Pristine Fire

We had a small wildfire late Monday afternoon into early evening and the follow up today altered my regular schedule so very short post today.

The circumstance of the weather, proximity of structures and fire behavior was such that I couldn't get any neat pictures. We got there, had to pull hose and get water flowing quickly. It was a great outcome, we knocked it down quickly. The red engine on the right, 41 is from a neighboring department. We also had help from Prescott National Forest Engine 330.

A Different Take On Risk Parity

Risk parity has always intrigued me but I've always thought the plain vanilla way to implement it is either hard to do or not that good of a strategy with the results of the Risk Parity ETF offered up as Exhibit 1, although it did well in 2024.

Risk parity is about weighting the holdings such that their respective contributions of risk to the portfolio are the same. A portfolio of an S&P 500 fund and T-bills would have just a small portion in the S&P 500 and the vast majority in T-bills. The generic implementation is to leverage up on debt to balance out the equity exposure similar to what RAPR does but it includes commodities too. The block I was having is that it must be used in simple stock/bond portfolios but that's not the case. Playing around more with Finominal helped me figure that out. FTR, I have no affiliation with Finominal, this is more like getting a new toy to play with. 

Finominal has a portfolio optimizer that lets you optimize for various things including risk parity. In playing around with some of the others, I didn't find anything useful yet but I think I did when optimizing for risk parity. A couple of different ideas to experiment with, here's the first one;

The results of the two are very different. I threw in VBAIX for context;

Portfolio 1 has much better returns with just a little more volatility than VBAIX while Portfolio 2 had 80% of VBAIX' CAGR with less than half the downside and a 1/3 of the volatility. 

Here's another one;

And again compared to VBAIX;

In the second one, Portfolio 1 looks good of course but Portfolio 2 is more interesting, the return equals VBAIX but with less than half the volatility and smaller drawdowns. I believe the reason that the risk parity optimization gives interesting results is that the intuitive versions don't use plain vanilla bonds to manage/offset equity volatility. Putting in 60% SPY/40% AGG into the risk parity optimizer tells us to put 22% in SPY and 78% in AGG which decade to date has compounded three basis points less than inflation. SPY/AGG is not the right input for this exercise. 

The utility of the approach we took today can be a path to cracking the 75/50 code (targets 75% of the upside with only 50% of the downside). 

A cautionary note is that the optimization process ignores the risk of having enormous weightings to one fund like 47% in APHPX or 46% in FLXIX. That much in one liquid alt?

From there, anyone interested in this would need to add more funds and spend some time adjusting the weighting. No matter how great some alternative strategy/fund might be or how glorious the backtest is, it will take a turn being a poor performer and a huge weighting when that happens is an unforced error waiting to happen.

One final item is from today's ETF IQ show which included a segment on the Hedgeye Fourth Turning ETF (HEFT). The idea is to position for big changes in how things are going now and the manager said the strategy has a ten year time horizon. HEFT is a multi-asset fund that includes long/short equity. 

I really don't know anything about the fund but the reason to bring it up at all is differentiation. We've used that word several times lately and HEFT is a good example of what that looks like.

It looks nothing like 60/40, rarely looks like the Permanent Portfolio and occasionally looks similar to other long/short equity. Who knows if that can persist but it has been a good example of the concept thus far. 

The information, analysis and opinions expressed herein reflect our judgment and opinions as of the date of writing and are subject to change at any time without notice. They are not intended to constitute legal, tax, securities or investment advice or a recommended course of action in any given situation.

Here We Are Now! Unconstrain Us!

On Saturday I started to think about and then do a little research on unconstrained bond funds in relation to what we talk about here and what I do in client accounts. Then as a coincidence, Idea Farm included this short paper from PIMCO about unconstrained bond strategies in their weekly email.

The paper doesn't use the word unconstrained but it talks about trying to differentiate a bond portfolio with foreign exposures. PIMCO didn't focus on a specific fund so I am guessing the paper connects to the PIMCO Dynamic Bond Fund (PFIUX) which repeatedly says it is flexible without using the word unconstrained.

