Recently I was reading an article called “The Realized Volatility Puzzle”, by Harel Jacobson, a writer here on Medium. While I enjoyed the reading a lot, I was feeling the need for more visualizations. So let me directly start with a quote of the just-mentioned article:
The selection of length/frequency is probably the first and foremost important factor of the realized volatility measurement. While we sometimes put little to no thought into the realized volatility length/frequency, this is far from being a trivial question. Let’s say that we want to analyze the 1-month volatility. Which lookback window should we look…
It was a lot of work but finally I managed to release a new alpha version of the pandas machine learning utilities and quant finance extensions. Here are the major changes.
Pulling a lot of dependencies which are probably not always needed was a draw back. Instead of introducing various optional dependencies I have opted for independent modules. This will also allow a diverging release cycle in the future and will allow to support different versions of 3rd party libraries like tensorflow 1 and 2.
The new modules are:
Usually, it is very tough to make a reasonable good model for financial markets. This is because one of the biggest problems with price time-series is the data normalization part. It should be rather obvious that we can not simply use the prices as they are. Not only because the magnitude is most likely outside of the preferred -1 to 1 range.
If developing a good model is difficult in general, it is even more difficult to build one to predict continuous outcomes vs. discrete ones. So let us first formulate a binary toy problem that we can use throughout…
Have you ever wondered if it is possible to draw these kinds of trend lines by an algorithm?
Well since this article is named Algorithmically drawing Trend Lines on a Stock Chart, you might already have guessed the answer ;-)
Now whether these lines are useful or not is a whole different topic. You probably know this saying: “Draw two random lines on a stock chart and you will see support and resistance lines”. On the other hand, people are using these lines so there should also be some kind of self-fulfilling prophecy component with them. The question is if…
Recent discussions whether the stock market is near to crash or not let me think one more time of Sornette’s Log Periodic Power Law Crash Indicator. I do not remember how many times I have implemented this algorithm in various languages. I am working quite a lot with Keras these days so I thought why can’t I just use Keras to fit the LPPL. Or even better just fit any function? Since Keras provides full vectorized math and CUDA support I would also expect some kind of performance gain.
I am a technology enthusiast and especially like challenges which are “not possible”. I like to reproduce interesting papers and to share my learnings.