Wes McKinney at AQR and how he uses python for financial analytics

I watched a great PyCon 2010 demo where Wes McKinney at AQR showed how he uses python for financial analytics.


Other Useful tidbits I picked up from the demo

Now for a bit of a tangent 😉

He made a comment about transitioning from prototyping to a “Production” ready system which is relevant to what I do professionally. In the scientific or finance world this often involves using tools such as Mathcad or Mathematica to prototype your work and then having to take that prototype and build it into your production application.

This may mean handing the prototype off to another team entirely.  There are many places for errors to creap in and the process around making enhancements or updates can be slow and combersome.

I’m doing significant research into how to improve this now so I’ll write about this separately as I have something substantial.  Let me know if you have found any methods that work for you.  Stay tuned!!!

John Rizzo

Technology, problem solving, learning, business, society and where those things intersect is what I am always thinking about. From my hobbies to my profession I attempt to combine those interests in a way that makes the sum greater than the whole. Find out more about me at linkedin, http://www.linkedin.com/in/johnrizzo1.

You may also like...

Comments

  1. […] This post was mentioned on Twitter by johnrizzo1, johnrizzo1. johnrizzo1 said: New post: Wes McKinney at AQR and how he uses python for financial analytics http://johnrizzo.net/blog/2010/03/31/python-finance/ […]

  2. Josh Hemann says:

    I met Wes at PyCon and heard his talk. Smart guy, and his Pandas package is great for slicing and dicing time series data.

    Wes’ talk touches on an issue I see a lot: analytic groups prototype stuff in MATLAB, R, and increasingly Python, but often have to move the analyses to a lower level language like C or C# for myriad reasons (e.g., speed, scalability, IT requirements with respect to existing applications, etc). There is typically a lot of pain and cost in that process.

    Python being a much more general language than R or MATLAB, it is a much better language in my opinion for doing analytical work. I am on the PyIMSL Studio team at Rogue Wave Software and we wrap out IMSL C Numerical Libraries in Python so in addition to all of the great stuff in packages like SciPy and Pandas, you have access to hundreds of mathematical and statistical algorithms that you can call from Python, or call the exact same function and API from C (or C#) if you need to move your Python prototype. Personally, I try and stay in Python as much as possible, but I am more of a modeler than an enterprise application developer. It is nice knowing though that a lot of my work can be easily handed off to the C developers down the hall and that they can actually deploy the analyses I have come up with.

    Is anyone else out in the ether using Pandas? I have played with it a little, but the problem is that the vast majority of Python functions expect just the data values, not an R-like data frame with one column being datetimes. Wes has included a bunch of descriptive statistics routines (and I think it will fit a linear regression model too) that work with these data frames but once you need to step outside of Pandas’ functionality you need to detach the data from the datetimes so typical Scipy or PyIMSL Studio functions can operate on it. Not a big deal, just something to be aware of.