backtesting-frameworks

Avoid look-ahead bias by using point-in-time data and shifting signals before applying them to returns. In a vectorized backtester, shift signal series by one period so decisions only use information available at that time.

Walk-forward analysis splits historical data into sequential train and test windows, optimizing parameters on each training window and validating on the following unseen test window. It supports anchored or rolling windows to simulate realistic strategy re-optimization over time.

Use a vectorized backtester for fast evaluation of simple signal-based strategies on large datasets. Choose an event-driven backtester when you need realistic order handling, position tracking, limit orders, and detailed portfolio accounting.

Model transaction costs by applying per-share commissions and slippage in basis points to each fill. In vectorized approaches, multiply position changes by the combined commission and slippage rate and subtract from strategy returns.

This usually results from overfitting parameters to historical data, ignoring transaction costs, or survivorship bias in the dataset. Use out-of-sample testing, walk-forward validation, and Monte Carlo simulation to estimate realistic performance distributions.