A “penny wise, pound foolish” application of the Trend Model

I received a number of responses to the post on the 2021 report card on my investment models. While most were complimentary, one reader asked me for a more aggressive formulation of the Trend Asset Allocation Model. 

As a reminder, the signals of the Trend Model are out-of-sample signals, but there are no portfolio returns to publish, mainly because I don’t know anything about you. I know nothing about your return targets, your risk tolerance and pain thresholds, your tax situation, or even the jurisdiction you are in. If I offered an actual portfolio, it would be a formal prospectus document outlining what to expect.


Instead, the backtested returns are based on a specific formula for constructing a balanced fund portfolio based on Trend Model scores and reasonable risk assumptions of an average investor with a 60% stock/40% bond asset allocation.

  • Risk-on: 80% SPY (S&P 500), 20% IEF (7-10 Treasuries)
  • Neutral: 60% SPY, 40% IEF
  • Risk-off: 40% SPY, 60% IEF
An advisor or portfolio manager could then change the equity allocation by 20% depending on the Trend Model score without Compliance tapping him on the shoulder.


The historical backtest of the Trend Model using this portfolio construction technique yielded excellent results. An investor using this approach could achieve equity-like returns while bearing balanced fund-like risk. Needless to say, this backtest is just a proof of concept. Every investor is different and your mileage will vary.



A reader then asked me to backtest a more aggressive approach to portfolio construction. Instead of a 60% SPY and 40% IEF benchmark, he suggested a 100% equity position, based on 60% SPY and 40% defensive equity substitute for bonds. The defensive portfolio consists of an equal-weighted portfolio of XLV (Healthcare), XLP (Consumer Staples), XLU (Utilities), and XLRE (Real Estate).


The results turned out to be a case of “penny wise, pound foolish”.



The value of diversification

I re-ran the backtest using the same signals but with different portfolio construction rules. In my results, I dubbed the 60% SPY and 40% defensive equity portfolio the “Hybrid 60/40” and the portfolio constructed using the Trend Model using the defensive equity portfolio component the “All Equity Model”.



The good news is the Trend Asset Allocation Model worked as expected.
  • The All Equity Model had the best returns. It beat the Hybrid 60/40 and the 100% SPY benchmark over the study period.
  • The All Equity Model outperformed the Hybrid 60/40 benchmark, indicating that it was able to distinguish between risk-on and risk-off episodes.
The bad news is there are major caveats to the results.
  • The outperformance exhibited by the Original Model was far better than the All Equity Model. The Original Model beat its benchmark by 4.2% over the study period compared to 1.6% alpha by the All Equity Model. Moreover, the Original Model’s outperformance was far more consistent than the outperformance of the All Equity Model against its benchmark.
  • The All Equity Model exhibited equity-like maximum drawdowns. Even though returns were better, the maximum drawdown for the All Equity Model was similar to the S&P 500 benchmark.
  • The Original Model had better risk-adjusted returns. Even though the All Equity Model had better returns, an investor using the Original Model could achieve a similar return level by employing modest leverage with lower risk, as measured by either standard deviation or maximum drawdown.
The last point illustrates the value of stock-bond diversification. The Original Model used a combination of stocks (S&P 500) and bonds (7-10 year Treasuries) as portfolio building blocks. Stocks and bonds are less correlated to each other than the S&P 500 and a portfolio of defensive stocks, mainly because the latter are stocks and therefore more correlated to the S&P 500. Applying a market timing model to less correlated assets is more valuable than applying the same timing model to correlated assets.


In conclusion, this exercise was a lesson in the value of diversification. Applying a market timing model which works to less correlated assets will add more value than applying the same timing model to correlated assets.


The moral of this story? Pay attention to diversification when constructing a portfolio. Even though you may have a tool such as the Trend Asset Allocation Model that works, using it improperly can lead to a “penny wise, pound foolish” result.


10 thoughts on “A “penny wise, pound foolish” application of the Trend Model

  1. I’ve been using 80% SPY, 10% IEF, 10% DBC. When the Trend Model changes to Neutral, I plan to to 60/20/20. Thoughts on this moving forward?

  2. Since the Trend Model and Ultimate Timing Model results are back tested, how are they useful to a reader/investor? What possible approaches one should consider for their use in real life investing? If not, they are essentially a signaling/advertising tool. This may be too harsh a view but I am looking for guidance in using these models, not investing advice.

    1. Cam said the performance in the scorecard is out of sample. You can checkout the scorecard page. Though I’m unclear if there are any changes to the model since 2013 when it began. Though he did score well since then so I personally wouldn’t be too critical on that.

  3. Cam, you are assuming past correlations will hold. Would you change the model if IEF yield was negative already, like Bunds?

    1. If yields go negative, which I doubt would happen, will there be an effective alternative to the 60/40 model? If so, I will change.

  4. Looks like everyone everywhere is into hedging and asset allocation change, on first trading day of 2022. Anyone can smell it from hundreds of miles away. It just feels super strange. Have never seen this for a long time. And yet indices are at ATH levels.

    What is it? Even the long disappeared and discredited Dennis Gartman is out today with a prediction of 15% slow and laborious decline because of aggressive rate hikes (four hikes, in his opinions). No comment on his opinions. But I actually feel better after seeing the headlines.

    The one special attribute of modern markets is that there is no slow decline. It happens at lightening speed whenever major changes are in the making. Algos will drive the price movements in the speed of electrons, aka time compression.

    1. yeah strange how everyone saying they want to lighten up in January. Maybe we can be contrarian bullish or since everyone’s trimming at around same time in addition to weak confidence the lightning fast 10%+ drawdown will happen this month. Just brainstorming.

    1. As for me – we left for vacation on December 30. I moved to cash ahead of departure (TIP/IEF closed). That turned out to be prescient. Less prescient was driving to SoCal early that morning. I-5 was closed due to snow, which necessitated backtracking to Bakersfield and detouring through the high desert. Our first time splashing through flooded sections of the highway in the midst of the Mojave desert! One of the worst storms to hit SoCal – sections of Topanga Canyon were closed due to mudslides.

Comments are closed.