I have mentioned this a few time in earlier posts, but I thought I would begin writing more frequent comments and observations about my long-term technical indicator. This will include a somewhat detailed review of the indicator, insights and potential recommendations, and the performance results from a model portfolio constructed using the indicator as a timing tool.
For a brief overview, my long-term technical indicator results from multiple attempts over the years of constructing a timing indicator using standard statistics. By themselves, standard statistical models, in of themselves, are abysmal timing tools. Any performance figures for timing derived using these models are prone to wild swings, are inconsistent, and require so many exceptions that it makes them unusable. That said, I found that with a few corrections and adjustment to the standard statistical models, their usability are vastly improved.
The above graph shows the price trend of the S&P 500 since the late 1980's, with the long-term technical indicator over-imposed. Generally speaking, a low reading on the indicator is a bullish-signal. In that spirit, I created a simple model on the S&P 500. The model states that you have double exposure to the S&P 500 in any period where in the last two years the indicator was below a certain critical level. For any other period, you passively except the market return. The results of this are shown below. The dark blue line represents the model while the pink is the market.
One note of caution that I will continue to reiterate in nearly every subsequent post I write concerning this indicator, it does not work in all situations, for all stocks, and all the time. No perfect indicator exists. That said, the indicator does appear to provide clean signals in a majority of the cases (actually 60% to 70% of time in the subset of the universe I tested the data against), and I thought I would catalog my progress with it.
In other back-testing analyses, I found a portfolio consisting of names purchased when the timing model was in the critical range outperformed the market over a 10 to 20 year period. With that in mind, I set up a model portfolio in Marketocracy to track the model's real-time performance. The model portfolio is constructed using a combination of the timing model and other fundamental/technical analyses, the later to weed down a list typically consisting of a few hundred names. The holding period for any stock is expected to be about two-years and the portfolio is expected to be low turnover. You can track the performance of the portfolio here.
The graph above shows the results so far. The dark blue line is the portfolio while the green and brown lines represent the S&P 500 and NASDAQ, respectively. Since early July, when the portfolio was initiated, the portfolio is underperforming the market by about 300bps, not unexpected considering many of the names in the portfolio are beaten down laggards. I would expect short-term performance to lag considering most stocks follow their recent trend. That said, the portfolio has a Beta of only 0.82 and an R-squared of 0.39%, according to Marketocracy estimates.
For a brief overview, my long-term technical indicator results from multiple attempts over the years of constructing a timing indicator using standard statistics. By themselves, standard statistical models, in of themselves, are abysmal timing tools. Any performance figures for timing derived using these models are prone to wild swings, are inconsistent, and require so many exceptions that it makes them unusable. That said, I found that with a few corrections and adjustment to the standard statistical models, their usability are vastly improved.
The above graph shows the price trend of the S&P 500 since the late 1980's, with the long-term technical indicator over-imposed. Generally speaking, a low reading on the indicator is a bullish-signal. In that spirit, I created a simple model on the S&P 500. The model states that you have double exposure to the S&P 500 in any period where in the last two years the indicator was below a certain critical level. For any other period, you passively except the market return. The results of this are shown below. The dark blue line represents the model while the pink is the market.
One note of caution that I will continue to reiterate in nearly every subsequent post I write concerning this indicator, it does not work in all situations, for all stocks, and all the time. No perfect indicator exists. That said, the indicator does appear to provide clean signals in a majority of the cases (actually 60% to 70% of time in the subset of the universe I tested the data against), and I thought I would catalog my progress with it.
In other back-testing analyses, I found a portfolio consisting of names purchased when the timing model was in the critical range outperformed the market over a 10 to 20 year period. With that in mind, I set up a model portfolio in Marketocracy to track the model's real-time performance. The model portfolio is constructed using a combination of the timing model and other fundamental/technical analyses, the later to weed down a list typically consisting of a few hundred names. The holding period for any stock is expected to be about two-years and the portfolio is expected to be low turnover. You can track the performance of the portfolio here.
The graph above shows the results so far. The dark blue line is the portfolio while the green and brown lines represent the S&P 500 and NASDAQ, respectively. Since early July, when the portfolio was initiated, the portfolio is underperforming the market by about 300bps, not unexpected considering many of the names in the portfolio are beaten down laggards. I would expect short-term performance to lag considering most stocks follow their recent trend. That said, the portfolio has a Beta of only 0.82 and an R-squared of 0.39%, according to Marketocracy estimates.
I will check it out.... thanks
ReplyDelete