“Past performance is not a reliable indicator of future results” is a standard disclaimer we often see. So is there any point in back-testing?

We believe that it all depends on the size of the historical sample that you’re using. The past performance of a single share is certainly no guarantee that it will continue to perform in the same way. However, both the insurance and casino industries have enjoyed centuries of profitability by relying on large sets of historical data to judge the probability of some event happening in the future. It’s also the method that economists use to forecast the future.

The reason is The Law Of Large Numbers. This fundamental concept in probability and statistics states that, simply put, the larger the number of historical samples used, the more likely it is that those samples will be a useful guide to the population as a whole.

In the context of back-testing, the key is the number of shares included in the test multiplied by the number of data-points (e.g. price movements, company results, etc.) in the period over which the test is run. So provided that the number of shares and the period of the test are large enough, then we believe that back-tests can be a useful indicator of the future performance of a trading system.

But there’s obviously an important caveat. A back-test’s results will not only reflect the performance of your trading system, they will also reflect the macro-economic environment over the period the back-test was run. Clearly a deep recession is going to have a different impact to a rapidly growing economy. This is another reason why we believe that back-tests should be run over as long a period as possible, in the hope that it will cover these different environments.

One further thing you may like to consider: there has been rapid growth in back-testing and algorithmic trading in the financial services industry over the last 40+ years. If it works for them, maybe it can work for you.