I think of that result as being differentiated. It held up much better in 2022 but still went down some. Per the paper, PIMCO thinks foreign exposure is the way to differentiate currently, the way to add alpha. Managing a bond mutual fund, I'm sure they are  trying to isolate outperformance versus something like AGG or a ten year Treasury which makes sense.

The primary objective here is to build a portfolio sleeve that behaves reliably in terms of offsetting equity market declines with a steadier result than plain vanilla bonds have had over the last few years. If prevailing interest rates go down a lot from here, then my approach would lag and if rates go up then my approach will outperform but I am very confident that the strategy of pulling together low volatility strategies each with their own risk factors and with very little interest rate sensitivity will continue to be much steadier than AGG or the ten year.  

I asked Grok for any five star unconstrained bond funds to see whether they do differentiate from AGG. The point is not to pick an unconstrained bond fund but to see if I can learn something new about my approach. It gave me T. Rowe Price Dynamic Credit Fund (RPELX), Carillon Reams Unconstrained Bond Fund (SUBFX) and Artisan Global Unconstrained Fund (APHPX) but the info about their being five star might be stale. Grok threw in MetWest Unconstrained Bond (e.g., MWCIX) with batch of secondary choices. There could be others.

The area  in the black circle is when people needed differentiation the most. A couple of the funds clearly avoided the worst of it in 2022 while a couple only had slight improvement. They have obviously all outperformed for the entire backtest which is nice of course but the volatility numbers are more interesting to me as well as the drawdowns. RPELX and APHPX have meaningfully lower standard deviations than AGG, MWCIX is somewhat lower and SUBFX isn't that much lower. 

APHPX is an obvious outlier, it looks nothing like the others, so what's the story there? I found this on the fact sheet;

That doesn't seem very bond like-unconstrained or otherwise. I asked Copilot if it really is an unconstrained bond fund or some sort of macro strategy. "No, Artisan Global Unconstrained is not really a bond fund. It is a global macro / absolute‑return credit strategy with hedge‑fund‑like tools, exposures, and return profile."

It focuses on foreign exchange, interest rates, various types of credit spread trades, has small equity exposure and can be long/short just about anything. I'm intrigued so a next step is to try to figure out how it compares to a few things. 

It seems pretty alt-ish, with some similar results to other alternative strategies we look at but with different risk drivers. Copilot says they are completely different "risk engines." A couple of the comparison outputs from Copilot;

The fund does take quite a bit of risk but that is not apparent in the numbers, just the strategies it runs. The way I think of the risks that the fund takes is that it has done a good job so far of balancing and mitigating those risks. 

Finominal weighs in on Portfolio 2;

The simplification suggestion is a little more interesting than what we looked at the other day but the suggested portfolio is clearly not superior. It has a similar result with much more volatility. 

Is there enough differentiation with this idea? What this has going for it is it gets a slightly better return with just 50% equity exposure versus 60% for VBAIX with much lower volatility. The drawdowns are generally shallower. 2022 is truncated due to APHPX's start date but I think my conclusion is that there is a decent amount of robustness without massive differentiation. The conclusion is about Portfolio 2, 50% in one alt like in Portfolio 1 is a nonstarter in real life.

The title of this post is a reference to the Nirvana song Smells Like Teen Spirit. 

The information, analysis and opinions expressed herein reflect our judgment and opinions as of the date of writing and are subject to change at any time without notice. They are not intended to constitute legal, tax, securities or investment advice or a recommended course of action in any given situation.

Fun With Finominal!

I spent some time playing around with the portfolio evaluation tools available to the free tier at Finominal. I thought it would be a good way to circle back to some ideas we blogged about in the past and see how they stood up.

