A Quantitative Strategy For Trading Inverse Volatility With Impressive Backtested Results

Includes: SVXY, UVXY, VXX, XIV
by: David Easter


Historical backtesting calculates impressive hypothetical gains.

Objective buy and sell triggers eliminate most of the guesswork.

Individual judgment must still be exercised whenever a volatility spike is known to be imminent.

This article is an expansion and extension of a comment that I initially posted here on Seeking Alpha in response to an article by Nathan Buehler. My comment summarized the results of quantitatively backtesting Buehler's backwardation/contango trigger strategy for shorting volatility, but I added my own objective exit strategy.

Principles of the Backtesting Model

I applied the following rules in backtesting analysis. These rules are objective and quantitative; therefore the reported results can be verified independently by anyone who is up to the challenge of writing the appropriate Excel code.

1) I based my evaluation on the (theoretical) purchase of VelocityShares Daily Inverse VIX ST ETN (NASDAQ:XIV) long, an inverse VIX futures ETN, at the market closing price. The returns of ProShares Short VIX Short-Term Futures (NYSEARCA:SVXY) are indistinguishable from those of XIV, but XIV was created first, so its real prices go back earlier in time.

2) Real XIV closing prices, adjusted for splits, were used back to the ETN inception date of Nov. 30, 2010, available here from Yahoo Finance. For theoretical XIV prices prior to inception, I used the calculated/modeled prices of XIV back to March 26, 2004, posted here at the Intelligent Investor Blog. When downloaded, the data set was current through Sept. 26, 2014. The same spreadsheet was the source of my front-month and second-month VIX futures closing prices.

3) I define the backwardation of VIX futures as being when the second month futures price is less than the front month futures price at close of trading (4:15 p.m. ET); I define contango in VIX futures as being the reverse situation at close of futures trading. Only closing prices are considered in this analysis.

4) I define a "buy signal" for XIV as being when the VIX futures first move into backwardation (based on closing price) one day, and subsequently move back into contango one or more days later, following the model developed by Nathan Buehler. A "buy" signal is generated the first day that the futures prices reverted to contango as defined above, and XIV is (theoretically) purchased at its closing price the same day.

5) No exceptions were made in the analysis for subjective economic or geopolitical considerations. By nature, such exceptions can neither be objectively nor quantitatively backtested.

6) I tested to determine the optimal limit sell price to maximize overall gains. I also tested to determine the optimal stop sell price for losing trades. Finally, in order to avoid unnecessary losses during major spikes in volatility (as in 2008 and 2011), I tested to determine an optimal trailing stop strategy. The results are all summarized below.

7) XIV was theoretically bought/sold at the market close price the same day that the buy/sell signal was triggered.

8) The model assumes that 100% of the account balance is committed to each trade. All gains or losses are parlayed into the next trade. Results are reported based on an initial investment of $25K.

9) Theoretical gains calculated by the model are pre-tax, and do not account for brokerage fees or margin interest charges. Real gains would have been less after these expenses were paid.

Results of the Modeling

1. The Buy Cycle. After the previous buy cycle has been completed, a new buy cycle begins when the VIX futures enter backwardation at close one day and then revert to contango one or more days later. The buy signal is immediately triggered, resulting in the purchase of XIV (inverse volatility) at the market closing price the same day. A buy cycle ends when either, based on the initial trade price, (1A) the closing price of XIV drops below its stop loss price (see 2A below); or (1B) the price of XIV exceeds its limit sell price (see 2C below). Although the position may be sold based on a trailing stop as described in (2B) below, this event does not end the buy cycle. A new buy signal cannot be generated until the previous buy cycle has been closed.

2. The Exit Strategy. The exit (selling) strategy has three considerations, presented in order of priority. (2A) The optimal stop loss limit was determined to be -37% relative to the original purchase price. Absorbing this large of a loss will be uncomfortable to many investors. (2B) Once the investment gains surpass a defined threshold, a trailing stop is implemented. Modeling identifies the threshold limit as +38% gain relative to the original purchase price, and the trailing limit as a loss of -32% relative to the highest closing price prior to any downturn. (2C) The position is closed when the share price exceeds the established limit sell price on market close. The optimal limit price was identified as +132.8% relative to the initial purchase price.


buy date

sell date

cycle end


balance / $M

max loss

days held

















































































Source: Created by author.

