The Fortress Portfolio

I cannot recall how long ago I discovered @NomadicSamuel on Twitter. My definition of “Canadian Travel Blogger” might not be 100% accurate but take it as is; he’s a great chap who, for some reason, got the finance bug along his journey. In particular, he got the Picture Perfect Portfolio (the name of his blog) bug, an interest in combining great ETFs to build even greater portfolios. I love it.

He made me discover ETFs that employ innovative strategies and combined those insights with interviews with fund managers, to offer even deeper enlightenment. What Talk Your Book by Animal Spirit podcast should have been but never was.

The only issue I have with his approach is that he is too enthusiastic at times, rushing to create portfolios that seem great but are a bit too data-mined. It is easy to find funds that performed well in the past, it is harder to determine if that performance will last in the future. He also relies too much on PortfolioVisualizer, a tool that can test the past we lived through but not all the possible scenarios that could have happened (yes, yes, PortfolioVisualizer has a Monte Carlo simulation tool but I bet not even the person that coded it has ever used it).

Recently, he created a sort of lazy portfolio called The Fortress Portfolio and today I am going to review it.

The Fortress Portfolio

Even if Samuel benchmarks its fortress to a 40% stocks / 60% bonds portfolio, I prefer to see it as a modern take on the Harry Browne Permanent Portfolio. While Harry had 4 equal, 25% slices of stocks, bonds, gold and cash, Samuel put together the following 40% equal buckets:

• stocks
• bonds
• managed futures
• diversified alternatives

Yes, 40%*4 = 160%. This is viable thanks to our old pals, leverage and capital-efficient funds. Here are the details of the funds:

20% USML – ETRACS 2x Leveraged MSCI US Minimum Volatility Factor ETN
40% RSBT – Return Stacked Bonds & Managed Futures ETF
10% BTAL – AGF U.S. Market Neutral Anti-Beta Fund ETF
10% CAOS – Alpha Architect Tail Risk ETF
10% FLSP – Franklin Systematic Style Premia ETF
10% GLD – SPDR Gold Shares

The additional 60% comes from the embedded leverage included in USML (for $1 invested you get $2 of the MSCI US Low Vol) and RSBT (for $1 invested you get $1 of AGG, a total bond market fund, and $1 of a managed futures strategy).

Before delving into each component of the portfolio, I want to show you what is theoretically achievable. I say theoretically because some of the ETFs employed by the strategy were born not a long time ago, so I have to use proxies to show you a bit of a backtest (I will discuss related implications later):

In the last graph, I have added the S&P500 performance because…I know you 😉

Pretty impressive, uh? And if you are not impressed, just leave now because this is clearly not the content for you.

Now, let’s go and face the harsh reality, a.k.a. why the results I just showed you are not exactly what investors will get (unless you are a millionaire?).

USML

A minimum volatility factor stock portfolio has proven to produce better risk-adjusted returns compared to a market-cap-weighted one. And it does so mainly because, guess what, low volatility. I hope I do not have to dissert further why in a defensive portfolio, it makes sense to employ a tilt towards low-vol, let’s move to more interesting topics 😉

The main issue in levering this portfolio is that there is no future market, as for the S&P500; therefore the cost of leverage would be higher than other capital-efficient solutions. How high? Maybe too much:

The implied cost of financing for NTSX, to pick an example, is very close to the 3 months SOFR while USML pays 1% (and change) more than that! The reason for the difference, I believe, is related to the fact that liquid future markets can maintain low leverage costs, as market participants would arbitrage away significant deviations; in this situation, the ETN issuer has to borrow and pay its own credit risk. It is quite telling that if you buy USMV, the unlevered ETF that tracks the MSCI US Min Vol factor, and borrow from InteractiveBrokers at their highest tier (Fed Funds + 150bps) to lever it up, overall you only pay a 75bps cost of financing. And there is the additional saving in expense ratio, 15bps for USMV compared to 95bps for USML.

At least, USML resets the leverage on a quarterly basis, not daily, therefore reducing the volatility drag effect (for long-term investors). A quick test on PortfolioVisualizer reveals that the difference between the two funds is c130bps in CAGR:

This means the CAGR in Samuel’s backtest should be reduced by 52bps (130 * 40%), from 8.17% to 7.65%.

Managed Futures

Samuel backtest uses PQTIX, the PIMCO TRENDS Managed Futures Strat Instl, while in “real life” the only capital efficient solution we can access is RSBT. It is difficult, if not impossible, to fully capture the real beta of managed futures because it is an investment strategy and each fund, by definition, employs its own.

PQTIX is useful in this test because it has a long history but it is reasonable to also assume that it has a long history because it survived. The trend sleave of RSBT employs a proprietary model developed by its manager: there is no guarantee that, if started in the past, RSBT would have performed as PQTIX and there is even less guarantee that the past of PQTIX would represent a good proxy for the future of RSBT. It is also possible that RSBT would perform better than PQTIX…

If it is of any reassurance, it seems that PQTIX and DBMF, an ETF based on a different but still trend model with the longest available history, do track each other quite well:

Starting from June-23, you can see how the 2 models can even make opposite choices. That happened also in June-22 and eventually the 2 funds converged back again.

