Realistic backtesting, not just raw price math

What’s configurable

• Instrument(s), strategy, timeframe and historical period
• Starting capital and position-sizing rules
• Brokerage, transaction fees, taxes/levies where applicable
• A slippage model and spread assumptions
• Margin/leverage rules where applicable

Why it matters

A backtest that ignores fees, slippage and realistic fills will overstate performance. my ALGO agent’s backtesting engine applies the same cost assumptions a live strategy would face, so results are a more honest estimate of what actually happened.

Bias prevention

• No look-ahead bias — a strategy can only act on data available at that point in time
• No future-data leakage into indicator calculations
• Correct handling of indicator warm-up periods
• Trading-session and calendar awareness
• Explicit handling of missing or invalid market data

Metrics reported

• Total return, absolute P&L, CAGR where meaningful
• Win rate, loss rate, profit factor, expectancy
• Maximum drawdown, average drawdown, recovery factor
• Sharpe, Sortino and Calmar ratios, volatility
• Trade count, average holding period, largest win/loss
• Monthly/yearly returns and performance by instrument

Reproducibility

Every backtest stores its exact configuration alongside its results, so results can be reproduced and audited later — an important property when a strategy graduates to paper or live trading.

Important disclosure

A backtest describes how a strategy would have performed on historical data under the modeled assumptions. It is not a guarantee of future performance. Markets change, and live execution can differ from simulated fills.