Black Tar Portfolio

Note: Please read the disclaimer. The author is not providing professional investing advice or recommendations.

I must continue to warn that back-testing and data-mining stock market simulations can be dangerously habit-forming! I speak from experience!

The 36 year database of historical asset class returns pieced together by has proven to be my black tar heroin of late. I just keep thinking of one more simulation to run… Just this one additional back-test, and then I’ll quit…

Truth be told, 36 years of annual returns does make for a small sample size. And of course the future may look nothing like the past. But the results of one of my simulations was so interesting I thought I’d do a quick post to share it.

Rule Number 1, Don’t Lose Money
You’ve heard the Warren Buffet quote on the rules of investing ad nauseaum. Rule #1 is not to lose money and Rule #2 is not to forget Rule #1. So this begs the question, based on historical returns, what portfolio asset allocation (with annual re-balancing) would have resulted in never having a return below 0% for any year?

Rule Number 2, Maximize Return
But… duh! We can already predict what that answer will be – something like putting all your dough in a money market fund. So let’s add a second constraint, we don’t just want to find any portfolio that never lost money. Let’s find the one that did so with the highest annualized return.

Well, after literally millions of iterations of backtesting random portfolios (with a little genetic algorithm optimization to speed up convergence), here was the result:

The Black Tar Portfolio

63% US Short Term Treasuries (such as VFISX)
17% Emerging Markets (such as VEIEX or EEM)
12% Commodities (such as PCRDX)
08% International Value (such as VTRIX or EFV)

This particular portfolio had the following characteristics over 1972-2007:

Best Return (1985) 25.23%
Worst Return (1994) 0.01%
Average (Arithmetic) Return 11.64%
Annualized (Geometric) Return 11.43%
Sample Standard Deviation 6.84%
Skew -0.066
Excess Kurtosis -0.678

Note, if you type the above weights into’s backtesting engine, you’ll get slightly different results due to rounding.

Now most times when I perform such a backtest, I end up with some optimized portfolio that, to be honest, I don’t really know how to interpret. So the optimal asset classes turned out to be X Y Z using allocation weights of A B C… So what?!

But in this case the more I studied the results and asked “What does this mean?”, the more fascinated I became. I really like this portfolio, and at the risk of falling prey to some sort of Talebian narrative fallacy, I think I see an almost perfect logic to it. Let me pontificate…

The Life Vest: Roughly 2/3 of the principal goes into short term US treasuries. Not surprising as this asset class is time and again referred to as a virtually risk-free investment.
Spatial Diversity: All bonds come from my own country, all stocks from somewhere else. At the very least it’s a sort of currency hedge.
Risk Premium #1: Of the 25% devoted to stocks, a dominate portion (2/3) goes to the historically high risk (but high reward) emerging markets.
Risk Premium #2: The other 1/3 in stocks is spread around the most beaten down (value premium) of the various developed European countries.
Tangible and Intangible: I’m not an expert on the necessity of commodities, but PIMCO claims that a nice quality is that they represent “real” assets (stocks and bonds are “financial” assets). They also supposedly do well in high inflation (thus no need for TIPS).
Good Stats: The white belt will notice the good mean return. The yellow belt will recognize an accompanying low standard deviation. The green belt notes that skewness is near 0 so there shouldn’t be more returns below the mean than above. And the brown belt notes a negative excess kurtosis, so large surprises far from the portfolio mean should be rare. The black belt notices all of the above, but uses more esoteric language – impressively stating that the 3rd and 4th moments about the mean indicate returns of a symmetrical and platykurtic nature.

The Odd Couple
Now one way to get a stable portfolio with moderate returns is to invest in a bunch of medium-risk vehicles. Yawn! This portfolio achieves moderation in a different way, dumping 80% of the money into a marriage of the boring-as-watching-paint-dry treasuries with the Ay, caliente! emerging markets. I think this will make for more interesting bubble and black swan years as at least one of your asset classes will be trouncing the market returns during these times.

Other observations? Well given that this portfolio has both never lost money and has actually averaged a nice return, you could argue that it’s perfect for anyone, independent of age and risk tolerance. With 63% already in bonds, forget all that gradual shifting of asset class allocations as you approach retirement. And hang up worrying about whether you’ve taken on more risk than you can really tolerate when the market goes sour. If it continues to perform as it has in the past, you never lose money while your pot doubles every 6.4 years.

What about 2008?
Perhaps the final question is, given this year’s large drop so far, how has this portfolio performed year-to-date? As of November 22, the Black Tar Portfolio is down 16.8%. By comparison the S&P 500 (proxied by VFINX) is down 44.4%.

Yes, it’s not fair to just compare single year returns while ignoring that one portfolio might be riskier than the other. So here’s Black Tar vs. the S&P since the beginning of recorded history (OK…’s recorded history…)


Portfolio Annualized
Black Tar
S&P 500

Yep – slightly more return with much less risk.

So will the Black Tar Portfolio go the way of all data-mining anomalies or need to be revised? It’s certainly looking like it won’t yield a positive return this year, but the performance so far is incredibly respectable given the huge loses other so-called diversified lazy portfolios are delivering.

I’ll check back in with the Black Tar Portfolio after the end of the year just to see how it needs to be revised moving forward.

UPDATE January 2, 2009
By my calculations, Black Tar returned -13.4% for 2008. Not bad compared to the S&P’s -37%, but we certainly can’t say that it didn’t lose money. I’ve incorporated the asset class returns for 2008 and am currently simulating to find the updated Black Tar for 2009. Planning to do a new post when I have the new portfolio so stay tuned…

8 thoughts on “Black Tar Portfolio”

  1. Your writing is fun to read! And thanks for some actionable info in your posts. Liked them & linked them as well. I used some of your info in a post of mine. Hope that is ok with you.

  2. Lumi, thanks for posting. Have you tried to backtest your Black Tar Portfolio with a data set that goes back farther?

  3. Hey Phaser,

    No I haven’t. You run into problems if you go too far back b/c some of the indexes haven’t been around for more than a couple decades.

    I’m guessing if you could go back 100+ years the constraint of never losing money may force you into almost 100% treasuries. In that case it may make more sense to alter the criteria a little – maybe requiring the portfolio to not lose money 95% or 98% of the time.

    – Lumi

  4. Hi Momo, has a table of correlation coefficients here.

    You can also read here about how they pieced together index return history going back to 1972.

    You’re right that some of the funds and indexes haven’t been around since 1972 but they seem to have pieced together the best (free) proxy for what returns would likely have been.

    It just seems to me that you’d probably get a more realistic picture by using their database due to the larger sample size than if you only have data going back a few years.

    By the way, I’d probably pick PCRDX or PCRCX over GSG for my commodity allocation. I recently discovered an article here that shows how different the commodity indexes can be.

    PCRDX is sort of an equal split between metals, energy and agriculture whereas GSG has almost 2/3 allocated to solely energy.

    – Lumi

  5. hey momo,

    that’s a good point. if you want to test short asset classes too i guess you could create your own by multiplying the historical long asset class returns by -1 (assuming you’ve imported assetplay’s data into your own simulation environment).

    GRZZX had a great year but I’m estimating a “since inception” (2000) annualized return of about 3%, which is about the same performance as a money market fund over that same time period.

    given the market’s long-term upward trend i’m not sure you’d want to include short asset classes in any sort of long-term buy-and-hold strategy. but they’d certainly make sense if you had a market-timing technique to tell you when to move in and out of them.

    – lumi

Leave a Reply

Your email address will not be published.