A six-month run of a retail focused strategy that seeks to capitalize on the tendency for implied volatility [IV] of S&P 500 component stocks to mean revert is definitely floundering (see study review information at the bottom of this article). Slicing and dicing the data in a variety of ways displays one pretty important factor for most systems: the need for a significant number of profitable trades when all costs are incorporated. Although trend-following systems can benefit when large profits are taken on fewer profitable trades, this strategy has not created those higher average gains over losses.
Fortunately, the results are far from disappointing. Here are a few things to consider:
We continue method review here in an on-going search to capitalize on IV mean reversion in a systematic way. The discussion here looks at intermediate steps necessary before abandoning an approach. They can be applied to other strategy reviews or as a launching point for you to continue with this one.
Current IV versus Median IV
Optionetics Platinum can be used to generate a list of S&P 500 component stocks that provides current IV for different timeframes (i.e. 7-149 days), along with median values for the same period. This data can then be transferred to an Excel spreadsheet where the difference between the current IV and median IV is ranked in percentage terms so there is an apples-to-apples comparison of the data. The strategy used examined long straddle results for 10 stocks with current IVs that were furthest below their one-year median IVs. The assumption is that the component stocks IVs will increase sufficiently to generate profits for the combination position.
The two long positions in a straddle make it particularly vulnerable to losses due to time decay. In addition to off-setting this impact by using next month instead of near month options, longer term options were reviewed, including LEAPs (Long Term Equity Anticipation Securities). Tabular results of the approach were reviewed last week. This week, a closer look at the underlying data using January options is completed using scatter plots.
Since the trade approach is based upon the relative relationship of current IV and median IV, the first scatter plots display position Profit/Loss (P/L) versus the Current/Median IV. The specific formula used for the IV percentages is: [(current IV Median IV)/Median IV]. Two trade approaches were evaluated:
By increasing the holding period, additional time was provided to allow for the current IV to move upward toward the median IV. Although this allows for additional time decay, the longer term to expiration helps partially off-set the impact. On all of the charts provided, P/L is displayed on the horizontal axis (x-axis).
Figure 1: Scatter Plot of Profits/Losses versus Current IV Relative to Median IV (IV%), Held 1 Month
(click here for larger view)
The first thing that should strike you is the amount of data points to the left of the y-axis, the breakeven point for the trades (with commissions and slippage). Next, note that the two data points with the lowest IV% represented losses. Finally, the profitable positions dont appear to be in an IV% range that may represent optimal conditions for the strategy. That is, if we limited trades to those between -10% and -30% IV%, a majority of the trades still result in losses.
Since the strategy was tested with relatively high commissions (only 1 contract traded with full commissions), trades to the right of the first interval (-50) would realize profits if scaled up. Even though this approximately doubles the profitable trades, the majority represent larger, unprofitable losses.
Figure 2: Scatter Plot of P/L versus IV% with a Regression Line, Held 2 Months
(click here for larger view)
Although results appear to shift moderately when held two months, part of that impact results from a change in the y-scale. The profitable trades have become more profitable while the losses diminished more slightly. The liner regression line displays the impact of IV% on profitability. The flatness of the line is the strongest image for the lack of potential in the system. The slight slope in this line is likely due to the outlier values on the lower portion of the plot.
A trader can certainly look at other data cuts to note if there is anything distinctive about the points that lie in the profitable region. Using 90-day statistical volatility [SV] and the Probability of Profits function in Platinum, a view of how historical movement in the stock may be predictive of trade results was reviewed.
Figure 3: Scatter Plot of P/L versus Probability of Profits with a Regression Line, Held 2 Months
(click here for larger view)
Had the results been more compelling, the Probability of Profits measure would be discussed in more depth; however, it is not merited in this case.
Two last scatter plots are provided as part of a look back process. We anticipate that changes in price and changes in IV will impact the results. Since increases or decreases in price will benefit a straddle, the absolute values of the price changes were used so that a linear regression line was still valid to display the results.
Figure 4: Scatter Plot of P/L versus Absolute Change in Price with a Regression Line, Held 2 Months
(click here for larger view)
Figure 5: Scatter Plot of P/L versus Change in IV% with a Regression Line, Held 1 Month
(click here for larger view)
Data for trades held 1 month was used since the long holding period created deeper in the money and out of the money IVs that skewed results (IV=0 in each case). So it appears price changes and IV% changes did impact profitability. No startling conclusion, but ideally the use of various scatter plots provided here will provide some insight on how to examine backtest data you generate.
Three factors that remain for me to explore include:
So much for this being the end of the discussionsince the outstanding issues impact other trade strategies, it remains a valid pursuit. The next installment will appear in June.
To see the other articles by this author, please click here.
Clare White, CMT
Contributing Writer and Options Strategist
Optionetics.com ~ Your Options Education Site
See Analytical Toolbox articles: Volatility Mean Reversion (Dec 31, 2006) and Strategy Follow-Up Volatility Mean Reversion (Mar 9, 2007) for additional information about the strategy methodology.
See Analytical Toolbox articles: Probability of Profits in Practice (Sep 2006), and Two Keys to the Probability Calculations (Sep 2006) for more information about the Probability of Profit function in Platinum.
Study Review
In Options Returns and the Cross-Sectional Predictability of Implied Volatility, by Amit Goyal of Emory Universitys Goizueta Business School and Alessio Saretto of Purdue Universitys Krannert School, the authors examine volatility trends for SPX component stocks along with strategy assessments to capitalize on these trends. Using the index components rather than the index itself is a somewhat unique approach and was determined to be a better means for predicting future volatilities.
One of the main findings of the study was that a stock with an implied volatility [IV] below its 12-month moving average value tended to have a higher IV the next month. On the other side of the volatility picture, stocks with IVs higher than its 12 month moving average tended to have a lower IV the next month. These results were based upon 10-year of data (120 monthly data sets) and used decile groupings to identify the component stocks with the highest relative IV and those with the lowest relative IV.
The study identified a group of 50 relatively high IV stocks along with a group of 50 relatively low IV stocks each month on the trading day prior to expiration. On the Tuesday following this scan, long straddles were entered for the relatively low IV stocks since mean reversion implies a boost in premium from rising IV levels. To capitalize on decreasing IV levels using the relatively high IV stocks, short straddles were established for this group. While rational, this represents 200 individual option trades and significant margin for 100 short option positions.
Although spread mid-points were used for transactions, the study also separately examined increased transaction costs to more accurately model actual costs. Even when thinly traded options (with greater spreads) were evaluated using higher transaction cost methods, the results remained impressive. An average monthly return of 15% declined to 5.3%. A significant decline, but the approach still justifies further investigation.
Modified Approach
The modified approach makes three modifications. First, it reduces the number of trades to allow for smaller allocations. Rather than breaking the 500 component stocks into 10 x 50 stock groupings, 50 x 10 stock groupings are used, with long straddles created for the 10 relatively low IV stocks. It should be noted that this group excludes deal stocks; those with IVs that have recently decreased significantly due to pending corporate actions such as private equity takeovers or spin-offs.
The second significant modification makes use of a comparison of current IV versus median IV over one year rather than mean IV for the year. The advantage of using a median value is that it minimizes the impact of outlier values on the results. A very brief period of unusual, high volatility will skew mean results for the entire period, while having much smaller impact on median results.
The final modification was in expiration month selection. Rather than using near month options, the approach tested options for
This was done to minimize the impact of accelerated time decay within 30 days and within 45 days, respectively. The study portfolio included an equal number of short options which benefit from time decay.