If you are devising a currency trading system you will probably want to back test and if you do you need to aware of the dangers of curve fitting which is the subject of this article. Let's take a look at the dangers of curve fitting in more detail.

A common mistake when traders back test a currency trading system is to try and improve the performance by adding more indicators or rules to their systems method of generating trading signals but there is a problem with this.

All the programmer is doing on the software is bending the system to fit the data and this is known as curve fitting. The problem with curve fitting is that data never repeats exactly the same way again and the system which has been bent going backwards, collapses going forward.

You will for example, see a huge amount of Forex Robots for sale online which all claim they can make hundreds of percent per annum or more but they never make the same gains for users and the reason is the systems are curve fitted. Most of the Forex robots sold online curve fit on purpose, just to get system track record they can sell to naïve or greedy traders but in most instances curve fitting is done by traders, who just don't realize the problems it causes.

To a degree any trading system you back test will be curve fitted but you can restrict its impact by the following

Keep Your System Simple

Keep the system simple, simple systems work better than complex ones and have fewer elements to break – its a myth that complex systems work better they don't because Forex markets are an odds based market and in this type of market, simple system are more robust and likely to make Forex profits than a complicated one.

Same Rules for All Market Conditions

You should use the same rules for all market conditions, long term trading or short term trading, you don't know when the market conditions are going to change but you do, when your doing a system back test and different rules for different market conditions - is curve fitting!

Test it Over the Long Term

Many traders test their systems over a few months or a year but the time period should be much longer than this and we would recommend 10 years or more.

Final Words

No matter what system you use, its never going to be perfect and the search for the system that is is futile so don't try and make your system perfect through curve fitting. If you keep your system simple, make it trade all market conditions in the same way and back test it over 10 years of price data or more, you will have a robust currency trading system which will make a similar performance in real time, as your back tested results.