Back to Community
What more shortable vehicles can be used for a hedged convexity strategy?

Here's an implementation of the strategy described here:

http://seekingalpha.com/article/2110753-part-v-hedged-convexity-capture-continues-to-be-the-worlds-best-performing-etf-strategy

Looks great on the backtest, but according to Schwab both TZA and TMV are hard to short. How do I go about selecting similar but more easily shorted vehicles for this algorithm?

Clone Algorithm
349
Loading...
Backtest from to with initial capital
Total Returns
--
Alpha
--
Beta
--
Sharpe
--
Sortino
--
Max Drawdown
--
Benchmark Returns
--
Volatility
--
Returns 1 Month 3 Month 6 Month 12 Month
Alpha 1 Month 3 Month 6 Month 12 Month
Beta 1 Month 3 Month 6 Month 12 Month
Sharpe 1 Month 3 Month 6 Month 12 Month
Sortino 1 Month 3 Month 6 Month 12 Month
Volatility 1 Month 3 Month 6 Month 12 Month
Max Drawdown 1 Month 3 Month 6 Month 12 Month
# Backtest ID: 541c919ff163a9089c0e4eb5
There was a runtime error.
26 responses

Alan,
This is really interesting, I just read Harry Long's book, it's really short. He calls his concept "structural arbitrage," it is basically an arbitrage between two common equity hedges, short term VIX volatility futures, and long term bonds.

He reasons that if VIX futures are related to equity markets and long term bonds are related to equity markets, then VIX futures and long term bonds must also be related to each other. VIX ETPs and long term bonds are both used as insurance to hedge equity risk, but the former is usually more expensive than the latter, so the strategy is to sell the expensive insurance, and fully re-insure the risk you have taken on by buying the cheaper insurance in the long bond market.

With this fundamental concept in mind, it looks like you don't need to find any assets to short. For example, XIV is short volatility, and TMF is 3x long on 20+ year bonds, buying both assets will achieve the same goal. Shorting the inverse assets is a way to get more bang for your buck, but the strategy does not depend on it.

Thanks for sharing this, you have likely sent me down quite the rabbit hole.

David

Clone Algorithm
91
Loading...
Backtest from to with initial capital
Total Returns
--
Alpha
--
Beta
--
Sharpe
--
Sortino
--
Max Drawdown
--
Benchmark Returns
--
Volatility
--
Returns 1 Month 3 Month 6 Month 12 Month
Alpha 1 Month 3 Month 6 Month 12 Month
Beta 1 Month 3 Month 6 Month 12 Month
Sharpe 1 Month 3 Month 6 Month 12 Month
Sortino 1 Month 3 Month 6 Month 12 Month
Volatility 1 Month 3 Month 6 Month 12 Month
Max Drawdown 1 Month 3 Month 6 Month 12 Month
# Backtest ID: 541cff998de0cf089c850af1
There was a runtime error.

Hi David (and Alan),

I took a quick look at the algo you posted immediately above, and looking at the Yahoo chart (hopefully in-line below), it is not obvious what's going on here. It is clear that over the last year or so, XIV has shot up dramatically (any idea why?), but TMF basically kinda tracks SPY (S&P 500), with a lot of volatility. So, I'm wondering if there is a bias here in picking XIV? Is there really "arbitrage" in the sense that the returns are risk-free (or at least less risky than an investment in an appropriate benchmark)?

Also, what is XIV? The Yahoo Finance summary is:

The investment seeks to replicate, net of expenses, the inverse of the daily performance of the S&P 500 VIX Short-Term Futures index. The index was designed to provide investors with exposure to one or more maturities of futures contracts on the VIX, which reflects implied volatility of the S&P 500 Index at various points along the volatility forward curve. The calculation of the VIX is based on prices of put and call options on the S&P 500 Index. The ETNs are linked to the daily inverse return of the index and do not represent an investment in the inverse of the VIX.

Whew! Can anyone explain how XIV works? If I buy XIV, what exactly am I buying?

Grant

Grant,
XIV works by selling medium term VIX derivatives and buying short term VIX derivatives, so it is effectively short volatility. If you buy XIV, you are only likely to make money if the volatility in the S&P 500 stays level or goes lower, which would explain why the price has gone up through the last year+. Simon Thornington does a good job of explaining these assets in his first reply here.

