What is backtesting?

THE SHORT VERSION
Backtesting uses historical data to test a trading idea before you risk money, but a good backtest is not enough. You need to avoid bias, include costs, and confirm with forward performance testing on new data.
KEY TAKEAWAYS

What is backtesting?

Backtesting is the process of running a trading strategy on historical data to see how it would have performed. It tells you what worked in the past, not what will work tomorrow, so you test before you risk real money.

You backtest to avoid losing money on a bad idea. The process simulates trades using past prices, volumes, and rules. It gives you a number, like a return or a Sharpe ratio, before you commit a cent. That result is hypothetical performance, not realized profit.

How a backtest works

You split your data into two sets. The in-sample set is for building the strategy. The out-of-sample set is for checking it. A walk-forward test rolls the window forward, re-fitting each time on new data. You also need to include transaction costs. Slippage, commissions, and spreads eat returns.

A backtest that ignores costs will overstate profits by a lot. Prices must be adjusted for splits and dividends, or a stock split breaks your math. Use data from many market conditions, including stocks that went bankrupt, to stop survivorship bias from flattering your numbers.

The traps that fool you

Overfitting happens when you tweak a strategy until it fits the past perfectly. It fails on new data. Look-ahead bias sneaks in when you use info that wasn't available at the time. Survivorship bias comes from testing only stocks that still exist today. That flatters results.

Data dredging means testing many strategies on the same data until one looks great. Cherry-picking trades, like skipping a loss, also lies. Both make a backtest look better than reality. Stick to honest rules.

How to test honestly

After a clean backtest, you move to paper trading. That means you trade on paper, with no real money, in live markets. This is forward performance testing. Track expectancy, which is your average gain per trade, and the profit factor, which divides winners by losers.

Sharpe ratio tells you how much return you get per unit of risk. A higher Sharpe means more reward for the swings you endure, so you can compare one strategy against another on a fair basis before you risk money.

Max drawdown shows the worst peak-to-trough loss. Both numbers matter. A deep drawdown tells you how much pain the strategy can cause at its worst point. It sets your expectations for the hardest stretch you may have to sit through.

Stick to your system's logic during forward testing. If you skip a trade because it lost on paper, you are fooling yourself. Honest results come from honest execution, so apply the same rules you would with real money.

Backtesting vs. Scenario Analysis

Backtesting uses real historical data. Scenario analysis uses hypothetical data to test extreme cases, like a crash or a rate spike. Sensitivity analysis shifts one input at a time to show which assumption your result leans on. Each answers a different question.

Use both to see the full picture. Scenario analysis adds stress tests that history may not cover. Combine them for a stronger view of risk. Backtesting anchors you in what actually happened, while scenarios show what could happen.

What is Monte Carlo simulation?

Monte Carlo simulation runs your strategy thousands of times with random variations in price, volatility, and order timing. It shows a range of outcomes, not one clean line. The strategy that survives the bad rolls is the one worth trusting. It tests luck, not just history.

A backtest alone gives you one path through one past. Monte Carlo gives you many paths. Combined with scenario analysis, you see how the plan holds when the numbers turn. Trust the outcome that survives the widest spread of results.

The long-term investor's edge

Think of a backtest as a rearview mirror. It proves only one thing: that the strategy worked in one past. The honest bridge to live money is walk-forward plus forward performance testing on new data.

Hold the discipline to walk away from a curve-fit result. If a backtest looks too perfect, that is the signal to suspect overfitting rather than celebrate. The honest edge is knowing when to stop trusting a number.

Manual or automated: same honest rules

Manual backtesting means you study a chart and apply your rules by hand, then add up the results yourself. Automated backtesting runs software that scans every trade and tallies the score for you. Both test the same strategy. Neither removes the risk of a biased past.

You cannot let the tool tell you your idea is proven. A clean run in software still shows only one past. It may tempt you to keep tweaking until the chart looks perfect. That is the same overfitting trap, wearing a faster coat.

Why the data deserves respect

Detailed histories once cost more than most traders could pay, so only large institutions backtested. Web platforms changed that. Cheap data is a gift, but only if the feed is clean. Garbage in, garbage out, and a backtest hides it well.

Ask where the numbers came from before you trust them. Prices must be adjusted for splits and dividends, or the math silently breaks. Include every transaction cost you would really pay. Your result is only as honest as the data you feed it.