Profiting from Volatility

I have spent the past few days testing the Connors RSI pullback strategy.  It performs very well with certain issues and certain timeframes.  My objective was to find settings that I could use for everyday trading, and I was particularly intrigued by the strategy’s stunning returns during the crash of ’08.

I won’t go into detail about the strategy itself.  It is a mean reversion strategy, and it can produce a win ratio above 75%.  It has a number of parameters that you can tune to be more or less conservative.  The downside is that it may not trade at all for long periods.  This brings us to the “issues and timeframes” part of my study.

I chose a long, net flat timeframe, and ran optimizations for the nine sector ETFs individually and as a portfolio.  There was no set of parameter values that did well for all nine.  The strategy was suited to XLK, for instance, and not to XLU.  You can tune it for any issue but now you’re curve fitting and besides, ten trades in ten years is not statistically valid.

Next, I loaded the 41 issues from Howard Bandy’s book and ran some tests on the portfolio.  Here’s where it gets interesting.  Continuing with some vanilla settings, we find this equity curve around 2008.  Vol1Never mind about my position size.  The point is that equity doubles in roughly three months during the crash.  Remember that this is a strategy that can go dead flat in a bull market, for years.

Mean reversion runs on volatility, and this was obviously a volatile period.  You can see the market volatility in the chart below.  In his later work, Connors adopted historical volatility as part of the strategy. doesn’t have HV so we are looking at standard deviation.


I tried different lengths, and STDEV(40) > 50 seems to be a good indicator of when the strategy makes money.  The peak in November corresponds to the big money months and, if you look carefully, you can see that all the little equity bumps correspond to STDEV bumps.

The problem with STDEV is that it’s not scaled.  Its magnitude changes with the stock price, so you can’t make comparisons, even with the same issue in different time periods.  That’s why HV is better.  It is the standard deviation of proportional (log scale) price changes.

I was intrigued by the little bumps matching up, so I loaded the daily returns into Excel.  They needed smoothing anyway, so I used SMA(40) which is the length I liked for STDEV in the previous chart.  Then, I loaded SPX and calculated HV(40) on it.  Voila!

HV Experiment

Looking at SMA(40) on the daily returns is a lot clearer than squinting at jumps in the equity curve.  Plus, it synchronizes the series with HV(40) so I don’t have to correct for lag.  You can see clearly how the strategy returns are governed by volatility, every little bump.

Bear in mind that the blue line represents the strategy operating, day by day, trading 41 different issues, while the red line is plain simple volatility in the general market.  I was pleased to see such a clear correspondence.

As an added bonus, Excel’s chart scaling feature decided all on its own that HV(40) = 25 is the zero line for strategy returns.  If you are trading this strategy, you might want to stop until volatility picks up.

This entry was posted in Quantitative and tagged , , , , , , , . Bookmark the permalink.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

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

Google+ photo

You are commenting using your Google+ 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 )


Connecting to %s