I would not consider this idea an arbitrage in a traditional sense, and it is certainly not risk free, but I don't think there are any truly risk free investments out there. It requires looking at the seemingly unrelated bond and volatility markets as having a common feature, they are both used to systematically hedge equity risk. The arbitrage comes in because the aim is to extract returns from mispriced sources of hedging between two markets.

Thanks David,

Well, I guess I have a few things to learn. If I buy XIV, is it backed by anything tangible? If the issuer, Credit Suisse, goes belly up, would XIV retain its value? I guess you are saying it would, since the "VIX derivatives" have intrinsic value in the "volatility market" ("I'll take 3-sigmas with my cheeseburger, please."). All a bit abstract to me at this point.

Grant

One thing to be careful of here is determining what is value being surfaced by the strategy, and what is a result of the fact that VIX volatility has decreased so dramatically since 2008. Any strategy that in some significant way shorts volatility is going to do well on many metrics. But (changing to David much more implementable ETF pairing) this still looks pretty good. Here's a test looking looking at the period around September 2011, which is the biggest increase in volatility since 2008. (The pre-2008 period don't seem to be available for the backtests, which is limiting since we are only testing in bull equity markets). It shows that the strategy still does well for alpha, beta, and sharpe. I'm also graphing the individual ETFs (adjusted to the same start point) so you can see they have an almost perfect inverse correlation which accounts for the low beta despite the high volatility of the components.

Clone Algorithm
349
Loading...
Backtest from to with initial capital
Total Returns
--
Alpha
--
Beta
--
Sharpe
--
Sortino
--
Max Drawdown
--
Benchmark Returns
--
Volatility
--
Returns 1 Month 3 Month 6 Month 12 Month
Alpha 1 Month 3 Month 6 Month 12 Month
Beta 1 Month 3 Month 6 Month 12 Month
Sharpe 1 Month 3 Month 6 Month 12 Month
Sortino 1 Month 3 Month 6 Month 12 Month
Volatility 1 Month 3 Month 6 Month 12 Month
Max Drawdown 1 Month 3 Month 6 Month 12 Month
# Backtest ID: 541df9103ee85108c21c405a
There was a runtime error.

Yes, I too saw that 2011 bump from the TMF holding, but overall the good performance of the backtest is due to XIV. One thing I've noticed in backtesting on Quantopian is that both the 2008-2009 and 2011 time frames appear to be special events. We had the 2007-08 financial crisis and its immediate aftermath (http://en.wikipedia.org/wiki/Financial_crisis_of_2007%E2%80%9308), followed by its lingering impact in Europe (http://en.wikipedia.org/wiki/Eurozone_crisis). I'm wondering if there is a kind of underlying assumption of "linearity" here in that if TMF offset XIV in 2011, it will provide a proportionally smaller hedge for more typical corrections and cycles? Is there any evidence that outside of the 2011 time frame (and going forward), TMF paired with XIV is beneficial?

And what's so special about XIV? I've attached a backtest swapping out XIV for UPRO (ProShares UltraPro S&P500 (UPRO)), which is a 3X leveraged S&P 500 ETF.

It is not obvious what is going on here. It sorta seems like one is basically putting together a traditional stock/bond 50/50 portfolio, but with leverage on both the stocks and the bonds, but effectively there is more leverage on the stocks, so the return looks spectacular.

Generally, perhaps someone can explain how leveraged ETF's work. If I buy a 3X leveraged S&P 500 fund, have I effectively taken out a loan that I will need to pay back? Who is on the hook? I'm guessing that in the end, investors are somehow effectively being loaned money to buy securities, and it is investors who are on the hook. Or is the money being put up by the banks, but they then make it up by charging fees to investors?

