Backtesting framework for SMA crossover strategies across S&P500 stocks. Built with Python, Pandas, Matplotlib, and yfinance.
| Ticker | Sharpe | Max Drawdown | Strategy Return | Market Return |
|---|---|---|---|---|
| META | 0.81 | -51.58% | +253.8% | +92.8% |
| NFLX | 0.54 | -41.43% | +96.8% | +89.9% |
| NVDA | 0.28 | -80.25% | +6.4% | +1073.3% |
| GOOGL | 0.22 | -57.57% | +10.8% | +160.8% |
| TSLA | 0.12 | -78.28% | -39.3% | +96.7% |
| JPM | 0.02 | -45.92% | -11.4% | +99.5% |
| AAPL | -0.18 | -55.16% | -34.7% | +103.9% |
| AMD | -0.26 | -87.56% | -74.6% | +152.5% |
| MSFT | -0.42 | -64.44% | -50.0% | +64.9% |
| AMZN | -0.35 | -60.81% | -59.5% | +30.2% |
Key finding: SMA 20/50 crossover outperformed buy-and-hold on 3/10 tickers (META, NFLX, NFLX). Underperformed on trend-dominated names like NVDA (+6.4% strategy vs +1073.3% buy-and-hold) due to whipsaw — the strategy kept switching sides on pullbacks during a sustained uptrend. Transaction costs and regime changes identified as primary performance drivers.
META — Strategy +253.8% vs Buy & Hold +92.8%
NFLX — Strategy +96.8% vs Buy & Hold +89.9%
├── fetch_data.py # downloads OHLCV data via yfinance
├── strategy.py # SMA crossover signal generation
├── metrics.py # Sharpe, max drawdown, cumulative returns
├── visualize.py # 3-panel chart: price/signals, equity curve, drawdown
├── main.py # orchestrates full pipeline
├── data/ # downloaded CSVs (gitignored)
├── charts/ # output PNG charts
└── results_summary.csv
pip install -r requirements.txt
# download data
python fetch_data.py
# run with defaults (10 tickers, SMA 20/50, 10bps cost)
python main.py
# custom tickers and SMA windows
python main.py --tickers AAPL TSLA --short 10 --long 30
# custom transaction cost
python main.py --bps 5yfinance>=2.0.0
pandas>=2.0.0
numpy>=1.26.0
matplotlib>=3.8.0