Should I watch these 3 hours of Python videos to convince myself this should be the programming language of choice for quant trading research?
I ask myself about Python all the time. Here is an interesting set of 3 hours of videos I think I should watch. This presenter is the same guy at http://www.thepythonquants.com/
A couple of notes on the first video:
This is excellent and for free with Linux. It seems there is no advantage thus far for those using matlab or even considering f#. They also seem to have difficulties with package managers for their dependency packages which is a real turn off compared to what I am used to. Reminds of r.
Do you use synaptix as a package manager for all dependencies?
27:20 says to use NumPy to speed up algo and make it more compact without the the need of python loops
Outside of python, implement correct algo,as in use fast Fourier tams form in a CCR derivative model. It will be 900x faster vs naive version.
31:34.talks about Monte Carlo simulation algo
Yves hilpisch presenter has a book on this with code.
Pycuda mentioned for parallelism.
Pandas has cool time management against time series data. Very useful against time series with incomplete data vs r. At 54 minute, he talks about future changes in Pandas where it can impact your current scripts.
at 1:07 it seems a student struggles with pandas for excel integration where it is admitted it is a mess but thus was in2012 but these function changes can screw up your scripts with automatic package updates.
At 1:13 he explains how the matplotlib plot function will not translate the dataframe with the adjusted time format as compared to the built in pandas plot. You never worry about this in matlab as it all plays nice.
At 1:30, I am pulling out my hair watch these people have configuration issues downloading data from yahoo finance with pandas.
The rolling window with math calculations within pandas is cool.
PyTable is mentioned which works with pandas. It uses hdfs as the file storage which is quick.
Can use pytables with latex for custom automated reports after Monte Carlo simulation.