Back Testing

Many traders advocate back testing, and it’s easy to see why. Back testing usually entails taking several years of market data, setting up a program to work like your trading plan or strategy, and setting it to run through those years with a pretend account, making bets in accordance with your plan, and seeing what the outcome is. In other words, it’s a way of “proving” that your system will make money, and giving you some idea of how much, or perhaps how good your system is. Some spread betting books ignore back testing, but wouldn’t you rather have some idea of how well your system is going to work before you use it?

Now some strategies are easier to back test than others, but the point of a trading plan is that you should be able to rely on your results over time, so it shouldn’t involve discretion. Therefore, it should be programmable for you to check how well it would have done. This doesn’t mean that you can’t back test manually if you have to, but it will be time consuming.

You will find software that has been developed to enable back testing. It’s possible that any charting software you use has a back testing module, so start with seeing what is available at hand. Basically you are looking to develop statistical figures about how your chosen strategy has performed so that you can consider whether it suits your betting needs. If you don’t have or can’t afford the software, then you can always back test by hand, and even if it’s only over a limited range of time because it takes so long, it is much better than doing it at all. You can get free charts of pretty much any financial security that you are considering betting on, over the Internet.

That’s not to say that you would necessarily have bet in the way that the back test checks – for instance, you might have found it hard in practice to “pull the trigger” on a particular bet, and the back testing software would not suffer from any such emotions. But the very process of proving your strategies by back testing should give you confidence to overcome any hesitation, as you must do. If you know that the system has proved itself over time, you are better equipped to deal with your own psychological blocks, and become a better trader as a result.

Having established that back testing is a good idea, and hopefully convinced you to do it, there are a few pointers you have to bear in mind.

First, make sure that your back testing includes the spread, perhaps based on a typical value, as this is effectively the cost of making your bets. Any back testing that you do based on single values is bound to give you a rosier picture than what you would have encountered in real life.

Secondly, it’s not enough to plug an account value in at the start and run the test over a number of years coming out with a profit. You may think that you have a great system if there’s a good profit, but what you might not realize is that it takes too many risks for your peace of mind, and has large swings in your account value which in practice you would find hard to bear, and may even be tempted to second-guess the system. It’s vital that you look at the amount of account “draw-down” that you might be faced with, and decide whether the system is right for you.

Finally, don’t be tempted to over-optimise the trading system. You may find that you are experimenting, say, with the number of days in a simple moving average that you use as a signal, and that’s good. But inevitably you can “tune” your strategy too far, so that you get “perfect” results from the historic values you are using, even though they may be less than optimal with the markets going forward, when you are actually using and relying on them. A good way to guard against this is to split your historic data into two parts. If you are testing with ten years of data, use just five years for your system tuning. Once you think you have it set, run the system on the other five years of data which has had nothing to do with the way you played with the numbers, and see how it does.