Introduction To Backtesting

This reading offers a summary of backtesting and other methods of assessing the risk of investment strategies. Backtesting allows investment professionals to use historical data to simulate the performance of strategies, and to analyze the risk and return properties of these strategies, prior to making any actual investment.

Increases in the availability of data along with dramatic increases in computing power have facilitated quantitative assessments. Today, widely available software allows an investor to backtest and simulate infinite combinations of possible investment strategies, create multifactor models, and build investment portfolios, all before making any actual investment.

Objectives of Backtesting

Backtesting is a process by which we use actual historical data to emulate the investment process, in order to determine whether a particular investment strategy or technique would have delivered the expected excess returns.

The primary goal of backtesting is to assess the risk and return of an investment strategy by simulating the investment process. This practice helps to reassure investors that investing strategies and models are likely to perform well. Additionally, backtesting helps investors to refine and improve their investment process.

Backtesting has been extensively used in the financial industry for decades. While backtesting is a natural fit for systematic and quantitative investment styles, fundamental managers also make extensive use of backtesting.

The backtesting process is intuitive, because it mimics how investing is accomplished in reality: we formulate investment strategies, test those ideas using historical information, and then assess how well those strategies would have performed.

Backtesting relies on the assumption that the future will, at least to some degree, look like history and thus that a strategy that performs well in backtesting should perform well in real life. For various reasons that we will explore later, this is often not the case: methodologies that perform well in a backtest sometimes fail to deliver outperformance in actual use. (Conversely, a model that would have performed well in reality but that doesn’t exhibit predictive power in backtesting will probably never be implemented.)