fetching with parquet-based local caching
• Portfolio Movement Tracking: Daily portfolio values with peak and max drawdown annotations
• Backtesting Engine: Single-year and multi-year backtests with performance comparison charts

Installation

cd /path/to/sp500-momentum-screener
python -m pip install -e .

For test dependencies:

python -m pip install -e ".[test]"

Usage

Screen for Current Year Portfolio

python main.py --mode screen
# or, after installation:
sp500-momentum --mode screen

Identifies the top 10 momentum stocks for the upcoming year based on the past 5 years of performance.

Run Historical Backtest

python main.py --mode backtest --start-date 2024-01-01 --transaction-cost-bps 5 --slippage-bps 5

Tests the strategy starting from a specific date, using 5-year lookback for stock selection. Backtest runs write a deterministic JSON manifest under output/ (or to --manifest PATH) with parameters, code version, output file checksums, selected stocks, and summary returns.

Multi-Year Backtest

python main.py --mode multi-backtest --years 2016 2017 2018 2019 2020 2021 2022 2023 2024 2025

Runs backtests across multiple years and generates a comparison chart.

Portfolio Movement Analysis

python main.py --mode movement-analysis --years 2020 2021 2022 2023 2024 2025

Generates daily portfolio movement charts with peak return and max drawdown markers for each year.

Command Line Options

Option	Description	Default
--mode	Operation mode: screen, backtest, multi-backtest, movement-analysis	Required
--start-date	Strategy start date for single backtest (YYYY-MM-DD)	-
--years	Years for multi-year analysis	-
--lookback	Number of years to look back for stock selection	5
--top-n	Number of top stocks to select	10
--workers	Number of parallel workers for data fetching	10
--transaction-cost-bps	Per-side transaction/commission cost applied to backtest entry and exit	0
--slippage-bps	Per-side slippage estimate applied to backtest entry and exit	0
--manifest	Optional path for the reproducible backtest manifest JSON	output/*_manifest.json
--no-show	Save charts without displaying	False

Project Structure

SP/
├── main.py                    # CLI entry point
├── pyproject.toml             # Packaging, dependencies, CLI entry points, pytest config
├── requirements.txt           # Legacy dependency list
├── README.md                  # This file
├── sp500_momentum/
│   ├── __init__.py
│   ├── constituents.py        # S&P 500 historical constituents
│   ├── data_fetcher.py        # Multithreaded data fetching with caching
│   ├── strategy.py            # Momentum filter and scoring logic
│   ├── backtester.py          # Backtesting engine
│   └── visualization.py       # Chart generation
├── cache/                     # Cached stock price data (parquet files)
├── data/                      # S&P 500 constituents cache
└── output/                    # Generated charts and CSV files

Testing and CI

python -m pip install -e ".[test]"
pytest

GitHub Actions runs the test suite on Python 3.9 and 3.11 for pushes and pull requests.

Strategy Logic

1. Universe Selection: Get S&P 500 constituents at the start of the lookback period
2. Filter: Identify stocks that beat the S&P 500 index in ALL 5 years
3. Score: Calculate weighted momentum score:
   • Year T-1: 40% weight
   • Year T-2: 30% weight
   • Year T-3: 30% weight
4. Rank: Select top 10 stocks by momentum score
5. Hold: Equal-weighted portfolio for 12 months

Performance (10-Year Backtest: 2016-2025)

These figures are a historical backtest over a specific period, not a live trading claim. Treat them as a research artifact to review for data quality, transaction costs, survivorship handling, and regime sensitivity before drawing any investment conclusion.

Metric	Value
Average Portfolio Return	20.00%
Average Benchmark Return	14.09%
Average Alpha	+5.91%
Win Rate	70% (7/10 years)

Important Caveats

• This project is for research and education only; it is not investment advice.
• Yahoo Finance data can be revised, delayed, missing, or adjusted differently over time. Re-run manifests record parameters and output checksums, but they cannot guarantee an external data vendor returns identical data in the future.
• Historical S&P 500 constituents are used to reduce look-ahead/survivorship bias, but constituent data quality still matters and should be audited before relying on results.
• The default backtest cost assumptions are zero. Use --transaction-cost-bps and --slippage-bps to model entry/exit frictions; taxes, borrow fees, market impact, liquidity limits, and corporate-action edge cases are not modeled.
• The strategy selects stocks using completed lookback years ending before the strategy start year. Tests cover the basic no-look-ahead date windows, but they are not a substitute for independent validation.
• Past performance and parameter sensitivity do not imply future returns.

Dependencies

• yfinance: Yahoo Finance API for stock data
• pandas: Data manipulation
• numpy: Numerical operations
• matplotlib: Visualization
• pyarrow: Parquet file support for caching
• requests: HTTP requests
• tqdm: Progress bars
• lxml: HTML parsing for Wikipedia data
• scipy: Statistical calculations for out-of-sample validation

License

MIT License