Backtesting Your Style Across Regimes

A backtest is not a forecast. It is a sample of how your setup would have behaved across past conditions. The point is not to find a number that makes you feel confident. The point is to find out where the setup breaks.

Three things a useful backtest tells you, and a live account is an expensive place to learn them:

• The shape of the drawdowns. Average drawdown matters less than the worst one. If the worst historical drawdown is deeper than you can sit through emotionally, you will abandon the setup at the bottom every time.
• Where the setup struggles. Most setups have a regime they hate: trend systems hate chop, mean reversion hates strong trends, breakout setups hate the slow open of a holiday week.
• Whether to believe the win rate at all. Small samples lie. A 70% win rate across 30 trades is noise. The same rate across 300 trades is data.

Mark Douglas's frame applies here too. You are not validating a prediction. You are validating that the edge exists across a series, so that you can sit with the inevitable losing run when it shows up.

If a real backtest is not possible for your setup, paper trade it for a defined sample. Either way, do not trust a setup you have only watched for a week.

You do not need an institutional-grade research platform. You need an honest sample.

Define the setup first. Write the entry, the invalidation, the exit, and the size rule before you look at any data. Setups defined after the fact always look profitable.

Pick a sample that matters. A few hundred trades across at least two different regimes (trend and chop, at minimum). Cherry-picking a strong year proves nothing.

Track the right numbers. Win rate, average winner in R, average loser in R, max consecutive losses, worst drawdown. Expectancy in R per trade is the single number most worth knowing.

Read the drawdowns out loud. Six losses in a row, peak-to-trough seven R. Could you sit through that with real money? If not, the setup is not yours, or the size is too big, or both.

Stress the rough conditions. What happens to the setup in 2008, 2015, 2020? You do not need to predict the next crash. You do need to know whether your setup is the kind that gets carried out on a stretcher when one shows up.

Then forward test small. Backtests assume perfect fills and no missed signals. Live execution does not. Run it small until the live numbers look like the backtest.

Get clear about which environments your setup actually fits.

This is an interactive exercise. The reflection and structured worksheet open in your dashboard tools. Read through the prompt below first, then come back to complete it.