Condor Options

Because implied volatility is forward-looking, it tends to lead backward-looking historical (or realized) volatility readings.  This relationship stands to reason: options prices can and do respond in real time to new information, while any historical measurement – even a short-term one – will by definition lag as day-to-day changes are gradually incorporated into the period under review.

The chart above (click to enlarge) displays the 30 day historical volatility (HV; the dotted line) and the nearer-months average implied volatility (IV; the solid red line) in the S&P 500 over the past 4 years.  As you can see, IV tends to run ahead of HV readings, though this is entirely consistent with the fact that the two calculations cover different timeframes.

In our prior work on mean reversion in implied volatility, we noticed that violent IV swings often tended to be quite early when used to predict price movement.  Instead of using the movement of IV itself to help moderate that tendency to be early, we examined whether HV would be helpful as a confirmation of the predicted move:

Is there any significant correlation between price movement in the underlying and crossovers of historical and implied volatility levels?

There are actually two questions here that we wanted to test:

1. Does HV crossing above IV provide a reliable short signal?
2. Does IV subsequently crossing back above HV provide a reliable long signal?

Parameters

We used the 15 day rather than 30 day historical volatility, as this perked the indicator up a little bit and generated some more trades; however, the results are fairly robust with.  We used the daily average weighted implied volatility of the near-month options.

Testing this as a buy-and-hold, signal-to-signal strategy would introduce too much noise, so we imposed some stops and time limits: each trade has a maximum 15-day limit after which it is closed, and we used stops of 50 S&P points on the upside, and 100 points on the downside.  Optimizing these parameters further should improve performance, but we aren’t so much interested in building a trading strategy here as we are in confirming a hypothesis.

Results

The table below presents the performance of a strategy applied using $100,000 in the S&P 500 ETF (SPY), using the parameters stated above, since 2005.*

One surprising result is that the short signals significantly outperformed the long ones, with a profit factor of 5.54 vs. 2.31 for the long side. This suggests that waiting for realized price action to confirm and exhibit the behavior expected in the options market (i.e., waiting for HV to cross above IV), may provide a non-negligible edge versus reacting to swings in IV alone. While our data set is too limited to draw any firm conclusions, we wonder whether the “bullish VIX spike” thesis that has become so popular is, over the very short term, precisely backwards: once price action in the underlying begins to conform to expectations, the edge seems to be to the downside.

We also noticed that signals tended to clump together, starting with a short trade as HV crossed higher, followed by alternating long and short positions, usually ending in a buy signal that expired at our 15 day limit.

Some of the losing trades came from “false positive” crossover signals generated during otherwise quiet bull markets.  Taking trades on the basis of relatively tiny fluctuations in volatility proved unprofitable, so we retested the strategy by adding a IV “floor,” such that signals generated when implied volatility was below, say, 15% would be ignored.  (Obviously, using any particular absolute IV number is capricious, but 15% is pretty safe as a bright line dividing active from quiet markets.)  The IV floor improved the results somewhat.

Unpleasant Surprises

With regard to the situation in September-October 2008 (the right edge of the chart above), notice that the spike in HV was not preceded by a similar rise in IV.  A short crossover signal did occur in mid-September and was extremely profitable, but this case is clearly different from the other signals, as the HV cannot be said to be “digesting” or reflecting volatility that had been implied earlier.  If anything, options prices were significantly underestimating future volatility, and had to play catch-up to realized price action.

It might be instructive to look at instances of this other type – that is, cases where HV spikes without any real warning in the options market.  We’ll leave that as an area for possible further research.

Conclusion

Our study of the relationship between crossovers in historical and implied volatility readings suggests that waiting for realized price action to reflect the movement implied in the options markets presents highly favorable short and long side opportunities in the short term.

* We ran these tests on SPY rather than $SPX only because the volatility data available from Tradestation for $SPX was incomplete at the time of testing.  SPY data prior to 2005 was plagued by the same problems.  Prior work has shown that while there are substantial differences between SPX and SPY when applying optimized strategies, they are similar enough for our purposes here.