Financial model of Efficient Frontier for 2k of NASDAQ companies and funds for 2013-2014. We will see a theory in reality, see some interesting stuff beyond (Median vs. stdev) and as always enjoy a beauty of the data (e.g. “NASDAQ Fish”).
Reminding a bit a course in economics, I wanted to play with some amount of financial data, so I have extracted 2088 NASDAQ stocks/indexes since beginning of 2013 (done at mid of March 2014: 2088x250days = 522000 of values).
If you did not have a chance to hear about Efficient frontier, you can read about this amazing theory, developed by Harry Markowitz that got Nobel Prize. Bottom line – taking gains versus risk is giving max gain per risk, nicely below some boundary, called “Efficient frontier”. Even simpler: You cannot get higher gain without raising a risk level. Even Simpler: Want more – play harder.
Let’s take a average of daily gains versus stanard deviation of gains… Kind of efficient frontier baseline. See – it is on the line. Nice.
The standalone index is CombiMatrix Corporation – publicly listed warrants (CBMXW), while CBMX itself is within. What the #$%@ is going on there? Hm… Is it fraud? Is it mistake? Is it anomaly? It’s worth a check. Some people just recently did 4000% on that.
Zoom-In to get the NASDAQ fish (“topological” lines are non-par density of JMP to expose the percentage “from above” – beautiful, right?):
A small notice – yes, I did not take dividends as our fellow Markowitz would suggest (feel free to do it).
Interesting observation is that gain MEDIAN values are giving a different perspective. While 40% of stocks got median value of zero median gain, about 50% are positive and about 10% below zero:
So looking at Median versus average gains you can see that there are two types of behaviors – the one with zero median (i.e. precisely half of gains are positive and half are negative, even though average is big due to long tails) and those with non-zero median (mostly positive with good adjustment to average):
Interesting? We see a clear distinction between those two groups. What is unique about each group? Commonality is not about type (e.g. Stocks, ETFs…), nor the sector or industry:
Drilling down further we will see that within non-zero median there is a high degree of cross-correlation.
We will run the cross-correlation and will build a “NASDAQ network” of high-correlation indexes.
But that’s in the next time! From the teaser you can imagine who is going to be extremely connected: