In my blog post last week I dove into complex multi-model trend applications on the S&P 500. In that post, I modeled out trend systems that are either long or short—betting either with the uptrend or the downtrend. You can read that post by clicking the link:
In that post I largely found that going short—or betting with the downside—is a futile effort for many equity index investors. This is even the case when the higher costs of short-selling are excluded. Today, of course, we have lower cost options such as E-mini futures contracts and inverse index ETFs.
Successfully betting with the downside on equity markets is incredibly difficult. I believe a large part of this difficulty stems from the increased volatility the equity markets experience during downtrending periods. Prices can move sharply one way or another, quickly shifting trend directions and eroding the possibility of compounding gains that can be crystalized for the investor.
This is not to say that gains cannot be made betting with the downtrend. When rising markets slowly rollover and then gain momentum on the downside, long/short trading systems can see enormous gains in a very short time. However, these systems fail to hold onto most of the gains as reversals are often sharp and unpredictable. Timing the transition from being short the market to going long the market is an epic challenge.
In contrast, uptrending periods in equities tend to be much less volatile. A buyer can often hold the market for a long time and see beautiful compounding gains during these holding periods. Strategic re-balancing strategies allow investors to hold onto the gains.
Following Uptrends in Equities
The dichotomy of trend characteristics within equity indices provides investors with an opportunity to make the best of the trends they can more easily profit from—the uptrending side of the market.
Since uptrending markets are generally less volatile, we can buy the market as it enters an uptrend based on the trend systems’ definition or identification of an uptrend. We can often stay in the market for a long time, holding the market until the frequently-observed slow rollover in the trend occurs. Then we exit. Of course, this isn’t always the case. October 1987 is the classic reminder that uptrends don’t always have gentle transitions.
In the following backtests we identify short-term trends as seen from 2.5 weeks to 3 months of price movement. Long-term trends are identified from 3 months to 12 months of price movement. As a long/flat trend model, the factor scale goes from 0 to +10 in both trend models. When the factor is at 0 the investor will be fully in Treasury bills; when the factor is at +10 the investor will be fully invested in the S&P 500.
Long/Flat Short-term Trend Model
Short-term trends, as one would expect, involves a lot of monitoring of the trends for signal changes. There are relatively long periods of holding the same position without any changes; however, they are often limited to a few months in duration.
In this backtest I used my short-term trend model on the S&P 500 price going back to 1950. It does not include any distributions.
The second chart in this section shows the drawdowns of my short-term trend model. The drawdowns are very muted into the 1980s. Entering the 1990s the drawdowns are noticeably larger. Some of this effect may coincide with the shift in the effectiveness of short-term trends after 1987, which I previously observed in my post on long/short models.
Despite the increase in drawdowns beginning in the 1990s, the overall drawdowns of this system are still much lower than the underlying index. For example, the S&P 500 index saw a price decline in excess of 55 percent in the 2007-2009 period.
Long/Flat Long-term Trend Model
The long-term trend model in this section measures trends exceeding 3 months in duration. As one would expect, the holding periods without any change are much longer—many exceed one year in duration.
As in my short-term trend model, this backtest is applied to the S&P 500 index price. Distributions are not included.
Well into the 1960s my long-term trend model was performing very comparably to the S&P 500 (though with smaller drawdowns). The trend model began pulling away from the underlying index beginning with the 1969 market correction. The performance during the 1973-1974 downtrend period further solidified the gains of my long-term trend model.
While my long-term trend model performs very favorably compared to the underling index during larger downtrending periods, this trend model still does experience relatively large drawdowns. The following chart shows the drawdowns from 1951 to 2018.
The deepest drawdown period with my long-term trend model occurred in the October 1987 crash. My model caught nearly the full brunt of the sharp decline. In the other major market drawdowns—2000-2003 and 2007-2009—the long-term trend model held drawdowns at 20 percent. This is a very impressive performance compared with the much larger drawdowns of the underlying index.
Short-term vs. Long-term Trend Models
Choosing between the short-term trend model or the long-term trend model is not easy. During the nearly seven decade backtest period, both models have generated nearly the same overall return and have clearly outperformed the S&P 500 index. My long-term trend model is easier to manage and does a great job of reducing drawdowns while achieving index-beating returns. With long holding periods it is probably a better option for most self-directed investors. It would be more tax-friendly as well in regular trading accounts.
However, my short-term model also is impressive. It does a better job getting out of bad situations quickly. We saw this clearly in 1987. The short-term model had completely exited the market by October 9, two weeks before the October 19 crash. Aside from trading much more frequently, one of the drawbacks of my short-term system is the larger overall drawdowns in the big market downtrends. This is because the model re-enters the market during reversals, losing some money before exiting back to Treasury bills.
The following chart shows the total performance of both my short-term and long-term trend models applied on the S&P 500.
My short-term trend model handily outperforms my long-term trend model at the beginning of the backtest period. This performance gap begins to shrink after 1987. In recent years my long-term model has significantly outperformed my short-term model (though since 2000 returns have been more subdued for both).
A chart tracking the annualized 3-year rolling returns of each trend model shows this performance difference much more clearly. The darkened area represents the difference between the rolling returns of my long-term trend model compared to my short-term trend model. As explained, it shows the outsized returns of my short-term model at the beginning of the backtest period. The long-term model begins to outperform in the latter half of the backtest.
Examining the models, my long-term trend model and short-term trend model both show great returns that have beat the underlying S&P 500 index. They both demonstrate excess returns over full market cycles and have greatly reduced drawdowns. This translates to superior reward/risk ratios.
In recent decades, the advantage in overall performance seems to have shifted somewhat in favor of my long-term trend model. This was not always true. Until the 1980s my short-term trend model on the S&P 500 was very effective. The logarithmic chart for my short-term trend model shows a very neat staircase effect of large returns followed by mellow periods.
Using a long/flat trend model appears to be much better than a long/short trend model when applied against broad equity indices. The S&P 500 is not the only broad equity index, but it serves as a good marker of equities in general. I have done backtests on a range of other broad market indices which show similar results.
In the coming weeks I will complete a post on my long/flat trend model applied to other large indices in U.S. and foreign markets.
Comments & Questions
All comments are moderated before being posted for public viewing. Please don't send in multiple comments if yours doesn't appear right away. It can take up to 24 hours before comments are posted.
Comments containing links or "trolling" will not be posted. Comments with profane language or those which reveal personal information will be edited by moderator.