Coming Soon: Volatility ETF Strategy
As I occasionally talk about on the blog, I’ve been successfully trading volatility ETFs like VXX and XIV for going on two years now, and about 9 months ago I began tracking my returns in a standalone account (see my verifiable performance).
[actual real-time verifiable results through 07/2012]
I had intended to open up the strategy to subscribers last week, but it seems Interactive Brokers made a change in how they handle conditional orders (something the strategy makes heavy use of).
That’s pushed me back a few weeks as I test out a fix. Annoying yes, but better now than with new subscribers on board.
In any case, I’ve been getting a ton of inquiries as to when the strategy is going to be opened up to subscribers, so this is a heads up that it’s right around the corner.
Watch this space for updates, or send us an email and ask to be notified when we’re officially live.
This is the first strategy I’ve made publicly available since the YK Strategy. YK crushed it for over 3 years before the daily momentum/mean-reversion play petered out and I retired the model in October, 2011.
I’m happy to be able to put something out there again for readers, as always with actual real-time verifiable results, and especially trading such new and exciting products like VXX/XIV.
More to follow…
Happy Trading,
ms
. . . . .
To stay up to date with what’s happening at the MarketSci Blog, we recommend subscribing to our RSS Feed or Email Feed.
Filed under: Random Stuff | 19 Comments
Could you also share details on the risk/return relation, e.g. what’s the Sharpe ratio of the strategy? As you will have noticed yourself, the drawdown is big even compared to the SPY benchmark. Could the strategy still be interesting for mid- to long-term investors or would you say it’s only suitable for less risk-averse speculators? I’m asking these questions because I’ve been developing a few trading models myself, and so I’m always interested in the thought process that went into other people’s models. Thanks!
Where to begin…
1. I highly doubt folks with experience in real world trading for any length of time would say that a max peak-to-trough drawdown of HALF the annualized return is big. Both of the strategies I’ve previously made public, the original MarketSci and YK strategies were at one point the top dogs in managed account land, and both breached that level.
2. Is the program appropriate for risk-averse investors? You asked the wrong question. VXX/XIV are almost 3 times as volatile as the S&P 500, so returns (and volatility) are inherently magnified relative to that benchmark. So no, putting all your eggs in that basket doesn’t make sense for anyone (but neither does putting it all in the S&P 500). I say you asked the wrong question b/c the real question is what % of the total portfolio you should allocate to the strategy given your level of risk-aversion.
3. RE: Sharpe Ratio. If you mean real-time, it’s included in the link provided (1.34): http://www.marketsci.com/strategy.VT.html
If you mean backtested, I don’t release backtests for my proprietary strategies (ever). Investors should judge the performance of a strategy on it’s real-time verifiable results. I don’t have enough fingers and toes to count the times I’ve seen someone in this business wow folks with a sexy backtest, only to come up short. I can only say that the strategy has lived up to my expectations in real-time trading.
michael
As it happens, I’ve been trading a VXX/XIV strategy through IB for a few months, but have also had recent problems with conditional orders. Warning, if you use the VIX index as a condition, it only triggers every 15 seconds.
Actually, I missed a great (9%) gain over the last week, because I could not get shares to short before a large drop. This trade is getting crowded!
Hello Zack – thanks for the thoughts. See response to Jim below. michael
Curious, what change are you referring to at IB?
Hello Jim – the strategy uses conditional market-on-close orders (so that users aren’t required to sit in from their computer at the close).
You can’t execute a true conditional MOC because the exchanges require the MOC (without conditions) to be submitted X minutes before the close (I don’t recall the actual number of minutes…15?)
In any case, you can mostly work around that limitation by using a conditional market order through IB that doesn’t begin until say 10 seconds before the close.
That worked perfectly for well over a year, until the last week or so. The conditions aren’t being satisfied and the orders aren’t executing. I’m testing starting the orders earlier (but still as late as possible) because I think the problem is related to what Zack mentions above (problem with the time interval the index is reporting).
Missed trades were painful last week as the model missed out on what should have been a juicy XIV gain. Grrr…
michael
Just a follow up on the closing issue orders. It took 3 days of back and forth with IB support to track this issue down. Specifically, their conditions appear to only trigger on a new tick – not on the last reported price. I had mostly traded ultra-liquid instruments before this, but would occasionally have a trade not trigger and never knew why before. Since the last ^VIX tick before the close is 14:59:45 CST, any order after that time will never fire.
I would rather leave my orders on their servers, but I may well have to shift to sending them from my automated platform.
Hello Zack – what’s strange is that I had been using 10 seconds before the close for a long time and never had an issue. In fact, the trade always showed as executed at the same second the order went into effect.
I’m wondering if that 15 second update sometimes doesn’t happen on the quarter minute (and/or has changed over time). For example, I’ve done tests in the last week where I set 1 share to buy/sell at a given minute and it now executes at say 8 seconds after the minute.
I’m now testing setting multiple 1 share buy/sell “conditional MOC” orders at 5 second increments from 13:59:30 – 13:59:55 to get a handle on this thing.
