What Is Backtesting? A Practical Guide for New Algo Traders

Backtesting is the process of running a trading strategy against historical market data to estimate how it would have performed. It is the cheapest filter we have: most strategies look great on a napkin, fall apart in a backtest, and never reach live capital.

Why bother

The honest answer is that most trading ideas are bad. Some are bad because the edge is illusory. Some are bad because the edge exists but is too small to overcome costs. A backtest does not tell you which strategies will work — it tells you which ones definitely will not. That is enormously useful.

What a backtest cannot do

A backtest cannot promise future returns. Markets evolve. Edges decay. The data you tested on may not resemble next year. A clean backtest is necessary but not sufficient — it earns a strategy the right to be considered, nothing more.

How to read results without fooling yourself

Look at the equity curve first. A strategy with smooth equity and a deep, late drawdown is more dangerous than one with a noisy curve and frequent small losses, because the smooth one teaches you to size up before it breaks.

Then look at trade count. Twenty trades over five years is not a backtest, it is a story. You want hundreds of independent trades before you trust any statistic.

Finally, separate in-sample and out-of-sample results. If they diverge sharply, you overfit.

Where to start

Pick a simple strategy — moving-average crossover, mean reversion on RSI, momentum on a 12-month lookback — and backtest it across multiple instruments and decades. The goal is not to find the best parameters but to develop intuition for what real edges look like.