The VIX is Very Predictable
In The VIX isn’t Magical I showed that much of the VIX is comprised of information we already know: price changes and past volatility. Here I’ll show that because these two factors are somewhat predictable, the VIX itself is extremely predictable.
But here’s the important part: if we can demonstrate that the VIX is highly predictable, and we already know that the VIX includes a lot of information about what’s already happened in the market, you have to wonder: does that then mean that by predicting the VIX we can predict the markets?
The strategy: Buy the VIX at today’s close if the 11-day EMA of the VIX crosses under the 11-day SMA. Close that position and short the VIX when the EMA crosses over the SMA (see geek note 1). Yes, I realize that you can’t directly trade the VIX (see geek note 2), but bear with me - I’m trying to demonstrate predictability, not tradability.
Results (01/1990 – Present)
The graph above shows the VIX index (dotted purple line) and long (red), short (green), and long/short (blue) versions of our strategy for “trading” the VIX.
As you can see, the VIX itself doesn’t trend very well. It tends to oscillate around some general value. But despite that fact, look at the “gains” we were able to pull from it. That’s about as predictable an instrument as you’ll find. This is purely a theoretical exercise, but just for giggles – that’s a 7,397,408% gain in 18+ years.
We asked at the beginning of this post, if we can predict the VIX, does it then mean we can predict the markets? In a future post, I’ll try to answer that question with my own idea for using this strategy to trade the S&P 500.
Happy Trading,
ms
P.S. to Mr. Luby at VIX and More – I hope that I’ve redeemed our little oscillator after my less than flattering post last week. We (the VIX and I) have an on again off again analytical relationship =)
To stay up to date with what’s happening at the MarketSci Blog, we recommend subscribing to our RSS Feed.
Geek Notes:
(1) There are two generally accepted ways to calculate an EMA that produce slightly different results. Here I have used the ((1/Period)*2) method. If your charting program uses the (2 / (Period + 1)) method, simply reduce my period by one. For example, if I’ve used an 11-day EMA, the alternate EMA would be a 10-day EMA.
(2) There are futures and options available on the VIX; however, they’ve proven to track the actual VIX very poorly. If they didn’t, I wouldn’t be posting this strategy, I’d be printing money.
Filed under: Trading Strategies, VIX & Volatility | 12 Comments
Very interesting ideas. waiting for your future trade where you put these ideas to test on trading S&P.
Michael,
I use BioComp Dakota to increase the degree to which my models adapt to changing market conditions. I’d like to add some R&D that I hope your readers and yourself will find interesting and possibly useful. I have built two Dakota systems based on the application of the VIX model to the SP futures contract.
The systems use reverse adjusted continually linked SP futures data sourced from Pinnacle Data Corp. The VIX (S&P 500) data is from the Pinnacle IDX database. Hypothetical trades are opened and closed MOC on the trading day following the current trading day. No transaction costs are included i.e. frictionless trading.
The hypothetical historical performance of both systems is measured from 3/16/1997 to 10/7/2008. I did intend to include images of the equity curves produced by both systems, but I don’t know how or if I can insert them here.
The first system has the EMA and SMA periods fixed at 11 trading days. The equity curve is decent enough. A total of $533,312 in hypothetical profits was accumulated over the period.
The second system has both the EMA and SMA Min periods set to 8 and the Max periods set to 14. I subtracted 3 from 11 to come up with 8 and added 3 to come up with 14. This gives the system the potential to adapt to changing market conditions. The second swarm produced a hypothetical profit of $695,962 and the equity curve was smoother.
As Dakota ‘walks-forward’ through the data from March 1997 up to date, it calculates the trading signal by averaging the signals from 35 ‘trade bots’. A trade bot is equivalent to a particle i.e. Particle Swarm Adaptation. You can define the number of trade bots to use. I usually use 35. Each trading day, any given trade bot can potentially have any combination of parameter values that fall within the parameter ranges. For example, trade bot number 1 on trading bar number 1345 is using an EMA period of 10 and an SMA period of 11. On trading bar number 1346 trade bot number 1 is using an EMA period of 11 and an SMA period of 11.
Two key components of Dakota are the Equity Engine and the Swarm Adaptation Engine. The Equity Engine measures the performance over the ‘Lookback Period’ of each trade bot on each trading day walking-forward. This information is passed to the Swarm Adaptation Engine. The job of the Swarm Adaptation Engine is to calculate new ‘positions’ for each trade bot within the parameter space for the trading day being processed. For the second system above, the parameter space is 8 to 11 for both the EMA period and the SMA period. I have built my own Swarm Adaptation Engine based on studying papers written by experts in the field that ‘plugs in’ to the Dakota application. Some original concepts of my own are in there too.