Thanks for the thoughts. Good to hear from someone else who has been having this same problem!
michael
Let me know if you find something. I used to have ES futures orders that would begin at 15:14:58, and they would always trigger, but occasionally DOW 30 stocks would miss a trigger at 14:59:55 (CST). It does seem like the misses are more frequent now, but volumes are WAY down across the markets. It is entirely possible we are just seeing a random skew that has always been around, but is just now coming to light. Good luck!
Buying and holding XIV from 1 Nov 2011 would have turned $25,000 into $64,000 (as of today). More volatility than your strategy though.
Hello Nick – buying & holding XIV would have outperformed the strategy over the last 9 1/2 months. B&H volatility and drawdowns would have been significantly higher, but B&H has still outperformed on a volatiiltiy-adjusted basis.
The problem is of course that XIV has been on an unusually strong streak and when it turns (and it will turn), those B&H gains are guaranteed to be wiped out:
http://marketsci.wordpress.com/2012/04/16/trading-volatility-etps-obey-thy-stops/
michael
Let me explain myself and the flaw in your system more fully.
There are two components to the XIV return – the roll yield and VIX return and I see that you are trying to get them both. Else you would just follow the strategy of holding XIV when the futures market is in contango, VXX when in backwardation.
The VIX is predictable but the VIX futures take this into account. So VIX futures are very unpredictable. The VIX is very variable so if you CAN predict the VIX futures then you will have very little return and a LOT of volatility.
On the other hand the roll yield is very predictable and not so variable. So when you combine this element of the strategy with the VIX element the VIX element gets VERY LITTLE WEIGHT.
Therefore, optimally, you should only go for the roll yield. That means B&H since November because the market has been in contango.
Also it’s better to look at the shape of the futures curve to decide whether to put in a stop or not rather than the past of XIV. The market can go from contango to backwardation quickly but any stop system based on XIV technicals will have a lag.
Thank you for gracing us with your sweeping, all-encompassing conclusions (and super prickly demeanor…but more on that later).
First, most folks (including myself) don’t have the intestinal fortitude to simply hold XIV/VXX when the futures are in contango/backwardation.
It’s very easy to sit on the other side of this little bull run and proclaim that you should have been long and strong VIX, it’s an entirely different thing to build a system that’s going to work for the coming years.
A simple test back to 2004 of blindly following the term-structure shows excellent terminal returns, but incredibly painful drawdowns/volatility (regardless of whether stops are used).
So I think following the VIX futures term-structure makes sense (and it’s something I do), but a bit more sophistication is required.
Second, you’re correct that the VIX futures compensate for the obvious predictable elements of the VIX – but that shouldn’t be a surprise to anyone – that’s the nature of markets.
Our job as nerds is to find those elements that the market hasn’t accounted for, no differently than we do with any other trading vehicle.
And I would submit to you that anyone that’s done the legwork understands that the VIX futures do not account for all the predictability of the VIX.
Lastly, while I appreciate smart discussion, your points are utterly lost when you come over my house pontificating about how you’re going to “explain the flaw in my system”.
I very much appreciate your thoughts, but leave the chest puffing at the door. We don’t do that here.
I’ve been eagerly anticipating hearing more on your new system ever since you started talking about it several months ago. It’s very interesting, and congrats on the live trading success it’s shown so far. As we both know, backtesting and live trading are two very different animals.
I’m sure both of our systems share some commonalities, but my parameters seem to be a lot more relaxed then yours as I find myself in many more trades each month than your system shows you were in. On the flip side though that may mean I will see deeper drawdowns during less than ideal times, but only time will tell.
Anyway, my question is regarding your choice of benchmark. Do you really feel the S&P 500 is the best choice for measuring your systems success? As was mentioned before, a buy and hold on the XIV itself has returned about 160% since last November. Now obviously we’ve been in the most ideal time period since 2004 for a long XIV position so it’s hardly predictive of future success.
However, shouldn’t a system that trades the XIV be measured against the XIV in some way or another?
For me personally, the benchmark I try to stay ahead of is a B&H on the XIV itself, adding a single parameter: Long when the 1st and 2nd month futures are in contango, cash when they are not. It seems to me that anyone trading the XIV on any serious level will at least know that much.
Perhaps that would be a more accurate way to measure the success of your system? If you can take that benchmark, which is a clear winning strategy over time, and apply your “geeky research” as you like to put it to not only improve returns but also reduce risk, you’d really have something.
Just some thoughts… Great to see this thing finally rolling for you though. I’m sure many people like me are following along with great interest.
Hello Brent – I’ve given this a bit of thought. A rundown of some options…
B&H XIV: inappropriate because when volatility turns it’s going to fall through the floor and make for a useless benchmark.
30-day VIX futures price (mix of front/second month prices that maintain a constant 1 month maturity): fine for short-term but not long-term benchmarking (due to mean-reverting nature of VIX).
Anything that resembles a strategy (i.e. your idea above, VQT, XVIX, XVZ, etc): benchmarking a strategy to another strategy is a little wonky. Fine in theory, but when the “benchmark strategy” starts making bad trading decisions, it calls into question the whole benchmark (I would suggest a simple test back to 2004 of a “blind” long XIV when contangoed/long VXX backwardation to see what those bad decisions would look like).
