Backtesting Basics: Proving (or Killing) a Strategy Before You Risk Real Money

Part of the Strategy Blueprint series. This step only works if your entry and exit rules are written down clearly enough to actually test.

Your strategy has an opinion. Backtesting is where reality votes.

You’ve defined your edge. You’ve written entry and exit rules. At this point most traders do the worst possible thing: they go live with real money to “see how it does.” That’s not testing a strategy — that’s gambling with extra paperwork.

Backtesting exists to answer one question before any money is at risk: applied consistently across real historical data, do these exact rules actually make more than they lose? Per Investopedia’s definition of backtesting, it’s the process of testing a strategy using historical data to see how it would have performed — a way to evaluate viability without risking capital.

What backtesting can prove — and what it can’t

A clean backtest can tell you whether your rules have positive expectancy over a large historical sample. It can surface an edge that isn’t obvious from staring at charts, and it can kill a strategy that felt great anecdotally but never actually made money once every trade was counted honestly.

What it cannot do is guarantee future performance, account for slippage and real fills perfectly, or prove a strategy works in live conditions with your actual emotions attached. That’s exactly why backtesting is step one, not the finish line — it’s a filter for bad ideas, not a certification for good ones.

The four ways traders fool themselves with backtests

1. Curve-fitting

Tweaking rules repeatedly until the backtest results look perfect on one specific dataset. A strategy with 15 optimized parameters that all happen to fit the last two years of one instrument isn’t an edge — it’s an elaborate description of the past that will fall apart the moment conditions shift even slightly.

2. Too small a sample

Thirty trades is not a sample size — it’s a rounding error that can look like anything. A real read on expectancy needs enough trades, across enough different market conditions (trending, ranging, high and low volatility), that the results represent behavior rather than luck.

3. Ignoring costs

Commissions, spread, and realistic slippage all eat into results. A strategy that’s only profitable with zero friction assumed isn’t profitable — it’s a spreadsheet exercise. Build costs into the test from the start, not as a footnote afterward.

4. Look-ahead bias

Accidentally letting the test “see” information that wouldn’t have been available at the time of the trade — like using a session’s full range to define a level that triggers a trade mid-session. This is the sneakiest error because it makes a bad rule look great without anyone intending to cheat.

What a usable backtest actually measures

Beyond simple win rate, track: total trades (sample size), win rate, average win vs. average loss, expectancy per trade, maximum drawdown, and the longest losing streak in the data. That last one matters more than most traders think — if your backtest shows a possible 12-trade losing streak and you can’t financially or psychologically survive that in real time, the strategy might have positive expectancy and still break you before it pays off.

A passing backtest is a hypothesis, not a green light

Even a clean, honest backtest with a real sample size only tells you the rules worked on data that already happened. It doesn’t confirm the rules will hold up when you’re executing them in real time, with real emotions, on data that hasn’t happened yet. That’s the entire reason the next step exists.

Takeaway: A backtest’s job is to kill bad ideas cheaply. If your rules can’t survive an honest test against a real sample size with real costs included, they don’t deserve real money — full stop.

Next up: Forward Testing on Demo: The Step Traders Skip and Regret — because surviving history is not the same as surviving live conditions.