A High-Return, Low-Risk Portfolio
A blend of strategies that select different types of stocks should be able to provide high average annual returns with less volatility than a single strategy. We examine one such blend submitted by a reader that showed an average annual return of 53% from 1986 through 1999 in a backtest, with a much higher risk-adjusted return than the S&P 500.
By Ann Coleman (TMF AnnC)
November 9, 2000
Today's article will be simple. I hope.
I'm getting a little worried. I have a feeling that the last two articles on Workshop strategies may be confusing those of you who aren't normally Workshop investors.
Before I was asked to start supervising the Workshop (in June of 1999), I used to feel a bit guilty for not really knowing what was going on in there. Every few months I would wander in and try to figure out what those guys were doing. After 10 or 20 minutes I would exit hastily, shaking my head, totally confused but vowing I would try again when I had more time.
I have tried, really tried, to make the Workshop more accessible to newcomers. Elan Caspi's FAQ has helped a lot too, as has his untiring devotion to answering "newbie" questions on the Workshop discussion board. Also helpful was creating the Mechanical Investing discussion board, where the Brain Trust can continue its collaboration without having to cope with those newbie questions -- or risk scaring the newbies out of their ergonomic seats by discussing three sigma events and linear regression models.
So, now that we are combining the Foolish Four and the Workshop, I worry that the rush of information will prove to daunting to many Foolish Four investors who, after all, were promised a 15-minute-a-year strategy. If you are feeling daunted, believe me, we all understand. We, those of us who hung in there until the mechanical investing idea clicked, all went through a period of finding this process terribly confusing. But, we are glad we stuck it out and we hope we can make it less confusing for the next batch.
On Monday in the Foolish Four report, we talked a bit about how to combine Workshop strategies into a blended portfolio that would give us diversity both in the numbers of stocks held and the types of stocks held. Jamie Gritton's backtester lets us backtest various combinations of strategies to see just how the returns and volatility have interacted over the past 14 years. We don't expect that the past returns are going to say much of anything about how well our real-money portfolio will do in next year or even over the next 15 years, but past performance can help us make an educated guess about which strategies might work well together to lower volatility while giving us a shot at pretty good returns.
At the end of Monday's Foolish Four portfolio report, I asked readers to play around with the backtester and familiarize themselves with how it works. I jokingly said that if anyone found a well-diversified blend that had average annual returns (CAGR) over 50% with low volatility (GSD less than 20%), they should send that blend to me right away. Lo and behold, I got one. Before I share it with you, though, let's review what those numbers mean.
The least-important number is the actual CAGR, which in this case is 52%. Yep, had one known about and followed this screen from 1986 through 1999, her portfolio would have grown by an average of 52% per year. Calm down. The important number is the GSD, which is a remarkably low, 18%. What that tells us is actually pretty impressive.
The Geometric Standard Deviation number says the returns in the backtest followed a distribution pattern that predicts that 67% of the time, the annual return will be within one standard deviation of the mean (which is 53%). In this case, that means two-thirds of the time the return should be 52%, plus or minus 18%, or within a range of 34% to 70%. Well, that sounds pretty good. GSD also tells us that 95% of the time, the average return will fall between two standard deviations of 52%, or within a range of 28% to 113%.
Here's the full table:
Blend(%) S&P500(%)
CAGR+3s* 152 65
CAGR+2s 113 47
CAGR+1s 80 32
CAGR 52 18
CAGR-1s 28 6
CAGR-2s 8 -5
CAGR-3s -9 -15
*s = sigma, the symbol for standard deviation
There is a problem with this. Naturally. For one thing, the sample period is almost all "bull market." Also, investing returns in general tend to have "fat tails." That expression refers to the trailing ends of a bell curve. The probability of getting a 3-sigma return, positive or negative, should be about 1.5%. But, in real life, we see extreme returns more frequently than these models predict. The bell curve doesn't taper off as much as it should.
For example, the numbers above predict that the S&P would lose more than 15% only 1% of the time. But, if you look at the returns from the last 74 years, we see 4 years of losses significantly greater than 15%. That's 5% of the time. There's considerable danger in expecting the returns of the past 14 years to project into the future, when they don't even adequately project into the past.
That's not to say that the GSD number is useless. Not at all. It does illustrate the range of returns fairly well, it just can't predict the extreme events that occur in investing very well, and it can't do much about changing market conditions.
So, with that rock of salt, let's see what Mike S. sent in:
This is a blend I am considering for 50% of my portfolio beginning Jan 2001.
30% Key100, Annual, stocks 1-3
30% Peg26, Annual, stocks 1-3
20% PlowRSW, Annual, stocks 1-2
20% RS13, Quarterly, stocks 1-2
(NOTE: You can learn more about these strategies on our Screen Explanations page. The PlowRSW is a variation on the Plowback screen that we follow in our Current Rankings and Workshop Returns. PlowRSW rankings are available at Brian Finney's website.)
Want to see how it works? Go to the backtester. Use the Blender and enter each strategy. Or, go to the bottom of the main backtester page and enter this code in the URL box to set up the strategy automatically: BLv86990101kc30n1213ps30n1213bw20n1212rq20n0312.
I gotta tell you, this is a pretty impressive blend. The yearly return stability is very attractive, and it's fairly diverse. Trading costs would be low, too, since three of the screens are annuals.
But, not as diverse as I would like. It selects only the top two or three stocks from each screen. That means it will have a maximum of 10 stocks and, some of the time, due to overlaps, it will hold fewer. Cherry-picking the top stocks is what gives it such a high return, and selecting diverse screens is what gives it such stability, but six to 10 stocks is probably not enough to give a portfolio real diversity.
Generally, between 15 and 20 stocks is enough to get good diversification; and that's what I'm aiming for, even if it means a somewhat lower expected return. But, Mike's portfolio is an excellent example of the type of blend we will probably end up with (although we will be adding the Foolish Four to the mix, remember).
Now it's your turn. What do you think of this mix and how much diversification do you think we need? Let's hear from you on the Foolish Workshop discussion board.
Fool on and prosper!