In summary, if you have a decent model to start with, then you can improve upon it by making use of particle swarm adaptation. I have been asked the question “Why not just calculate the best parameters over the lookback period and use those each trading day walking forward”. More often than not, the historically ‘best’ parameter values are not the best ones to use. Swarm Adaptation is a step up from this more basic type of curve fitting. Swarm Intelligence really does exist and it can work for you!
All the Best,
James
RE to James: I haven’t heard it referenced by the names you’ve used above, but this general concept is one that I’ve used in other adaptive systems I’ve worked on. I think it’s a terrific concept if each of those alternate lookbacks (i.e your 8 to 14 day range) show some persistence in beating the other combinations once they rise to the top. In other words, if 14 days has been the best lookback over some period of time, then is it more likely that it will continue being the best for at least a little while in the future? Or is the best combination unpredictable and non-persistent. I think the answer to that completely depends on the asset traded and the indicator.
Thanks for the great comment!
michael
[quote]if 14 days has been the best lookback over some period of time, then is it more likely that it will continue being the best for at least a little while in the future?[/quote]
Apologies for the late reply. I would suggest that the optimal parameter value will drift. For example, sometimes it might remain around 11 for a couple of years and then drift down to 9.
Another consideration is that what has been the absolute best set of parameter values over the lookback period (say the last 50 trading days) is unlikely to be the best over the next 50 days. However, often the best parameter values over the next 50 trading days will be in the vicinity of what were the best parameter values over the lookback period. So if the particles (in particle swarm adaptation) are surrounding the optimal particle in a ‘cloud’ then it is more likely that we will ‘catch’ the best set of parameter values moving forward. This helps the swarm to continue to track a decent set of parameter values moving forward.
Thanks,
James
Michael,
You can profit from predicting volatility using option straddles. Predicting volatility = predicting the VIX. Am I missing something?
Regards,
Max
RE to Max: in theory yes (or how about just applying it directly to the VIX options), but I haven’t dug into that whole world yet. My gut feeling is that “inefficiency” (because it is so glaringly obvious) is already traded out of the VIX options for instance.
P.S. appreciate all of the thoughtful comments on the blog lately. I’ve added your own blog to my rss reader and am following you (not sure if i mentioned that earlier).
Happy New Year!
michael
Very interesting that this contrarian-type trade has worked historically. Probably hasn’t done so well lately, with the VIX demonstrating more trend-like behavior?
A couple questions: how is the starting equity for each trade hypothetically calculated? For instance, if the rule says “buy VIX at 20,” then when you close out the trade, are you basing the return off “(closing price – 20)/20″? Thus accumulating equity in such a fashion?
Another suggestion to improve the Sharpe ration on this trade drastically: enter long positions only when EMA is a certain amount under SMA, then close it on the first day EMA crosses over SMA. Enter a short VIX position only when EMA is a certain amount above SMA, then close on the first day EMA crosses under SMA. In this way, you will be out of the market for periods at a time, and your drawdowns should be reduced drastically.
RE to LyricalOne: with the exception of the crash late last year (October, where most contrarian indicators fell apart) I don’t see any evidence that this contrarian-tendency in intermediate timeframes is abating. The version that we track on the State of the Market report has been doing very well.
RE “how equity is calculate” – it’s based on the closing value of the VIX that day (i.e. buying at the close). Of course, this is just a mental exercise as the VIX index can’t actually be traded (through futures/options it can, but like the SPY vs S&P 500 index, the tradeable product is an entirely different animal).
RE your variation – I agree – I did a post similar to your idea a while back here:
http://marketsci.wordpress.com/2008/09/22/test-of-condor%e2%80%99s-vix-based-trading-strategy/
Thanks for the thoughts.
michael
I must be missing something. Just eyeballing a 1 year chart, this strategy would be a disaster. Almost exactly the opposite of what the author predicts. Guess this is what LyricalOne meant by trend-like behavior. We must be mis-interpreting the author’s strategy.
Hey michael,
Very interesting post.
I went through the example with daily VIX data, and although it is a bit irrelevant, I think your return math is slightly off. You say there is a 7.4 MM % return, which would mean 1 dollar would grow to 74K, but it looks like your graph suggests 1 dollar grows to nearly 10MM (perhaps 7.4MM?), which would be a 1B% return over 18.6 years, which is something more like a 205% return.
Cheers,
Jon
RE to Jon: Yep, should read 739,740,843% total return (meaning the annualized figure is correct). Good catch! michael