A backtest is a simulation of a model-driven investment strategy's response to historical data. The purpose of performing experiments with backtests is to make discoveries about a process or system and to compute various factors related to either risk or return. The factors are typically used together to find a combination that is predictive of return.

While working on designing and developing a backtester, to achieve functionalities, such as simulated market pricing, ordering environment, order matching engine, order book management, as well as account and position updates, we can explore the concept of an event-driven backtesting system.

In this chapter, we designed and implemented an event-driven backtesting system using the TickData class, the MarketDataSource class, the Order class, the Position class, the Strategy class, the MeanRevertingStrategy class, and the Backtester class. We plotted our resulting profits and losses onto a graph to help us visualize the performance of our trading strategy.

Backtesting involves a lot of research that merits a literature on its own. In this chapter, we explored ten considerations for designing a backtest model. To help improve our models on a continuous basis, a number of algorithms can be employed in backtesting. We briefly discussed some of these: k-means clustering, k-nearest neighbor, classification and regression tree, 2k factorial design, and genetic algorithm.

In the next chapter, we will discuss Excel with Python, using the Component Object Model (COM).