What Constitutes a Statistically Significant Sample for Back Testing?

Hi folks,

I'm currently manually backtesting a breakout system - that means drawing those daily boxes, setting the highs and lows, measuring drawdowns, calculating the extent of price movements and determining the optimal values for allowed drawdown and TP targets to maximize profitability.

Because I'm doing it manually, I cannot backtest it all the way to 1970 and see how it performs. Given that, roughly how many trades would you want to backtest for before considering a given sample to be statistically significant?

This is what I do: http://www.forexfactory.com/showthre...45#post2834345

You have that much data? Which time frame? Which pair? If you don't mind....

i never validate a system with less than 20,000 trades. i do this by using synthetic data which i create.

mark brown

There would be no point going back four decades to 1970 even if you could, the markets have changed so much since then. 20-day breakouts were very profitable in the 1970's - a system that made Richard Dennis and William Eckhart rich doesn't even work today... About ten years is probably a decent amount of time IMO. If you're going to do a lot of testing (and as a trader you'll need to) I'd seriously recommend learning how to program something though.