A Critique of TradingMarkets.com’s VXX/XIV Strategy
This is a follow up to TradingMarkets’ A Low Volatility Strategy for Trading High Volatility (h/t The Whole Street) in which they describe a strategy for trading vol-ETFs VXX/XIV based on mean-reversion and the short-term indicator RSI(2).
Results of TradingMarkets’ strategy from 02/2009 – 05/2012.
[linearly-scaled, growth of $1, frictionless]
TM’s strategy: Divide portfolio into 6 units. Buy 1 unit at the close when RSI(2) of VXX closes above 90. Add 2 additional units if VXX closes above your entry price at any point. Add a final 3 units if VXX closes above your second entry price at any point. Close all positions when RSI(2) closes below 50. TM’s post assumes we shorted VXX, but this is unreliable in the real-world, so I’ve assumed that we instead went long the inverse ETF XIV.
Note uber-sexy results. TM says, they’ve “never seen numbers like this in equity trading.”
One BIG problem: TM’s strategy aggressively adds to the original position as the market moves against the trader, repeatedly doubling-down on a bad entry. That’s fine for well-behaving markets, but how would this approach have fared during times of real market stress, like 2008?
Luckily, readers know that we can estimate VXX/XIV returns with pretty darn good accuracy all the way back to 2004 (read more), so below I’ve shown the same strategy starting in 03/2004:
[logarithmically-scaled, growth of $1, frictionless]
Not nearly as impressive. Only a 9.2% annualized return with a whopping -32% max drawdown during the 2008 crises. Not a horrible result, but not the Madoff’esque numbers the initial test would imply.
This revised test demonstrates the weakness of this type of extreme overbought/oversold strategy. It works great until it doesn’t, and then it lops off your head. That’s especially true when trading vehicles like VXX/XIV that can move so far in such a short span of time.
Further, the strategy ignores an important characteristic of VXX/XIV: the always important impact of the VIX futures term-structure (but that’s a subject for another post).
IMHO, this strategy is a dangerous one. Extreme OB/OS strategies, that might be appropriate for equity indices or even individual stocks, are not appropriate for portfolio killers like VXX/XIV.
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: Trading Strategies, VIX & Volatility | 12 Comments
I wanted to leave a comment on their blog, but I quickly realized there’s a reason I can’t.
This is essentially martingaling, there’s nothing else to say about it.
Hello Michael, Is there any way you know of to apply these VXX/VXZ (etc) type system signals to SP500 timing using index funds (ETFs or end of day priced mutual funds) as vehicles? If so, what type of results have you seen?
Thanks for sharing your work. Jim P.
Hello Jim (long time no talk) – IMHO, a good VXX, XIV, etc strategy is going to take the VIX futures term-structure in to account (contango vs backwardation), which means that the resulting signal is going to be less effective trading equity indices than just analyzing the equity index itself (because you’d be including something in your analysis mostly unrelated to the performance of the equity index).
More:
http://marketsci.wordpress.com/2012/04/12/strategy-2-for-trading-volatility-etps-timing-the-vix/
http://marketsci.wordpress.com/2012/04/11/strategy-1-for-trading-volatility-etps-term-structure-following/
Now I do think analyzing the VIX index and trading an equity index might make sense b/c then you’re stripping out the impact of the term-structure.
Example:
http://marketsci.wordpress.com/2008/09/22/test-of-condor%e2%80%99s-vix-based-trading-strategy/
Just my $0.02.
michael
great post, they owe for actually doing real legwork on the merits of the strategy haha
They also say “Volatility is also auto-correlated. If volatility rises today, it has a higher chance of rising tomorrow.”
That’s not true and it also contradicts the statement that volatility is mean reverting.
In fact, VIX returns are negatively auto-correlated. If volatility rises today, it has a higher chance of dropping tomorrow.
On a different note: what’s your opinion on evaluating this strategy (or any other) based on 36 trades resulting from 37 signals? How comfortabel are you with such a (small) sample size? Do you have a certain threshold level that you regard as a necessary minimum?
Hello Greyzy – generally I wouldn’t be comfortable with such a small sample size. The only exception might be something like a long-term trend-following strategy where sample sizes are inherently small.
I don’t have a definite threshold level, but generally for my own trading, I’m looking for observations that happen so frequently as to take sample size off the table (i.e. I’m mostly a short-term active trader).
michael
Michael: another very interesting piece of research, congratulations.
By the way, the strategy you posted in the link below had an interesting timing:
http://marketsci.wordpress.com/2008/09/22/test-of-condor%e2%80%99s-vix-based-trading-strategy/
I duplicated it in WealthLab ( identical results ). However, just a few days after you posted it, it had the by far largest losing trade. From Sept 20, 2008, to October 20, it had a trade with 22.87 % loss. Something like a nine sigma event.
I agree that the Connors system is a martingale ( like many others he is commercializing ). But in general, counter trend trading has always been extremely dangerous. Yes, it has some good years but in the end it seems that we always end up zero.
Congratulations again for your posting.
eber
Hello Michael, I’m sure you’ve seen the paper by Simon and Campasano
http://papers.ssrn.com/sol3/papers.cfm?abstract_id=2094510
I’m curious about your thoughts on this hedging strategy with regard to your volatility MR strategy discussed in earlier posts.
Cheers, Steven
Time ago we made a basic video on VIX
Hope you like it
https://vimeo.com/41915075
It is really important to look at these results objectively. What I really notice is that it does well since the beginning of the new bull market but what system does not do that?? Be skeptical of track records. They are easily “manipulated” Good stuff by the way.