Grant

Clone Algorithm
27
Loading...
Backtest from to with initial capital
Total Returns
--
Alpha
--
Beta
--
Sharpe
--
Sortino
--
Max Drawdown
--
Benchmark Returns
--
Volatility
--
Returns 1 Month 3 Month 6 Month 12 Month
Alpha 1 Month 3 Month 6 Month 12 Month
Beta 1 Month 3 Month 6 Month 12 Month
Sharpe 1 Month 3 Month 6 Month 12 Month
Sortino 1 Month 3 Month 6 Month 12 Month
Volatility 1 Month 3 Month 6 Month 12 Month
Max Drawdown 1 Month 3 Month 6 Month 12 Month
# Backtest ID: 541eb56d442fe608aa474a27
There was a runtime error.

To see the value of these parings, just run a test on XIV (or UPRO) alone. The returns are still good (due to the time period) but the Alpha and Sharpe are quite a bit lower, and the Beta is much higher. So the pairing seems to be doing what it purports to do, which is capture the benefits of leverage while minimizing the risks. There's still the problem of really knowing what will happen in down markets, since we have limited data to test on.

As for explaining how the leveraged ETFs and ETNs work, my understanding is that the shorting is synthetic, in that the objective is to buy options that emulate a leveraged short while not actually being short. So nobody's on the "hook" for anything any more than usual.

When I go look at the VXX chart, I can't help but think that this can't put up the numbers it has in the past. VXX has come down from several thousand dollars per share at its inception to the mid 20's range today. How low can it go? XIV is the inverse VXX ETN, so it needs VXX to drop in order to be profitable, unless there is something that I don't understand. Does anybody have any thoughts on this??

VXX is pretty notoriously flawed. Here's a simple explanation of the biggest problem, which amounts to continuously buying high and selling low -- somewhat regardless of the movement of the underlying VIX index that is is supposed to track. VXX: Where Wealth Goes to Die.

So to answer your question, I think in theory VXX can happily keep reverse splitting and go on declining indefinitely, because much of the decline is "bleed" that will never be regained. To reiterate, though, what we don't know very well is what happens in a market of declining prices and increasing VIX volatility. I suppose it's possible that VIX can go out of contango.

Be careful relying on intuition of "how low can things go". VXX can halve every six months, forever. They will just keep reverse-splitting it.

My (uninformed) two cents...

To pick up on David's question, it seems like there should be some base level of intrinsic volatility, that cannot be overcome. There are N market participants and M shares across P securities, with N, M, & P being very large numbers. So, unless the market shuts down entirely, there will be a base level of volatility, right? But maybe that level is so low that it is irrelevant (i.e. "in the noise") compared to fear-uncertainty-doubt (FUD) driven volatility?

Sure, volatility cannot go down forever, but VXX can.

Thanks for shareing.I think you can get the similar result by long TMF and TZA

Long TMF and TZA, and rebalance

Clone Algorithm
39
Loading...
Backtest from to with initial capital
Total Returns
--
Alpha
--
Beta
--
Sharpe
--
Sortino
--
Max Drawdown
--
Benchmark Returns
--
Volatility
--
Returns 1 Month 3 Month 6 Month 12 Month
Alpha 1 Month 3 Month 6 Month 12 Month
Beta 1 Month 3 Month 6 Month 12 Month
Sharpe 1 Month 3 Month 6 Month 12 Month
Sortino 1 Month 3 Month 6 Month 12 Month
Volatility 1 Month 3 Month 6 Month 12 Month
Max Drawdown 1 Month 3 Month 6 Month 12 Month
# Backtest ID: 542dfd75a5b1c2090cd4e3bf
There was a runtime error.

I ran the backtest above, but over a more recent timeframe. It appears that the performance does not persist. --Grant

Clone Algorithm
5
Loading...
Backtest from to with initial capital
Total Returns
--
Alpha
--
Beta
--
Sharpe
--
Sortino
--
Max Drawdown
--
Benchmark Returns
--
Volatility
--
Returns 1 Month 3 Month 6 Month 12 Month
Alpha 1 Month 3 Month 6 Month 12 Month
Beta 1 Month 3 Month 6 Month 12 Month
Sharpe 1 Month 3 Month 6 Month 12 Month
Sortino 1 Month 3 Month 6 Month 12 Month
Volatility 1 Month 3 Month 6 Month 12 Month
Max Drawdown 1 Month 3 Month 6 Month 12 Month
# Backtest ID: 542e6774f025e908fb9e6589
There was a runtime error.

