(Automatic) Trading Frameworks (aka yabf)

[goodboy]

There's quite a few all geared at different things but I figure we might as well start a list of what's been tried as well as what has multi-broker support and/or (a) reasonable API(s) and can be run without a cloud platform.

yabf = yet another backtesting framework

There's obviously going to be a lot to choose from and many different specialties. I will not be surprised if we end up rolling our own since none of these seem to leverage modern Python's async features..

Python YABFs

• backtrader

  • hey look, its feature list is documented (how refreshing)
  • has a great list of alternatives (hopefully most of which we'll investigate)
  • has a large client repo base (i.e. projects that use it)
  • high level view of the GUI
  • uses matplotlib
  • unfortunately, it leverages multiprocessing + callback-hell

  • there's a lighter weight wrapper project, fastqant that has this medium pamp :roll_eyes:

• quantopian's zipline

  • the gold standard in python algo trading :roll_eyes:
  • they have an online web platform (with an IDE) that let's you do "online trading"
  • some coolio's made a live trading version, zipline-live
  • a pipeline system for peppering over the fact they don't have a streaming / async system underneath or care about generators...

  • they do have a ton of very cool quant libs including alphalens, pyfolio and qgrid

• jesse-ai/jesse

  • intro blog post talking about the simple design genetic algo for feature optimization
  • starter forum
  • amazing set of indicators mostly implemented with numpy / ta-lib
  • uses the peewee ORM for storage
  • interesting set of trading models

  • some neato data generators

• pylivetrader from alpaca

  • clearly focused for using alpaca as lone broker
  • works on a modified version of quantopian pipelines called pipeline-live

  • support for persistent storage via redit

• Quantconnect's Lean

  • biggest oof is it's mostly C#

• Rob Carver's pysystemtrade

  • he has a cool blog all about it.

• QTPyLib which has quite a cool set of features:

  • has support for live data and trading with IB
  • uses zeromq
  • has sms notifications via twilio
  • actual notion of common professional tools (like [blotters])

  • has a really strange async system that's for whatever reason re-written in this

• freqtrade (another crypto callback soup) but, GPL3

  • docs
  • has an included "hyper-optimization" subsystem for tuning signal/indicators
  • cool org

GUI YABFs

• backtesting which internally uses bokeh for plotting
• pyalgotrader which is an english port of the famous chinese vnpy
  • also uses pyqtgraph for charting (though not sure if supports real-time quotes; should be a lot of useful code in there tho)
  • author has some interesting S / R tooling work
• of course Quantdom which is the project I based the current charting code on :partying_face:

[goodboy]

I've got to say my main issue with most of these ^ is the overuse of classes to define both strategies and the data flow model.

[goodboy]

Rob Carver's pysystemtrade and blog look decent as well.

(added to above)

[goodboy]

• [x] TODO section on projects with UIs.

List I've got so far:

• backtesting which internally uses bokeh for plotting
• pyalgotrader which is an english port of the famous chinese vnpy
  • also uses pyqtgraph for charting (though not sure if supports real-time quotes; should be a lot of useful code in there tho)
  • author has some interesting S / R tooling work
• of course Quantdom which is the project I based the current charting code on :partying_face:

[goodboy]

vectorbt also looks interesting, in the sense it's not using pandas 😂

Probably worth some digging and peeking. Found it at this reddit thread.

[goodboy]

https://docs.nautilustrader.io/index.html is notable for having lower level components in cython and rust but it's still got a gnarly callback API.

Decent integration set including ib.