Assuming Python is mainly used by key players, I am wondering what is the most used back-testing library by hedge, institutions, pensions, big retail guys After extensive research I shortlisted these candidates: Backtesting.py, Backtrader, Zipline, VectorBT, VectorBT pro, and custom inhouse solution There is backtesting.py that looks quite limited, zipline is better but not a favorite, then in elitetrader history I found quoted a lot of times backtrader, which i also used a bit but it looks is not maintained anymore(even not backtrader2) and found in other forums specialized on algo trading that the best features and velocity/performance/complexity comes possibly with vectorBT. There is also a pro version with better features. Inhouse solutions I have no clue, I guess could be the way to go, but in this case would be really secretive to speak about. I am writing the post because 1) I do not know what industry uses 2) I am really surprised that if I run a query in elitetrader on vectorBT i do not find any post, maybe I am wrong in thinking that python is the most used language whereas C++ solutions may be faster?
...and I thought that Deltix with C# as language is the most widely used (and most expensive one too as I heard). https://www.deltixlab.com/ ---- Institutional-class data management / backtesting / strategy deployment solution: Equities, options, futures, currencies, baskets and custom synthetic instruments are supported Multiple low latency data feeds supported (processing speeds in Millions of messages per second on terabytes of data) C# and .Net based strategy backtesting and optimization Multiple brokers execution supported, trading signals converted into FIX orders
I am sure the "standard" at the most successful hedge funds is a custom solution that is just integrated into the general infrastructure. In the grand scheme of things a back testing engine is just not a hard problem to solve. With that said, VectorBT looks really nice. Doing everything with NumPy arrays and Numba is awesome. Even on the VectorBT page, half the page is data visualization of the backtest results. That is data visualization though and not really backtesting. Seeing that hodl BTC strategy from 2014 to today only has a Sharpe of 1 is interesting If anyone wants to try it out go with pip install -U vectorbt The optional dependencies install was installing basically every update of CCTX, some crypto library.