Historical Data for Backtesting

The quality of your backtest is bounded by the quality of your data. Bad data produces bad conclusions — sometimes catastrophically wrong ones. Before running a single test, understand what data you have and where it came from.

For US stocks, Yahoo Finance is the standard free source and works for basic backtests. For higher-quality data with proper corporate action adjustments, look at Norgate, Polygon.io, or Alpaca. For institutional-grade data, CRSP or Refinitiv are the gold standards.

Survivorship bias is the biggest hidden trap in stock data. If your dataset only includes companies still trading today, you've accidentally filtered out every company that went bankrupt or got delisted. This inflates historical returns significantly. Always use point-in-time or delisted-inclusive datasets.

Options data is far more expensive and complex. Free sources are almost nonexistent for historical options with full Greeks. Serious backtesting requires paid data from CBOE DataShop, OptionMetrics, or ORATS — expect to pay $500-$5,000 per year for a decent dataset.

Crypto data is generally free and readily available from Binance, Kraken, and CoinGecko APIs. However, watch for missing data during exchange outages, and remember that the crypto history is short compared to stocks — most strategies only have 5-10 years of testable data.