A general definition of backtesting is the process of applying a trading strategy or analytical method to historical data to see how accurately the strategy or method would have predicted actual results. It is used in ex-post circumstances to access the viability of a particular application in order to see potential success or failure in using it in implementing the same sequence in the future. Nowadays, complicated strategies such as those generated by machines are too difficult to evaluate empirically, so backtesting is one of the most effective ways to test whether or not the proposed approach is feasible in an application. It can be extremely beneficial in order to optimize a trading strategy and take a look at what the analysis of risks, results, and profits would be before actually risking any capital. If done successfully, the backtester can give an indication to the user that their hypothetical strategy continuously yields positive results and profits and that it may be worthwhile to pursue such a strategy for real-life usage. If the results are not to the user’s liking, it may be an indication that they should switch their strategy or try a different approach.
In many circumstances, traders will want to use computer software, in particular, using a proprietary trading platform language and seeking qualified programmers to morph their ideas into a testable form. In many cases, this is done by having the programmer adjust for input variables that allow the trader to alter the particular factors that perform best using historical data when backtested. Usually, the ideal conditions that would allow for the best type of model are relevant to the market conditions and relevant time period that the sample data is compared to. The beauty of backtesting is that it can be done for a variety of goals and predictions, including but not limited to predicting return percentages, volatility of stocks, or a company’s net income. However, it is important to note that despite the notion that if it worked in the past, it will work in the future, it is still risky to assume that past success or failure guarantees the same results in the future.
A good backtest factors in the most minute details no matter how seemingly insignificant they may be. These negligible factors culminate together over time, resulting in a large impact over the course of a period and drastically affecting the results. Examples of other supplementary, good models that address these inputs are out-of-sample testing and forward performance testing and when they have a strong correlation with backtesting, they add strength to the viability of a system. This is because backtesting alone has a few pitfalls such as the inherent tendency for individuals to include their own biases and preconceptions upon development. Thus, an effective backtest should be generated without consideration of reliance on the data that will be used to backtest, a caveat which is principally difficult because of the many traders will use historical data as an outline for constructing their models. This disrupts the validity of the results of the backtest, creating superficial results that do not give a true indication of an unbiased strategy.
In Forex trading and markets specifically, backtesting is one of the most popular and straightforward approaches used in the trader community. Forex is shorthand for the words, “foreign exchange,” and revolves around the buying and selling of currencies in the foreign market. For Forex traders, they will try to maximize profits through prediction of which currencies will appreciate or depreciate the most relative to others depending on their strategy.
One large aspect that comes from the technical analysis is the belief that past prices are strong indicators for performance in future markets.
Additionally, many of the highly important and frequently considered parameters while constructing a model include the total profit or losses, average profit or losses, success ratio of number of trades that were winners compared to the number of trades that were losers, maximum drawdown as a measurement of risk, and the measure of a risk-adjusted return calculated using the Sharpe ratio.