Results of backtesting are collected in the table. The end date differs from the sell date for cycles 3 and 7, when the trailing stop was triggered. The maximum running loss that would have been experienced in each cycle is also shown. The graph shows the running gain or loss (%) for each cycle vs. the number of market days the position is held. The cycle number is indicated at the end of each cycle. It should be clear from the graph that downturns are typical within each cycle.

Over the 10.5 year period since March 26, 2004, this backtested strategy would have resulted in only 10 round-trip trades. All 10 of the trades would have been profitable, with a median holding period of 247 (calendar) days and a median profit of +134%. Without the trailing stop strategy, two trades (cycles 3 and 7) would have been stopped out at a -37% loss. There were ten periods when no position would have been held (including the present), with a median duration of 61 days. Overall, the backtested strategy would have theoretically increased an initial balance of $25K to $41.7M, for an annualized percentage gain of 102% over 10.5 years. For comparison, buying and holding XIV for the entire 10.5-year period would have increased the same $25K balance to $496K, for an annualized percentage gain just under 33% over 10.5 years.

Future Adjustment of Parameters

Although this model would have resulted in impressive returns over the past 10 years, the four parameters were optimized with the specific goal of maximizing backtested performance. The parameter values will have to be reevaluated and tweaked as additional data are added with the passing of time.

The upper sell limit was set with the goal of capturing maximum gains. Setting the limit higher than 132.8% would have missed the completion of cycle 4. Setting the limit too low, however, would systematically diminish gains from all eight cycles that were sold at the upper limit price. Future market realities are likely to warrant setting a limit price lower than 132.8%.

The stop-loss limit was set to avoid being stopped out of a profitable cycle, while avoiding unnecessary losses. Importantly, it also triggers the signal for the end of a buy cycle in the event that a trailing stop has been activated (cycles 3 and 7). For example, if the stop loss were tighter than -31.5%, cycle 6 would have stopped out before reversing and making a big run.

The trailing stop threshold (+38%) was set to activate during cycles 3 and 7 before those trades turned downward. The trailing stop of -32% was set wide enough to avoid stopping out any of the other (profitable) cycles, while simultaneously exiting cycles 3 and 7 with maximum overall profit. The downturns during cycles 3 and 7 correspond to major spikes in market volatility. Whenever a major volatility spike is reasonably certain, investors would be wise to exercise their own discretion by exiting a long XIV position early.


This analysis confirms that the qualitative trigger signal for shorting ProShares Ultra VIX Short-Term Futures (NYSEARCA:UVXY) presented by Nathan Buehler has merit and can be applied quantitatively. Including objective limit sell, stop loss, and trailing stop parameters produce impressive backtested returns. This strategy is not suitable for day- or short-term trading. The holding period for the 10 round-trip trades over the 10.5 years of backtesting ranged from a minimum of 131 days to a maximum of 490 days. The strategy averages only one round-trip trade per year.

One cannot expect to make a profit on every trade. Although the 10 round-trip trades in this analysis all resulted in backtested gains, the same results cannot be expected in live trading. Short-term results can differ significantly from long term expectations. This strategy will not be as effective shorting UVXY or iPath S&P 500 VIX ST Futures ETN (NYSEARCA:VXX) as it will be for buying XIV or SVXY long. This is primarily because the maximum possible theoretical gain from shorting an ETF or an ETN is 100% per trade, whereas, for each cycle that ends by selling at the high limit, this exit strategy results in a backtested gain of 133% or more per trade. This strategy has not been evaluated for options trading.

As of this writing on Oct. 9, 2014, the last trade by this strategy would have been a sell on July 1, 2014, with a gain of 137.3% after holding the position for 16 months. The strategy would have had no open position for the last three months. However, VIX futures closed Oct. 9 in backwardation. Therefore, the strategy will signal a buy signal the next trading day the VIX futures revert, and close in contango.

Note: The long SVXY position that I currently hold was opened prior to the beginning of my analysis and development of the trading model presented in this article. I wish to thank Nathan Buehler and ianxpotent for their comments and for their encouragement.

Disclaimer: Past performance, including theoretically modeled performance based on backtesting, is no guarantee of future performance. The strategy involves a high level of investment risk. The author is not responsible for losses incurred by individuals who base their trades on any of these ideas.

Disclosure: The author is long SVXY.

The author wrote this article themselves, and it expresses their own opinions. The author is not receiving compensation for it. The author has no business relationship with any company whose stock is mentioned in this article.

About this article:

Want to share your opinion on this article? Add a comment.
Disagree with this article? .
To report a factual error in this article, click here