That’s the nature of the beast. The ideal solution would be to diversify using multiple managed futures funds…but there is no alternative in a capital-efficient wrapper to RSBT. It is possible to use (again) margin from IB but in this case I am confident the IB 150bps cost is higher than the financing cost implied in RSBT.

AVOLX / CAOS

AVOLX and CAOS are the same product, a mutual fund (AVOLX) that is now an ETF (CAOS). There has been much talking these days around the “holy grail” of investing and CAOS should definitely be up there on the podium (plus, a lot of bonus points for the ticker name). CAOS is a fund that provides tail risk protection, it goes up when stocks dive, BUT also generates income when stocks go up. It is an insurance that pays a premium to the insured! It sounds too good to be true but it has a more-than-10 years track record to back this claim:

I cannot show you the fund performance on PortfolioVisualizer because it is split between the period it was called AVOX and CAOS but it is there and you can check it. Here is the fund NAV (which does NOT include the dividends paid):

You can spot how the fund performed during the COVID crisis: as expected.

Would I trust to allocate 10% of the portfolio to this instrument? The risk is that I would get cash-like returns and then it would not be there during the next crisis…well, that’s not really a risk. The real risk is if the fund would not deliver during a crisis and would start to bleed already when things are calm.

Again, too good to be true?

Maybe a blend with the other tail ETF, TAIL, would be advisable but then bye bye magic, since TAIL bleeds like a character in a Tarantino’s movie. [FYI that’s what I do in my portfolio, a 5% split between CAOS and TAIL]

FLSP

This is an ETF that invests in four style factors (quality, momentum, value and carry), within and across multiple asset classes (equity, fixed income, commodity and currency). To achieve a “pure” exposure to those factors (i.e. to remove the beta component), the fund employs a long/short strategy.

If you take your plain vanilla value stocks ETF, the fund invests in companies that rank high on the value factor; by being long stocks, the fund also has a degree of correlation to the general stock market, the beta. FLSP wants to remove that correlation by taking a simultaneous short position on stocks that rank low on the value factor. The same principle is applied to all the other factors and all the asset classes.

This approach allows the fund to have a low correlation with traditional asset classes and to (ideally) deliver positive returns irrespective of how those markets are performing. The idea behind the fund is not new, AQR has been offering the same strategy in a mutual fund wrapper for decades; this is the first time someone put it inside an ETF.

Here is the ETF total return performance since inception against its benchmark (3-month T-Bill):

For some reason, PortfolioVisualizer just show the last 13 months ¯\_(ツ)_/¯

Here is the performance of AQR’s QSPIX mutual fund in the same period, the one Samuel used for his backtest:

While there’s a certain resemblance between the 2 lines, AQR crushed it with an almost 12% CAGR whereas FLSP’s CAGR was just less than 3%. Part of it is definitely linked to the difference in the models employed and part is by design, because FLSP targets a lower level of volatility (risk). This works well for the FLSP manager because it lowers the risk investors would flee when things go wrong (single line item bias FTW); it is not ideal for investors with opposable thumbs (and no access to AQR funds) because it severely limits the potential of the ETF in a portfolio.

I hope it goes without saying that Samuel’s backtest results are overstated again 🙁

BTAL

This ETF is the AGF U.S. Market Neutral Anti-Beta Fund, another long/short fund that provides exposure to the spread return between low and high-beta stocks. By offering consistent negative beta exposure, it can be used as an effective equity hedge to lower portfolio volatility and reduce the impact of drawdowns; it is a valid alternative to buying Treasuries, volatility products and low-volatility funds when seeking to reduce overall portfolio risk.

This is the single-line item bias poster child. Taken alone, it makes you scream Paaaaaaaaaassssssss:

And yet, when matched with the S&P500…:

Is it worth it?

Like this, probably not. But what if BTAL offers zero-cost protection for scenarios not included in this backtest? In the end, its inclusion does not hurt either.

A Fortress for?

This type of portfolio is ideal for replacing the standard glidepath strategy for retirement. A 7% CAGR with 8% vol and 8% max drawdown can still offer enough kick in the years before retirement and a meaningful SWR in the years immediately after.

In a Goal-Based framework, it looks superior to any strategy with a 10 to 15-year horizon; typically you would devote a big chunk of that portfolio to bonds, exposing yourself to failure if inflation rises.

It is the poor man’s “stay rich portfolio”…the perfect strategy for a person that does not exist? I mean, if you are rich you can probably invest in AQR’s funds or the Cockroach Portfolio but probably there is a group of lads that are rich but not rich enough (especially if they live outside the US).

All these considerations are valid as long as the portfolio delivers as expected. The good news is that the core of the strategy, the combination of stocks/bonds/managed futures, is well demonstrated to work.

The alternatives bucket is there to offer additional diversification and protection, it is highly customizable to fit personal preferences. GLD is the less controversial allocation, considering how well it works in other lazy portfolios like the Golden Butterfly or the Permanent Portfolio itself. It also comes in a capital-efficient package in GDE, if needed.

What I am reading now:

Follow me on Twitter @nprotasoni