As I stated in past weekly updates concerning the VIX Trading Portfolio, I been running a number of different scenarios on the standardize VIX data in attempt to smooth out the daily variation in the
VIX, which even in the standardized variation is exceeding volatility. I am doing this in order to provide cleaner signals. The data was just too noisy to make any calls useful. Take for instance some of the recent tradings days, where the VIX traded in range of +/- 2 points. These figures, when standardized, were resulting in excessive trading calls. There were even some days where the data would suggest a buy one day, only to go to sell the next. This just was not useful.
To subdue the daily noise, I began to model using a number of different scenarios and methods. I settled in on using a weighted average measure, compiled using data over the course of a number of weeks while putting more emphasis on near-term results. In this way, the data provides less noisy results and smoother transitions period between calls, while still providing timely calls in my opinion.
With that in mind, I came up with results that appears fairly robust and tradeable while keeping the systematic aspects of the strategy simple. I provide some details on the model and show the results below.
The following table shows the average rolling daily six month performance of the S&P 500 as compared to forward six month performance averages following certain criteria on the weighted-average standardized VIX data.
The above tables shows three columns, the first is the average buy-and-hold figures including the average 6-month market return, the median return, the number of instances where the market was up, the number of days in the sample, and the batting average or percentage the market increased over a six-month period. The second column shows the market performance figures following instances when the standardized VIX data greater than 1 but less 2. The third, in contrast, shows the performance figures following instances when the VIX data was less than -1.5 but greater than -2. I excluded those instances when the standardized VIX was +/- 2 due to the mix signals in these regions. The market has a tendency to become highly momentum oriented when the standardized VIX is an absolute 2 and the betting averages tend to move toward even odds. Hence, I excluded these results from the signals.
Using a weighted average also comes with its own problems. One of the most glaring is that sometimes signals are slow to develop. With that in mind, I also ran models for periods when the weighted-average standardized VIX hovered between -1.5 and 1, and used the standardized daily VIX data to discern buy and sell signals. Specifically, sell signals were generated when the daily standardized VIX is greater than 1 and buy signals when results are less -1.45. In both cases the standardized VIX was between the -1.5 and 1.
The results, in my mind, look compelling and I think they can provide cleaner trading signals (probably in conjunction with other data). With all that said, I should mention that the current weighted-average standardized VIX is around -0.09 while the current daily VIX is at about 2.32. This is a sell signal in my mind, noting that downside volume on the SPY has been picking up in recent days.
As for the VIX Trading Portfolio model, I am staying the course for now and keeping the short S&P 500 position. Even still, I am toying with the notion of using 2x and 3x ETF funds to gain greater long or short exposure depending on the signals and other market data.
VIX, which even in the standardized variation is exceeding volatility. I am doing this in order to provide cleaner signals. The data was just too noisy to make any calls useful. Take for instance some of the recent tradings days, where the VIX traded in range of +/- 2 points. These figures, when standardized, were resulting in excessive trading calls. There were even some days where the data would suggest a buy one day, only to go to sell the next. This just was not useful.
To subdue the daily noise, I began to model using a number of different scenarios and methods. I settled in on using a weighted average measure, compiled using data over the course of a number of weeks while putting more emphasis on near-term results. In this way, the data provides less noisy results and smoother transitions period between calls, while still providing timely calls in my opinion.
With that in mind, I came up with results that appears fairly robust and tradeable while keeping the systematic aspects of the strategy simple. I provide some details on the model and show the results below.
The following table shows the average rolling daily six month performance of the S&P 500 as compared to forward six month performance averages following certain criteria on the weighted-average standardized VIX data.
| Standard VIX | |||
| Buy-n-Hold | >1, <2 | >-2, <-1.5 | |
| Avg | 0.81% | -1.20% | 3.75% |
| Median | 2.56% | -3.11% | 3.83% |
| # Up | 1842 | 195 | 355 |
| Total | 3136 | 451 | 489 |
| Batting Avg | 58.7% | 43.2% | 72.6% |
The above tables shows three columns, the first is the average buy-and-hold figures including the average 6-month market return, the median return, the number of instances where the market was up, the number of days in the sample, and the batting average or percentage the market increased over a six-month period. The second column shows the market performance figures following instances when the standardized VIX data greater than 1 but less 2. The third, in contrast, shows the performance figures following instances when the VIX data was less than -1.5 but greater than -2. I excluded those instances when the standardized VIX was +/- 2 due to the mix signals in these regions. The market has a tendency to become highly momentum oriented when the standardized VIX is an absolute 2 and the betting averages tend to move toward even odds. Hence, I excluded these results from the signals.
Using a weighted average also comes with its own problems. One of the most glaring is that sometimes signals are slow to develop. With that in mind, I also ran models for periods when the weighted-average standardized VIX hovered between -1.5 and 1, and used the standardized daily VIX data to discern buy and sell signals. Specifically, sell signals were generated when the daily standardized VIX is greater than 1 and buy signals when results are less -1.45. In both cases the standardized VIX was between the -1.5 and 1.
| >1 | <-1.45 | |
| Avg | -8.59% | 2.26% |
| Median | -9.69% | 4.35% |
| # Up | 19 | 67 |
| Total | 91 | 99 |
| Batting Avg | 20.9% | 67.7% |
The results, in my mind, look compelling and I think they can provide cleaner trading signals (probably in conjunction with other data). With all that said, I should mention that the current weighted-average standardized VIX is around -0.09 while the current daily VIX is at about 2.32. This is a sell signal in my mind, noting that downside volume on the SPY has been picking up in recent days.
As for the VIX Trading Portfolio model, I am staying the course for now and keeping the short S&P 500 position. Even still, I am toying with the notion of using 2x and 3x ETF funds to gain greater long or short exposure depending on the signals and other market data.
No comments:
Post a Comment