Portfolio optimization of cryptocurrencies

Hello guys,

first things first. I am new to steemit and I want to introduce myself with this post.
I am a semiconductor physicist but I'm also very interested in financial markets.
Last Year I wrote an simulation about a portfolio optimization of DAX-stock and I was very surprised about the good output.

First I shortly explain the private project “Portfoliooptimization of the DAX-Data“ and then I want to discuss (with you) if it makes sense to apply this simulation on the cryptocurrency market.

The following program was developed in Shell Script and Python and features a
portfolio optimization program on historical stock data with simulations on the future.

A Shell Script was used to download the stock data from the finance.yahoo server. Once
modified it is executed in a time of a few milliseconds. It can easily be extended to
download every stock data of any arbitrary company or cryptocurrency. However, in this
project I just focussed on the German stock index (DAX).

Running the Python program the CSV-data downloaded from the Shell Script are implemented and afterwards the Simulated Annealing algorithm calculates an optimized portfolio to the chosen conditions of e.g. the investment capital or the selected risk. To customize the energy or cost function I become acquainted with the portfolio theory of
Markowitz. Hence, the following cost function was implemented:

Here, the quantity of a single stock is given by wi ∈ N. The first term of Eq. 1 represents
the chief constituent of the program, the profit one obtains by simply subtracting the
stock value at the start time (µt0) from the end value (µte). Since one is interested in steady increase of the deposits the second term represents the linear regression. To minimize the risk of a slump in prices of a portfolio consisting of stocks of only one special branch e.g. the automotive industry (BMW, VW, etc.) the third term was introduced.
The matrix A describes the correlation of the stocks among each other, which guarantees
diversification and thus risk reduction. The constant parameters C1,C2,C3, which describe
the significance of each term, has to be adjusted properly.

To verify if the program operates properly I calculated the portfolio for the past and
afterwards simulated the obtained configuration with the pseudo future. This pseudo
future are also stock values from the past but they haven’t been considered in the simulation.
Thus, one can check the result of the simulation, see Fig. 1. The following table
shows the simulation parameters.

Note that with sufficient computing power the swap value could be much higher to get
an even better profit. Figure 4 shows a graphical result of the simulated stock portfolio.

My question now is:

Do you think such a simulation makes sense applied to cryptocurrencies?

I'm glad to hear about the comments :)

If there are any questions in understand or if you want the source code feel free to comment !!!