Backtesting – What is it, definition and concept

Backtesting is a tool to analyze the effectiveness of an investment strategy. This, applying said strategy on historical data, so it is a retrospective method.

In other words, backtesting allows us to evaluate what the effects of an investment strategy would have been if it had been applied in a previous period. This is a simulation exercise using figures from the past.

The rationale for this tool is that if the strategy had worked in the past, it is likely to work well in the future. And likewise, in the opposite direction.

This method is used in the evaluation of investment strategies in the financial markets. For example, in the stock market or in the foreign exchange or forex market.

Backtesting steps

To carry out backtesting, first of all, you must select a period of time for the analysis (More than one period of time can be used to confirm the reliability of the result).

The historical data set must include a representative sample of values. This includes, for example, the shares of companies that may have gone bankrupt or were sold. If these types of assets were excluded, the backtesting results would be distorted.

For the application of this method, various parameters must be included, such as the investor’s initial capital, stop loss or take profit orders, the preferred position (long or short), among others.

Backtesting must also consider all transaction costs, however small they may be. Otherwise, the results of the strategy would be overestimated.

See also  Abenomics - What it is, definition and concept

For the procedure, a software is used. There is even the option to do it in Microsoft Excel.

What should we avoid?

It is important to avoid certain wrong practices in backtesting, for example, entering extra parameters and trading until the result is as expected. This would be adapting the procedure to fit what I want to get.

Another mistake that should not be made is to use backtesting for long periods with the aim of achieving better results, when in reality you plan to operate in the short term.

Similarly, the simulation of several possible strategies on the same data set should be avoided. This, because it may be that, if any of these strategies is successful, it is by chance.

In addition, one way to avoid bias is, as we have mentioned before, to apply the backtesting not only in one period of time, but in another outside the initially selected sample. In this way, the results can be verified.

Leave a Comment