Backtest

A backtest is a way of checking a strategy or program before putting it to real use, by running it against data from the past and calculating 'what result would this have produced if I'd used it back then.' It's similar to a chef trying out a new recipe a few times with ingredients already on hand before serving it to customers. The key point is that you can spot problems in advance without risking real customers—real money, real situations.

These days, backtesting isn't limited to strategies designed by humans. It's also used for systems where multiple automated programs split up roles, make judgments, and generate trading signals together. Deploying such a system directly into live markets could lead to unexpected losses, so it's first run against historical market data to see how profits and losses would have played out, and only then is real capital entrusted to it.

However, a strong backtest result doesn't guarantee the same performance in the future. A strategy that's been fitted too closely to past data can fall apart the moment real conditions shift even slightly—a phenomenon commonly called 'overfitting.'

It's hard to experience a backtest itself without dedicated tools, but you can get a feel for the concept by asking a chatbot something like this:

"If I were backtesting an investment strategy using five years of stock price data, what metrics (like returns, maximum drawdown, etc.) should I check?"

Asking this will get you an explanation—with concrete examples—of why the metrics commonly seen in backtests matter, and why past performance alone can't guarantee future results.