Backtesting

Every algorithmic trading strategy should be backtested before live deployment. Backtesting applies the strategy's rules to historical OHLCV (open, high, low, close, volume) data — and optionally tick-level data for more precise simulations — to generate a track record of hypothetical trades and performance metrics.

Key Backtesting Metrics for Indian Strategies

• CAGR (Compound Annual Growth Rate): The annualized return of the strategy over the backtest period
• Sharpe Ratio: Risk-adjusted return (return minus risk-free rate, divided by volatility). India risk-free rate ≈ 7% (RBI repo rate / G-Sec yield). A Sharpe >1.5 is the minimum bar for most institutional strategies.
• Maximum Drawdown (MDD): The largest peak-to-trough decline. Institutional strategies typically target MDD < 15-20%.
• Win Rate: Percentage of trades that were profitable. Counterintuitively, strategies with win rates of 40-50% can still be highly profitable if winners are significantly larger than losers.
• Profit Factor: Gross profit divided by gross loss. A profit factor >1.5 is generally considered acceptable.
• Sortino Ratio: Like Sharpe, but only penalizes downside volatility — more relevant for trend-following strategies.

Common Backtesting Pitfalls

Look-Ahead Bias

Look-ahead bias occurs when the strategy uses information that was not available at the time the trade would have been executed. For example, using the closing price to generate a signal that was executed at that same closing price. This is the most common and most damaging backtesting error — it produces inflated results that will never be replicated in live trading.

Survivorship Bias

When backtesting on a current stock universe, the historical data naturally excludes companies that went bankrupt, were delisted, or merged. This biases the backtest results upward because the strategy would have avoided companies that subsequently failed. On NSE, this is particularly relevant for mid and small cap strategies.

Overfitting

Overfitting occurs when a strategy is excessively tuned to historical data — its parameters are optimized to perform well on past data but have no predictive validity for future markets. The result is a strategy that looks exceptional in backtesting but fails immediately in live trading. Out-of-sample testing (walk-forward analysis) is the primary defense against overfitting.

Backtesting in Indian Capital Markets

NSE provides end-of-day OHLCV data going back to 1994 and tick-level data from approximately 2014. BSE similarly offers historical data for its listed securities. MCX (Multi Commodity Exchange) data is available for commodities including gold, silver, and crude oil.

For options strategies — particularly Nifty and BankNifty weekly options — accurate backtesting requires options chain data with Greeks (delta, gamma, theta, vega) across strikes and expiries. This data is expensive and harder to obtain, which is why institutional-quality backtests for F&O strategies require specialized data infrastructure.

EquiDrift61's strategy library includes strategies with verified backtest results, with metrics including Sharpe ratio, Sortino ratio, maximum drawdown, and monthly returns heatmaps. All strategies must exceed a Sharpe ratio of 3 in backtesting before inclusion.

Related Terms