On Friday, August 17th, $VIX made a new 1000 day low. Financial journals broadcasted this event, and headlines such as this and this are making the rounds. What can we take away from these articles?
…the VIX has spent over half of its time over the past two decades (from 1992 through Tuesday) between 10-20. So the level it’s at today is very, very normal.
“Be careful if you think the VIX has nowhere to go but higher”….the VIX has a history of remaining depressed during long periods of time — like they did between 2004 and 2007 when stocks slowly drifted higher.
Okay, easy enough. What we are witnessing with $VIX is not abnormal, but that isn’t information isn’t specific enough to help traders and investors. Let’s break the $VIX history apart so that we have information that is actionable.
We are going to start by looking at what happens when $VIX crosses beneath various moving averages. It is currently trading beneath its 10, 20, 50, and 200 day moving averages. We would expect this with it making a new 1000 day low.
The rules are simple: Buy $SPY at Close when $VIX Closes Beneath its X Day Average.
$SPY Buy-n-Hold is calculated by taking all $SPY history, breaking it into 50 day segments, and averaging the segments.
When $VIX is beneath the shorter (10, 20) moving averages, $SPY tends to track or slightly underperform its historical average performance. This is likely due to short-term mean reversion: as $VIX dips and stretches farther beneath the shorter moving average it will reverse for a brief period of time.
When $VIX is beneath the longer (50, 200) moving averages, $SPY tends to outperform its historical average performance. As $VIX stretches farther and farther beneath these longer-term averages, $SPY has tended to trend higher in a low volatility environment, enabling the outperformance.
Let’s dig deeper and look at the percentage of winning trades.
There have been 69 trades made¹ when $VIX crosses beneath its 50 day average. 68.12% of those were higher after 50 days.
There have been 55 trades made when $VIX crosses beneath its 200 day average. 60% of those were higher after 50 days.
The bottom line is that when $VIX is beneath its longer-term moving averages, $SPY has tended to outperform its historical average performance and there is a better than average chance that $SPY will be higher 50 days later.
The next post will look at what has happened after $VIX has made a new X day low.
¹Trades held the full 50 days. There are more than 69 trades made if each is not held the full 50 days.
5 Stars.
Thanks BRA.
Hey Woody,
MIT researchers developed an algo that helped a small robotic plane negotiate an underground garage at relatively high speed using just laser range finders and roughly the processing power of a tablet computer. No big picture GPS was involved.
Wondered if this might be relevant to your work, as these guys are essentially plotting future movement using a limited and rapidly changing set of short term data points.
And you guys are both working to avoid crashing. Hope it’s interesting in any case: http://bit.ly/NOK5LE
Here’s the paper about the algo itself.
It’s got that kind of math what has dead language letters instead of numbers and makes my head hurt: http://arxiv.org/pdf/1105.1186v1.pdf
lol…that is really cool, save for the dead language numbers…Thanks for posting it. I like your idea about adapting it to stocks. Will have to give it some serious thinkage.
Isn’t the conclusion of that plot that there is NO edge? Maybe my eyes are worse than the raw statistical analysis, but I would swear B&H looks like a smoothed version of all of those curves. What am I missing?
Technically, B&H is a smoothed version of those curves, or more specifically, those curves are part of the B&H curve.
What would you define as an edge?
“The bottom line is that when $VIX is beneath its longer-term moving averages, $SPY has tended to outperform its historical average performance and there is a better than average chance that $SPY will be higher 50 days later.”
I was confused by this conclusion, as the plots suggests to me that SPY tends to perform exactly in line with its historical performance regardless of the $VIX reading – rather than “outperform.” Hence, there is no apparent improvement in economic outcome (EDGE) by using this information. As I said, maybe there is something buried in the statistics I am missing. I was actually expecting to see something more like your conclusion suggests – I just don’t see it in the plot.
Zach, more statistics might help.
SPY performance over the average 50 day segment is 1.54% with a 62.89% win rate.
With VIX < MA50, the average SPY trade over 50 days is 2.00% with a 68.12% win rate.
Of course if we look out farther, say 100 days, this can change. My next post will look out farther.
I read in the newspaper today that the drop in the VIX was a ominous sign, because it will create a elastic effect if the announcement isn’t dovish. A rare sort of opinion to be reading in a newspaper I thought.
Might we get a look at what happens when VIX meets above listed conditions and TLT is trading below its declining 20 and 50 ma? Pretty please? BTW great post I shared it with some people.