This is a guest post by Graeme:
Babak recently wrote about the volatility put call ratio. This piqued my curiosity enough to ask him for the raw data to do a bit of backtesting to see if the VIX put call ratio provides any edge. All the tests you’ll see below were from July 2008 to the present.
First, I created some benchmarks to use as a comparison for further testing later on. This first test just buys the VIX on any day:
The first column is the number of days after the entry (n bar exit). The rest should be pretty self explanatory. All of the tests are done with Amibroker and as you may know, the payoff ratio in Amibroker is simply the average win/average loss. So any real tests run would have to beat these benchmark numbers to be even remotely interesting.
We have 637 days in the testing period. Just under 44% of the days in the sample are higher than the day before. And on average moves in the VIX are 1.4 times larger on up days than down. These numbers aren’t actually influenced by the massive spike during the financial crisis as much as you would expect. The VIX does generally spend more time going down than up, but the up days are a bit bigger than the down days.
Looking at the VIX put call ratio data from Babak, the first thing that really stands out is the daily numbers don’t oscillate up and down as much as I would have thought. The VIX put call ratio spends a lot more of it’s time showing relatively lower readings than higher ones (nearly 80% of the readings in the testing period are below 1).
A low VIX PC ratio isn’t necessarily bullish or bearish, it’s just what happens more often than not.
To start with, I’ll have a look at the day to day moves of the VIX put call ratio and see how the VIX trades afterward to see if anything interesting shows up.
Today’s PC ratio fell. Buy VIX at close. How volatility performs going forward on average for the next few days after the PC falls:
Very slight improvement, not even remotely significant though. And now for the reverse:
Really not much to report on there. There is a slight bias that shows that next day returns for the vix are better when more calls are traded the day before(and vice versa), but it’s not exactly an exploitable edge.
Maybe it’s a case of the readings’ need to be much, much more extreme before they start to make a statement. So firstly, let’s test low readings. What if it is below 0.5?
And finally, below 0.2%?
This one looks interesting. The next couple of days after an extremely low VIX put call ratio reading shows a small amount of upward bias in regards to the VIX rising over the next couple of days. It’s not an overly strong edge, but worth looking at a little closer.
Here are results for the 0.2 and below reading going out to 20 trading days (roughly one trading month):
Buying the VIX based on a low put call ratio reading (below 0.2) and holding for 20 days will give the following trades:
We have a profit of $155,000 on this test (assuming we use a $100,000 position size for a trade). However two of the huge winning trades came as the financial crisis really got rolling. It would be quite reasonable to assume that we won’t be seeing too many moves in the CBOE volatility index from 20 to 90 anytime soon. So after removing these outliers, we lose roughly $135,000 of the $155,000 profit, which basically returns the results back to a fairly random outcome.
Getting back to next day returns on the VIX, here is the equity curve for a next day exit following a reading of 0.2 or lower:
Pretty easy to see what has happened here. That first peak in the equity curve is roughly around March 2009. So it was possibly a handy little setup for next day returns for a few months during the bear market, but when that finished, getting long the VIX based on a low put call reading hasn’t exactly been a fruitful undertaking.
So it’s fairly safe to say that a one off, low reading on the volatility put call ratio is pretty well useless.
But, wait… what about high readings?
First, high readings aren’t anywhere near as common – roughly 20% of all day’s since July 2008 have been above 1.0). When the put call ratio is above 1.5 for the day:
VIX put call ratio above 2.0:
It would appear that there is a slight bias in the results here. But we have to remember that the sample size is pretty small. And made even smaller because the first few years had very low volume. So may be hard to have unshakable faith in the results going forward.
Here are trades from getting long the VIX after it spikes above 2.0 and then selling the next day:
I’ve organized the results by profit per trade. Once again what you see here is that a huge percentage of the profits are taken up by the largest couple of trades. Remove them and the whole story changes! Here’s the equity curve below:
What stands out here is that the equity curve basically tracks the moves in the VIX here. The put call ratio doesn’t seem to give us any additional, actionable information.
One day readings on the VIX put call ratio tells us pretty much nothing in regards to the future movement of the VIX itself.
Short Term Average of the VIX Put Call Ratio
Have a quick look at a rolling 10 day average of the VIX put call ratio. It is pretty obvious that things just don’t line up consistently. For details, see charts below, with the VIX turning points highlighted. The first chart is for 2008:
The next for 2009:
And for 2010:
In 2008 major peaks in the VIX are occurring at times when the put call ratio is at a low level, relative to recent readings. The first VIX bottom happened with a relatively high ratio, and the second one happened as it was turning up, but closer to a bottom. Pretty random stuff.
In 2009, the two VIX peaks had low put call ratio readings. The two troughs had high ones.
In 2010, the peaks had relatively low readings, with two out of the three troughs having relatively high readings. The third one was fairly random.
So based on some cherry-picking of the turning points in the VIX, what we are actually seeing is more calls being traded at times when the VIX is peaking.
The other problem here is that most of the readings are “relatively” high or low at some of the turning points. It doesn’t look there are any hard and fast levels that will immediately give us an idea about future direction of the VIX.
So if we see a “relatively” high or low value at vix turning points, then let’s try and code something up that might pick this type of pattern up a little:
Today was the lowest close in 20 days on the vix, and the VPCR was at a 20 day high reading:
A relatively small sample size, but the results do look promising. I also ran a test for the reverse conditions, but couldn’t get much of a sample at all.
One day readings of the VIX PC ratio are useless!
What might be worth tracking is a high PC ratio reading relative to recent data combined with a low VIX, but even then it’s a bit of a stretch.
I don’t think I’ll be tracking this going forward!