Backtesting of forex trading strategies is probably the most misunderstood area of forex trading research.  Most if not all forex newbies spend countless hours trying to discover the Holy Grail of forex trading, the magic forex trading formula that yields consistent returns if applied mechanically.  They brag in the Internet about the backtested results and guard the formula like a tiger guards its wounded cub.  And then they lose their money and disappear.   A particularly nasty breed of backtesters are smart enough to avoid using their formula, so they sell “the bullet-proof forex trading strategy” to the unsuspecting and disappear, too.

Now, what is backtesting? As the omniscient Wikipedia puts it, “Backtesting is a process, usually performed with aid of computers, by which traders try to estimate how financial instruments would have performed in the past had a particular mechanical trading system been employed to trade them.”   Sounds perfectly legitimate, but it isn’t.  

Backtesting is by definition performed on a finite set of data.  That could be anything, but usually it’s a set of price series.   Therefore, no matter how much or how long you backtest your forex trading strategy, the only result that you can get is that it behaves in a certain way in SOME circumstances – precisely, those that were used for the test.  It remains completely unknown how it behaves in ALL OTHER circumstances, i.e. untested – and that includes the future.  No method of backtesting, regardless of its complexity and involved precautions, can work around this.  No amount of backtesting can predict the future behavior of a forex trading strategy in real life.

So what is backtesting good for?  Well, if it is done properly and the sample is big enough, it is everything it is said to be – just if you use the word “may” instead of “will”.  A proper backtesting result is an indication of a tendency of a forex trading strategy to behave in a certain way.  Nothing more, nothing less.  It is not a guarantee.  Also, notice the word “proper”.   The many ways to get a result that does not indicate anything include using commercial backtesting software.  I’ll get into that some time.    

Still looking for the Holy Grail.