The secret art of backtesting

Whatever you trade with, moving averages, candlesticks, volatility breakouts, Fibonacci retracements, or a system entirely of your own devising, you need to backtest it thoroughly and objectively before you trust it with money. The point is not just performance measurement. It is the removal of doubt.

Test your system under the conditions it will actually be traded in, establish that it is reliable and robust, and something changes psychologically: you become capable of following it.

The question in every trader's mind

Mark Jurik put it well in Computerized Trading: after the arduous process of crafting a strategy, the question is simply, will it be profitable? A sound backtesting procedure is the closest thing to an answer available before risking capital. Done rigorously, it will:

• Pinpoint which approaches to the market are likely to succeed, and which are not
• Generate honest estimates of future performance for each strategy
• Create a historical performance record you can compare live results against
• Produce the data your position-sizing and allocation decisions depend on

Profit is not the only criterion

A profitable backtest can still describe a system you cannot live with. Before you open an account, the test needs to answer all of these:

• Is the system reliably profitable across different market conditions?
• Would its worst drawdown wipe out, or demoralise, your account?
• Does it trade in a way you can tolerate: not too often, not too rarely?
• Can you sit through its longest historical string of losses without breaking?

That last one matters most. Nobody abandons a system during the winning streak.

What lack of confidence does to traders

Traders who have not properly tested their systems inevitably question them, usually right after a string of losses, which is precisely when a tested trader holds course. The untested trader tinkers, or swaps the whole system for whatever whiz-bang indicator a forum is excited about this month. Anything that sounds too good to be true finds its easiest audience in a trader who never built real confidence in what they already had.

The fix has not changed in decades: test thoroughly, test objectively, and earn the confidence to commit time and money. Traders who do this put themselves in a small minority, and it shows in their results.

Testing across a whole portfolio

The subtler challenge: a strategy's true character only appears when simulated the way you will actually trade it, across a portfolio of securities with finite capital, not one clean signal on one chart. A system can look wonderful with an expert overlay on a single security and fall apart when twelve signals arrive in the same week and your capital can only take three.

This is exactly the class of problem MetaStock's System Tester and Explorer exist for: define entry, exit, stop and sizing rules precisely, run them across years of history and a whole market's worth of instruments, and read reporting honest enough to talk you out of bad ideas. Third-party portfolio simulators from MetaStock's add-on ecosystem can push the realism further still.

However you do it, do it before the market does it to you. A month with the free MetaStock trial and one honest backtest of your favourite rule will teach you more about your trading than a year of forum opinions.