How far into the past should you actually ‘back-test’ a method you’re convinced has a chance of working live in the market? This is a question high on the majority of traders’ agenda. Many experienced traders would immediately suggest that backward testing is irrelevant, only “forward testing is valid” and they’d be right, but only partially…
Back testing your method, in order to discover if the method and edge you’re working on actually ‘works’, is the right action for traders to take in order to put their theory to the test. Firstly, it encourages trading discipline; you’re being scientific in subjecting your method to forensic analysis and unless your back-test provides good returns, from an easy to implement plan, then you’re unlikely to put the method to work into the real market.
Secondly patterns, particularly those on higher time frames (such as the daily time frame) repeat, and they repeat time after time. They move in tandem with interest rate alterations, quantitative easing programmes, the carry trade etc. And the daily time frames and higher time frames tend to be devoid of rumours when measured over the longer term, as the impact of rumours is negated in preference to price dovetailing with fundamental policy decisions.
For example, traders pulling up a daily chart for the past twelve months would be able to pinpoint, within a day or two, the point at which the USA Fed announced their Q.E. to infinity programme, or the point at which various officials in the ECB began to talk up the possibility of an interest rate cut. Or the time when the ECB president, Mario Draghi, went on record as stating that he and his office would; “do everything it takes to protect the euro and the Eurozone”. And the point at which, in December 2012, the Fed and USA government finally reached an agreement on the fiscal cliff solution. There are waves and patterns of sentiment displayed on the daily chart and in many ways they’re no different to the waves and patterns painted on the previous yearly charts our back test will reveal.
What results to look for in your back tests and why ‘curve fitting’ isn’t necessarily bad practice
With our back test we have the ability to enjoy what we term “curve fitting”; we can clearly see where the market trend/swing sentiment changed and we can match our trading method to fit these turns exactly. And if you’re a trader who favours using an indicator based strategy, then back-testing provides a perfect back drop to curve fit.
For example, the main trend trading indicators should align perfectly with the previous turns in trend. The MACD, the RSI, DMI, stochastic lines, Bollinger bands, ADX and the PSAR on standard settings will no doubt fit perfectly in our curve fitting back test. Adjust your chart for Heikin Ashi candles and the swing/trend patterns will light up like it’s the Fourth of July.
We can add layers of sophistication to our back test and take a very different view of how to conduct it. We could use a Fibonacci retracement method, we could use trend lines, or significant moving averages such as the 50 and 200 SMA. However, the ‘results’ will be the same, you’ll probably get your method to ‘work’, but in many strategists’ opinions it’s the indicator based strategies that work best in terms of back-testing, therefore (in theory) they should work well forward testing in the live market.
Back test objectors
The main objection to forward testing is this; you can gauge the qualities of your back test for three years, but you can’t forward test for three years, unless you’re prepared to stay out of the market for that length of time whilst using a demo account. Therefore the claims of forward testing being the ultimate test are rendered pointless. None of us can forward test for three years therefore when putting your method into the market there’s always the chance that the live forward test won’t match the back test performance. Once you’ve back tested your method and you’re convinced that the method will work, it should immediately be out of test mode and into the real market.
High probability set ups are simple to detect when back testing and even simpler to put into operation.
We’ve accepted that we’re curve fitting when back testing a potential method; as such we can adjust our indicators and their settings to hopefully pinpoint, with some degree of accuracy, when price and the trend have turned. This then becomes a de facto high probability set up, logic dictates that if it wasn’t then quite simply we wouldn’t use the method. We’ve immediately developed a robust trading edge as if by accident.
It’s far too easy to dismiss back testing, but if we accept that the market repeats and rhymes to the rhythm of many of the fundamental policy and high impact news decisions that take place throughout any normal year, even when allowing for the abnormality of programmes such as monetary easing, then surely back testing has its place in a robust trading plan?