Most algo trading strategies that look brilliant in a backtest fail in live trading, and the reason is rarely a bad idea. It is a bad backtest. Overfitting, lookahead bias, ignored spreads and commissions, and testing on too little data can turn a random strategy into a beautiful equity curve that has no relationship with the future. A correct backtest is a scientific test, not a marketing chart. Follow this workflow and you will know, before risking a dollar, whether your strategy has a real edge or just a good-looking past.

Step 1: Clean Data and Honest Rules

Start with clean, high-quality historical data from a reliable source, covering at least two to three years and ideally a full market cycle, including a trend period, a range period, and a crash. Write your strategy rules down before you test them — entry, exit, stop loss, position size — and freeze them. The most common error is tweaking rules after seeing the results, which is like shooting an arrow and then painting the target around it. Record every assumption: spread, commission, slippage, and swap costs, because a strategy that makes 2 pips per trade on paper loses money once a 1-pip spread and commission are applied. If your strategy only works with zero costs, it does not work.

Step 2: Walk-Forward Testing Beats One Big Backtest

A single backtest over the full dataset proves nothing, because you can always find rules that fit the past. Walk-forward testing is the fix. Split your data into segments: optimise the strategy on the first segment, then test it untouched on the next segment, which the strategy has never seen. Repeat across the whole dataset. If the strategy is profitable on the unseen segments, it may have a real edge; if it only works on the data it was optimised on, it is overfitted and will fail live. As a rule of thumb, a strategy needs at least 100 trades in the out-of-sample test before you can trust the result, and the metrics to judge are expectancy and maximum drawdown, not win rate. A 45% win rate at 1:2 reward-to-risk is a profitable strategy; a 70% win rate that risks three times what it gains is not.

Step 3: From Backtest to Live Trading

Even a strategy that passes walk-forward testing should go through two more stages. First, forward-test it on a demo account for at least a month, trading exactly as the rules say, to catch data errors, execution problems, and logic bugs the backtest could not see. Second, go live with the smallest position size your broker allows and risk no more than 0.5% per trade for the first 50 live trades. Live trading adds slippage, requotes, and your own emotions, and all three degrade performance versus the backtest. Keep a journal that compares live results with backtest expectations: if live expectancy is less than half of backtested expectancy, stop and re-examine the strategy before scaling up. The goal of backtesting is not to find a perfect strategy, it is to find a robust one that survives contact with reality.

Frequently Asked Questions

How many trades do I need for a reliable backtest? At least 100 trades in out-of-sample data, and more is better. Fewer than 30 trades is not a test, it is an anecdote.

What is overfitting in trading? Overfitting is tuning a strategy's rules so precisely to past data that it captures random noise instead of a real pattern. It produces beautiful backtests and losing live accounts.

Should I include spread and commission in a backtest? Yes, always. Use realistic spreads for your trading session, add commission, and add a small slippage allowance. A strategy that only works with zero costs is not a strategy.