SPX/XBT correlation: It's meaningless w/o context

25 thg 4, 2018

I'll finish everyone's thought on this topic with something you can actually execute on.

Correlation coefficients are relative to two dimensions: Time and periodicity. Its important to understand that when plotting between these correlations, because in this case you could be missing over 50% of information. Bitcoin trades on a 24/7 market, while SPX/QQQ/NI225 or any other legacy index trades, at the most, 5 days a week, at the most 12 hours a day. Bitcoin keeps moving after your legacy market's closing bell to opening bell, so by the nature of sample data alone we can say that they Bitcoin is not correlated with legacy purely by the way that most market technicians are measuring it by: Periods of <1 day, against instruments that trade < 50% of the day.

That being said, sample data is sample data, it doesn't have to be perfect, but it is important to be cognizant of when thinking about the best way to digest and interpret any information. Thankfully, there are index futs, like your variations of /ES /YM /NY and so forth. those are much better data points to plot against because at least they trade 24 hours a day 5 days a week.

Again, nothing is perfect, so for this example i have elected to use the SPX against XBT to demonstrate not only how to interpret the data but also what it means and how it can change.

This is going to be a multi step post due to the need of multiple examples. Please comment if you have any questions. Enjoy.

25 thg 4, 2018

Ghi chú

Shown below is the first step that you need to understand when understanding what a correlation actually means: Correlation is plotted against SPX (again, for the examples sake).



You can see the variability between different periods of the correlation coefficient, giving you not only different values, but more volatile or dulled signals.

What actually matters? Which one should you choose?

The short answer is none of them, while the long answer is all of them. I know that answer is going to kill a lot of you, but again you have to understand what you are measuring:

BTC SPX
----- x ----
USD USD

You're measuring the relationship of the dollar (USD) between the two. And as such, it does make more sense to use a signal based off of the dollar. The original example shows, instead, of looking at the correlation between XBT and SPX, what would happen if we instead analyzed the relationship between XBT and DXY. We kept the default setting at 20 periods, and what we got was much more actionable than the original XBT against SPX

25 thg 4, 2018

Ghi chú

But this begs the question: Why?

If we subscribe to an theory that SPX, and in general Global Macro, does effect Bitcoin (as long as you can trade it w/ fiat pairs, it does) doesnt it make more sense to look at the drivers of the dollar too?

Because at this point in time, SPX is derivative of USD, and we've found a little more success with DXY.

So what if we looked at what DXY derives from: Interest rates and money supply

Below is a quick study plotted against TNX on the default period (20):



it doesnt seem like there is anything here.. until we start to apply some mean reversion strategies:



When the ten year yield TNX moves +0.5 on the correlation coefficient, a long opportunity presents itself.

This doesn't mean that when interest rates go up that you should should go long on bitcoin, This merely suggests that when bitcoin moves with treasury yields that you should look to go long. Interest opens up another dimension of strategic capital deployment, due to their ability to change market regimes. This insight is interesting, and so far actionable, but it ability to deliver consistent results is not proven. Do not assume this relationship will exist forever or possibly ever again

26 thg 4, 2018

Ghi chú

here is another idea on how to use this information, overlaid w/ gold on the daily, and we have a pretty consistent filter for a long only strategy: