While the model portfolio's "full cycle" performance since the market top in October 2007 is impressive (+63% vs -17% for the S&P 500), its performance since the historic market bottom (Spring 2009) has significantly lagged the market, which has been extremely frustrating. In this note, I'll pass along some findings from my recent post performance analysis, and explain some changes I'll be making to the model portolio to hopefully improve its overall performance going forward.
First, I'm reassured from my analysis that the model is again showing that it has the ability to consistently select stocks that will outperform the market. Consider the chart, which shows the "stock only" performance (no hedge) in my ACTUAL trading account for the 12 months ending April 22, 2011. As you can see, the stocks selected for the model portfolio have gained 42.5% while the S&P 500 has gained 11% over the same 12 month period, nearly a 4:1 advantage. These "excess returns" are consistent with the model portfolio's performance since inception in 2003.
The exception was the 12 month period immediately following bottom made in Spring 2009, where the stocks selected by the model significantly underperformed the benchmark index. The model was simply out-of-phase with the market during this period. Why? ...because the model selects growth stocks that are expected to outperform the market in a "normal" rally where money chases growth, whereas the 2009-2010 rally period was led by a bounce-back in large financial stocks that had been severely punished as the financial system unraveled. The model simply isn't designed to select down-trodden stocks... rather it selects high performing growth All-Stars, which happened to underperform the broader market during the 12 month period ending April 2010. This was a highly unusual period following a historic panic fueled decline, and now that money is again chasing growth, our model stocks are out performing the benchmark index.
Going forward you'll see some changes. First, I'm removing the hedge (TWM) from the model portfolio. Even without the hedge the model is uncorrelated with the stock market. e.g. the model has shown to have less than a 30% correlation with the market since inception, which is about as good as it gets. The portfolio's allocations will range from "100% long" to "50% short" depending on market conditions.
To drive the portfolio's long/short allocation, I'll continue to use the same third party timing model I've been using for years... the DecisionPoint Thrust / Trend Model (TTM). Google it if you'd like a detailed description. I've analyzed dozens of market timing algorithms and this is by far the BEST I've ever seen. Importantly, it enabled us to neatly side step the severe market decline in 2008-09 and also correctly got us on the right side of the market in early 2009 as the market began to rise from the ashes.
The TTM gives us three types of intermediate term signals which will drive the allocations:
BUY - 100% long
NEUTRAL - 50% long, 50% cash
SELL - 50% short
Going forward, I'll continue to post the hypothetical "Model Portfolio" stock selections and allocations, and I'll begin posting the ACTUAL weekly performance in my trading account rather than hypothetical model portfolio returns. This will make it easier to track your performance against what is actually achievable rather than what is theoretically achievable. You should be seeing less tracking error as a result.
Enough for now.
This chart shows the hypothetical performance of the new and improved Strategic Growth Model portfolio, since inception in May 2003 through April 22, 2011, including its performance during the different phases of the intermediate cycle (BUY-NEUTRAL-SELL).
While the model portfolio's "full cycle" performance since the market top in October 2007 is impressive (+63% vs -17% for the S&P 500), its performance since the historic market bottom (Spring 2009) has significantly lagged the market, which has been extremely frustrating. In this note, I'll pass along some findings from my recent post performance analysis, and explain some changes I'll be making to the model portolio to hopefully improve its overall performance going forward.
Posts
