raltin


Total Posts: 3 
Joined: Mar 2019 


Hi,
We stumbled this post from 10 years ago that asked for a tool that asks the question: === Hi guys. Do you know a tool to search the stock universe by correlation to a future? For example "search companies whose absolute correlation to the copper future is high/greater than x". http://www.nuclearphynance.com/Show%20Post.aspx?PostIDKey=133585#189300
===
We (Raltin) have exactly such a technology that does what the post asked for. It runs 2.5+ billion correlations across 40,000+ stocks, ETFs & mutual funds and shows the highest/least/inversely correlated stocks/ETFs/mutual funds for a given ticker.
Here are a few examples: (this example has the correlation charts as well)Mutual Fund: TRBCX: https://www.raltin.com/t/TRBCX/ ETF XHB: https://www.raltin.com/t/xhb/ Stock WMT  https://www.raltin.com/t/wmt/
You can search for any ticker and you can automatically get all the tickers that are highly correlated and inversely correlated.
Would love to get feedback on this forum on how you may be able to use this data/tool in your investing/risk management and other roles.
 Raltin




pj


Total Posts: 3434 
Joined: Jun 2004 


Yeah, and how are correlations calculated? What's the method? Assumptions? Outliers? Documentation? Or just trust us?

The older I grow, the more I distrust the familiar doctrine that age brings wisdom
Henry L. Mencken 

prikolno


Total Posts: 40 
Joined: Jul 2018 


+1 pj.
40,000 x 1,250~ as you're doing it is tractable in a few minutes on a commodity laptop. See this R package. You may find a more interesting use case if you're building a connectivity graph, because then a naive algorithm is quadratic in the number of vertices in the upper triangular of your output.




raltin


Total Posts: 3 
Joined: Mar 2019 


Hi PJ & Prikolno,
Thanks for getting back to us.
The core calculation between an any two securities (TRBCX & IWF) is a simple correlation/regression formula on daily closing price change. Here are the charts for this ticker which is transparent: https://www.raltin.com/t/TRBCX/
The challenge in what we are doing is not as much as the corelation calculations but the computation for ALL correlation for all possible combinations and then ranking. This requires a fairly sophisticated/ nontrivial computing resources and a lot of human/QC checking to see if the correlations are not purely junk. For e.g. In our shortlist of rankings, we exclude leveraged ETFs for certain securities since their correlations will be high but the hedge ratios will be very different.
Thanks, Raltin 