And finally, the S&P 500: imperfect, but as long as we’re comparing returns on a volatility/risk-adjusted basis (which are the only comparisons that matter), I think the S&P 500 makes the best benchmark for the long-term.
First off, ignore the difference in returns. Returns are meaningless – they’re a function of exposure and leverage. Vol/risk-adjusted returns are the only returns that matter. If we simply wanted to equate the vol of the two instruments, I would show the S&P 500 with leverage.
Volatility ETPs are risk assets that are highly correlated with equities, and like most strategies using such assets, the de facto benchmark is the S&P 500.
The “normal” state of the VIX futures in contango and the “normal” state of equities is bullish, so more often than not, vol trading strategies are going to be positively correlated to the S&P 500.
The downside is that the S&P 500 doesn’t capture the impact of the term-structure, but that’s part of the alpha on top of the highly correlated beta that these strategies are providing (timing entries/exits being the other part).
Again, imperfect, but if we ignore simple return for volatility-adjusted return (which we should all be doing anyways) I think the S&P 500 is the least bad choice as a long-term benchmark.
michael
Sorry to say that but so far it seems that your equity follows the market with a beta of 2 ~ 2.5. Do you have an estimate of the alpha of your strategy ?
Thanks,
Jean
RE to Jean: I’m surprised it took so long someone to make that comment.
VIX futures have been contangoed since I began tracking the strategy. That means trades have been biased towards short volatility. And that means that returns are going to be strongly positively correlated with the broader market.
When the next market crises occurs and futures move towards backwardation, the opposite will be true, and the strategy will have an inverse relationship with the market.
So in short, no, the strategy does not follow the market and over time correlation will fall significantly.
michael
All good points, and if it was any other trade vehicle besides XIV I would agree with you 100%. However this time, it may actually be different. The XIV is the first trade vehicle ( I dare say ever ) that allows regular people with very little knowledge of the markets or active trading to make outsized returns for the long term in a fairly predictable and repeatable way.
You mentioned that the benchmark I use, which is the long only XIV position when VIX futures are in contango would not have done very well if we go back farther then just this year. I guess it would depend on how you measure success, but in my opinion money is money, and that strategy actually would have made a significant amount.
A 10,000$ investment in XIV in March of 2004 would now be over 480,000$. Of course drawdowns are insanely huge with that simple strategy. That person would have been thrilled from 2004 to 2007, turning his 10,000$ into 130,000$ in just 3 years. The next 2 years would have been gut wrenching, taking him all the way down to 30,000$ again. Still pretty good returns actually, tripling his money in 5 years, but painful considering the highs he came from. If he sticks with the program, in the last 3 years his 30,000$ would have nearly touched half a million.
When the next crisis hits of course he’ll give a lot of that back again, and so on. As long as the financial system as we know it doesn’t dramatically change in the near term, that person can expect those up and downs to continue, but overall he can also expect to leave the S&P 500 returns so far in the rear view it hardly seems worth mentioning them in the same sentance.
However terrible the sharpe ratio is with that benchmark, and however painful those drawdowns would have been, it doesn’t change the fact that the strategy made an absolute killing over the course of the past 8 years. A simple matter of choosing a good entry would make the returns even better.
Which brings us to the point of “actively trading” the XIV compared to just B&H. It’s actually pretty easy to smooth out those returns by just applying a few extra filters to the trade. It’s also very easy to increase those returns, again just adding a few extra money making filters like occasional VXX positions, or using margin during low volatility times.
The real holy grail for XIV traders is going to be BOTH increasing those basic B&H returns, and smoothing out the drawdowns so it’s easier to sleep at night.
Shouldn’t we as active traders of the XIV be trying to take what regular people with very little financial knowledge and experience can do, and improve on it?
Shouldn’t our self made systems be able to backtest returns larger than turning 10,000$ into 480,000$, and also be able to do it in a much smoother fashion? And shouldn’t we also be able to live trade returns going forward that are also smoother and larger then that simple B&H strategy? If we can’t, there isn’t much point actively trading it. We mine as well just stick a portion of our portfolio in the super high risk catagory of Long XIV and ride the ride.
RE to VIX Trader: those are all fine points, but that isn’t a benchmark.
I traded YK for over 3 years, a long/short S&P 500/Russell 2000 program.
Did at any point anyone suggest I should benchmark my returns to say a trend-following strategy because that’s what “regular people with very little financial knowledge can do”?
No, that would be silly, b/c a benchmark is never a strategy. It’s a PASSIVE investment with postiive expectancy that bears some relationship with the thing you’re trading.
And yes returns are the only thing you can eat, but returns are NOT what matters when judging a strategy.
Returns are a function of leverage and exposure (I’m going to tattoo this on my head b/c too many folks seem to forget this point). Want more return? Increase exposure. Does that make sense to do? I don’t know, check the only stats that matter: returns relative to volatility/drawdown/risk.
michael