thanks. I find that the strategy have a beta almost equal to 1,so I add a short position of spy, and the new result may be a little more persist

Clone Algorithm
39
Loading...
Backtest from to with initial capital
Total Returns
--
Alpha
--
Beta
--
Sharpe
--
Sortino
--
Max Drawdown
--
Benchmark Returns
--
Volatility
--
Returns 1 Month 3 Month 6 Month 12 Month
Alpha 1 Month 3 Month 6 Month 12 Month
Beta 1 Month 3 Month 6 Month 12 Month
Sharpe 1 Month 3 Month 6 Month 12 Month
Sortino 1 Month 3 Month 6 Month 12 Month
Volatility 1 Month 3 Month 6 Month 12 Month
Max Drawdown 1 Month 3 Month 6 Month 12 Month
# Backtest ID: 542e9a2a50c08c08f9ed96cd
There was a runtime error.

This has a great return (6x BM) but can't tell from the code if there's any leverage - (sorry, new to Quantopian & Python but worked with other systems)
1) Is the portfolio allocation is 0.5 each or -0.5 each (which would be short) for XIV and TMF. Not sure why there's a "-".5 in the code?
context.sids = {
sid(37133):-.5,
sid(38292):-.5
}
2) How can i tell in the code if there's inherent leverage (i.e. cash being added to the portfolio over time)? Some of the other strategies posted on the forum have this issue.
3) Also, is there an easy way to get CAGR from the Metrics, so we can compare different strategies? Alpha=1.16 seems a bit confusing, as i guess it is 11 6% per year above benchmark?

thx
Kiran

Clone Algorithm
32
Loading...
Backtest from to with initial capital
Total Returns
--
Alpha
--
Beta
--
Sharpe
--
Sortino
--
Max Drawdown
--
Benchmark Returns
--
Volatility
--
Returns 1 Month 3 Month 6 Month 12 Month
Alpha 1 Month 3 Month 6 Month 12 Month
Beta 1 Month 3 Month 6 Month 12 Month
Sharpe 1 Month 3 Month 6 Month 12 Month
Sortino 1 Month 3 Month 6 Month 12 Month
Volatility 1 Month 3 Month 6 Month 12 Month
Max Drawdown 1 Month 3 Month 6 Month 12 Month
# Backtest ID: 5446c8dd1bcb20092dfb7185
There was a runtime error.

https://quantstrattrader.wordpress.com/2014/10/08/structural-arbitrage-a-working-long-history-backtest/

This is a good extended backtest of another one of Harry Long's "arbitrage" strategies. Caveat emptor.

You don't need more shortable vehicles, you can go long with IWM and TLT and it still looks ok, use TNA and TMF if you want the 3x versions. I think as long as you have a bull mkt, your are going to look good.

Hey Kiran,
A pretty good way of seeing if a strategy uses leverage is just to plot the position values with something like this for each traded instrument:

record(spy = context.portfolio.positions[context.spy].amount*data[context.spy].price)  

In order to see annual performance of a strategy, the easiest thing you can do is run a full backtest. After it is done, you can select "Returns" and it will show you all the 12-month returns.

I think there is something wrong with this backtest. Long TMF and TZA, 3x funds and a positive return?? Over the last 2 years TMF is DOWN 2% and TZA is DOWN 77%

Nakul, the strategy is taking short positions (50% position each) in both - as suggested by the -0.5 below .. hence its a positive return.
context.sids = {
sid(37133):-.5,
sid(38292):-.5
}

See Oct 2 source code:

This algorithm defines a target long-only diversified portfolio and rebalances
it at a user-specified frequency.

context.sids = {

    sid(38294):0.5,  
    sid(37515):0.5

 }  

Those are different sids

Nakul has a good point - why does the strategy backtest with similar results (pretty strong return) irrespective of whether you open Long or Short - i tried both +0.5 and -0.5. Both TMV and TZA have been in a bearish trend.

Is this a bug or is the algo missing something for accurate backtesting?

kiran

http://seekingalpha.com/symbol/TMV
http://seekingalpha.com/symbol/TZA

Hi Kiran, Nakul, both ETF have a downtrend but if you look at them you can see they fluctuate in opposite direction for certain period of time - Florent