Test of Condor’s VIX-based Trading Strategy

22Sep08

This is a test of a stock market trading strategy using the VIX offered up by Condor Options. The boys at Condor are all about (as you probably guessed) futures and options so they’ve applied their strategy to S&P 500 futures contracts.  I’m all about leveraged mutual funds, so I’ll spin my test that way.

Condor’s strategy is exploiting the market’s reaction to extreme VIX readings (read: very panicked or very complacent markets). The strategy I’m testing is as follows: go long the S&P 500 using double-beta funds at today’s close if the VIX will close above its upper Bollinger Band (12 days, 1.1 deviations) and close that position if the VIX will close below its 12-day moving average.  Go short the S&P 500 at today’s close if the VIX will close below its lower Band (12 days, 1.8 deviations) and close that position if it will close above its 12-day MA (see geek note at end of post).


[logarithmically scaled]

The graph above shows the S&P 500 (blue) and the Condor strategy long and short (red) and long-only (green) from early 1990 to date.  I’ve assumed a return on cash when not invested in the market equal to the nearest 13-week Treasury bill. 

The Condor strategy produced consistent returns over the entire test period and did a reasonably good job at sidestepping the early 2000 bear market and (at least up until last week) the current bear.  Stats below.

Condor’s strategy as I’ve applied it has produced outsized returns on both an absolute and risk-adjusted basis, but is considerably more volatile than the stock market and has suffered some pretty nasty (albeit not very lengthy) drawdowns. 

One additional observation.  Note how the strategy is asymmetrical in how it treats longs vs shorts; it’s much easier for the strategy to enter a long trade than a short one, but despite this, longs have been more consistently profitable than shorts.  I’m ok with this. With contrarian systems like this one, the market has historically been much more likely to spike up after market lows than spike down after market highs.  This is partially because the market has been mostly bullish and partially an intrinsic characteristic of the market regardless of the broader trend.

Our own long/short YK Strategy suffers from this asymmetry as well – the strategy is much more likely to take an aggressive long position in the face of a down market than a large short position in a strong up market. The takeaway from that is to be very careful in betting against a rising market just because it seems too high.

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.

Geek Note: leveraged mutual funds such as those from Rydex, ProFunds, or Direxion haven’t been around long enough to perform this test using actual fund data, so we’ve simulated the fund returns, conservatively assuming an annual expense ratio of 2.0%.  Because of the very high r-square exhibited by these funds to their underlying indices, these simulated returns are very accurate.



12 Responses to “Test of Condor’s VIX-based Trading Strategy”

  1. 1 dskills

    12 days sounds like it’s been optimized. What did it look like over different period days?

  2. 2 marketsci

    Good comment dskills. The parameters are very optimized (both the 12-day lookback and the 1.1/1.8 deviation cutoffs). To be fair, Condor does make clear that they optimized.

    Adjustments to the three parameters above produce different shades of “not quite as good”, but I think the concept is robust, especially for going long after relatively high VIX readings.

    michael

  3. 3 eber terandst

    I found your comment on simulating leveraged funds extremely intriguing. I tried something like that and found that the correlations drop very fast after only a few days. So, it might work as an acceptable proxy for a very high frequency trading system,say, in and out in a couple of days. Longer than that and the simulated fund will not reflect reality.
    Would you care to comment on the details of your simulated leveraged funds ?
    Thanks
    eb

  4. 4 marketsci

    RE to eber terandst…not sure what method you were using, but for broad indices like the S&P 500, this shouldn’t be any issue. For example, r-squared between S&P 500 index and ProFunds double beta fund ULPIX for daily changes runs about 99.9%, for weekly changes, about 99.8%. For more narrowly focused funds (like individual sectors) I’ve found the r-square to be much lower and much less accurate.

    You’re right that at very long horizons/hold times (say 1 year) the slow tracking drift would be any issue, but for very short-term strategies (like the one in this post), it shouldn’t be an issue.

    Thanks for the comment!

    michael

    P.S. I would do a more detailed demonstration of the simulated funds concept, but then only you and I would read it =) Got to keep it reasonably ungeeky. But I will share the formula I use for broad based funds where the beta is consistent:

    Simulated Price Change at Day 1 = (((Index Close at Day 1 / Index Close at Day 0) – 1) * Beta) – (((1 + Expense Ratio %) ^ (1/252)) – 1)

  5. 5 aleco

    I don’t think it’s fair to compare the SP500 to a(ny) strategy that uses a leveraged ETF. A buy and hold of the leveraged ETF would have similar returns, but without trading fees. By the way, how many trades did the system cause per year?

    In my local database (70 global ETFs, a total of 460 years of data) the returns 20 and 50 days after a spike in volatility [1] are rather high, and the returns are very/rather low when the volatility is very low. In both cases the difference compared to the average is less than 1%, meaning that trading frictions will eat the edge.

    You could slightly optimize the system by also using the RSI14, e.g. go long when vola is very high and RSI very low, go short when vola drops and RSI is high (you’d currently be long in large cap, technology, taiwan and africa btw).

    Still I’m not convinced that such strategies are an interesting option once trading fees are added.

    [1] volatility: standard deviation of open/close/high/low values for the last 20 days, as it has a high correlation to the VIX and will therefor allow you to build a kind-of VIX for any asset.

  6. 6 marketsci

    RE to aleco – I would humbly disagree on multiple points:

    First, regarding friction. The strategy as I’ve applied it is using leveraged funds (Rydex, ProFunds, and Direxion), not leveraged ETFs. As such, there are no additional slippage or transaction costs to consider.

    Second, that you could achieve similar returns with a buy and hold approach. Two issues here:

    First, it’s a misconception that over the long run, double beta funds will produce double the return of an underlying index. Remember, the funds are doubling the daily return of the index, not the long-term return – volatility and the asymmetry of up/down returns will eat away the long-term return of leveraged bull funds.

    Secondly, you’ve ignored risk-adjusted returns. Can you imagine the drawdowns/volatility of simply buying and holding a double-beta bull fund over the period we studied above? The gut wrenching drawdown of the early 00’s? I shudder to think…

  7. 7 John French

    Michael, I have coded this using Yahoo data with TA-Lib and I get nothing like the Equity Curve displayed in the article (anything else I have replicated from your articles has been fine). Would it be possible for you to send me the Excel file (presuming that’s what you used) when you have a chance?

    Thanks

    John

    • 8 marketsci

      RE to John: this is a long ago post – would be a real bear trying to hunt down the xls file now. if you send me your code i’ll try to diagnose. thanks, michael

      • 9 John French

        Michael, I found what my problem was. I was simply picking data up from my VIX sheet on a row by row basis where I should have been using INDEX()+MATCH() (or OFFSET()+MATCH()) as the Yahoo VIX data does not match their SPX data on a date basis (there are some extra rows for some market holidays).

        Thanks for the prompt attention though.

        John

  8. 10 John French

    Michael, thanks for the offer. I will have another look at it after work and if it still looks wrong I will send you what I have. Note I am not using an interest rate if in Cash and am “trading” the SP500 hence there will be differences for sure but not of the manitude I am currently experiencing!

    John

    • 11 marketsci

      RE to John: glad you were able to get it worked out…thanks for the follow up. michael


  1. 1 Technical Analysis at iBankCoin.com » Blog Archive » Weekend Wisdom

Leave a Reply

Please log in using one of these methods to post your comment:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s