There are a wide variety of open-source backtesting framework. This is due to the fact that everyone has its own idea and preference on how a backtesting engine should work, so they make their own implementation. This is kind of why I created this framework: I wanted to have something more flexible, and more complete than what I found on the market. I wanted to have a framework that is designed to be as close as possible to the real world, and that is designed to be as flexible as possible. I also wanted a framework that was well integrated, and could spare most of the non-datascience development time. This means that the datascientist or quant can invest more time at actually developing the strategy than setting up the code around to strategy to fetch the data, make the backtest work, or deploy the strategy. This is why I created this framework: A complete and flexible framework, the most realistic framework I found in the open-source, and the one that gives you the best bang for your buck in productive time (based on my opinion).
This framework was created by aiming to build a system that could test complex strategies that can work on multiple assets types (stocks, crypto, derivatives, etc.) at multiple time-resolutions and simulate conditions that looks as much as possible to the real world. This has been created for datascientists and quants that designed their strategy with statistical or machine learning models, and that want to test them in a realistic environment. This framework will provide accurate results on how a strategy would have performed in the past. It will also facilitate the deployment of the strategy in a production environment. In addition, the framework provides a module to facilitate the construction of complex data pipelines. However, this framework hasn't been designed to be fast. In fact, it is quite slow compared to other backtesting frameworks. This is due to the fact that it needs to make a lot of computation at every time-steps to simulate real-world conditions (Handling splits, dividends, computing indicators, etc.). However, if used correctly, it shouldn't be a problem. In fact, the backtest module of the framework was designed to be the last step of a strategy development. At prior steps, the strategy can be optimized on faster and simpler or vector-based backtesting framework (or even a simple loop over the data). Once we are confident that the strategy should be good, then we can test it on this framework to see how it would have performed in the past. This is the goal of this framework. If your strategy is used for high-frequency trading, this framework is not for you, for now. If there is a big demand for it, it might be implemented in the future.
As said before, the backtest module of this framework is designed to be in the last steps of a strategy development. It can provide you with powerful insights on how your strategy could perform in real-world scenarios. It can be the last line of defence before deploying a strategy in production. It can give you more confidence in your strategy, or help you find the flaws before deploying it. However, this framework can be integrated at every step of the workflow. Inf act, it comes with a package that helps you build complex data pipelines in a really intelligible way at the firsts steps of the workflow. This can save you a lot of time in the long run and help the maintenance team to understand the code more easily and fix the pipeline more quickly in production. Finally, it can help you deploy your strategy in only few lines of code for the last step of the workflow. In addition, it is a project in active development ,so if you believe some features are missing, you can contribute to the project and add them or add a feature request on GitHub. This is the power of open-source.
If you read this, and you think that this framework doesn't fit your needs, or you want to use it and want another framework for quicker computations, here are some alternatives that you can use:
If you like the initiative, but the features you need aren't implemented yet, you can contribute to the project or add a feature request on GitHub. We are open to any suggestions, and we will try to implement them as soon as possible. Welcome to the community!