First up is a portfolio I saved a year ago (I have a folder of screenshots on my desktop) that I called RR 75/50. The idea with 75/50 is to capture 3/4 of the market going up but only half of the market going down. It's difficult to pull off of course but the math works if you can do it. 

"Imported portfolio" is what I put into their template.

The goal is not to keep up with equities it is to significantly smooth out the ride. If done correctly, it will lag on the way up and outperform on the way down. The weighting to gold is pretty high, more than I have IRL, but was not too much of a problem in gold's recent selloff.

This is the portfolio simplification tab. It misses the mark. Finominal suggests a benchmark and for this it suggested gold, just gold which makes no sense since the portfolio only has 15% in gold. The suggested simplified version has 4x the gold I started with. The result of their suggestion delivers a completely different outcome than what we're going for.

Next is a mashup of the Permanent Portfolio and a loose replication of the Cambria Trinity ETF.

I'm not sure of the exact date but it is close to two years that I wrote about it.

Finominal suggested that replacing it with iShares Momentum (MTUM) for simplification. Again a miss, this idea is far from 100% equities but I think if we look at enough of these there will be some interesting simplification suggestions.

I have a link for the next one, the Yieldy Put which is from a research report by Man from two years ago. We had two versions, the more realistic one was 30% ACWI, 20% TFLO, 25% AQMIX and 25% QLEIX. The other version had 25% in BTAL. The starting point for Yieldy Put was that bonds don't diversify the way they used to but that sized correctly, managed futures and long short can help fill the void left by bonds. 

I cheated. I split the long/short sleeve to put 12.5% each into QLEIX and BTAL. The portfolio is not an equity proxy. The S&P 500 left this one in the dust. It compounded about 300 basis points behind VBAIX for the entire backtest with just over half the volatility. However, since yields bottomed in late 2021, Yieldy Put has done much better.

This shouldn't be a surprise, Man derived the concept in reaction to bonds losing their utility as diversifiers. 

The last one for now is an All-Weather from Man Financial that I built with 50% HEQT, 16% AQMIX, 17% MERIX, 9% SHRIX and 8% RSIR. 

All-Weather is not intended to keep up with equities, it is more of a 75/50 type of idea, maybe even more mild than that. We can't compare it to 60/40 or Dalio via Finominal so we pivot back to testfol.io.

Portfolio's 1 and 2 are not identical but I think they are very close despite being conceived many months apart. 

Taking a broader perspective on most of these that we've built over the last few years, they mostly all take very similar paths toward a portfolio with plenty of opportunity for growth but with a decent amount of robustness not necessarily to make something truly all-weather but help soften the blow to the downside. My perception of truly all-weather is closer to an absolute return without a "normal" allocation to equities, maybe 20% instead of 40-60% in equities. 

Twenty years ago, it was obvious that funds would evolve to democratize access to increasingly sophisticated tools leading to increasingly sophisticated portfolios and that's what has happened. It certainly would be easy to misuse useful tools and some of the funds coming probably aren't that useful so I do think there is a lot of work required to learn how to use them correctly. 

The part about using Finominal, they have a couple of resources not available on the other sites we use. One of the useful things is they have a sort of AI apparently that assigns a benchmark based on the holdings. Comparing what we've built to hedge fund benchmarks and seeing the results supports the validity of what were trying to do/learn. 

The information, analysis and opinions expressed herein reflect our judgment and opinions as of the date of writing and are subject to change at any time without notice. They are not intended to constitute legal, tax, securities or investment advice or a recommended course of action in any given situation.

Why Use Four ETFs When You Could Use 17 Instead?

Invesco maintains a suite of ETF model portfolios. The holdings appear to the same for the most part, subject to asset allocation which ranges from 0% equities/100% bonds to 100% equities/0% bonds. Here is their 60/40.

You can see it is tweaked a little to make room for managed futures which is taken from the fixed income sleeve and there's a little cash. Below, we compare the Invesco 60/40 to the following;

Both are obviously much, much simpler but generally target the same asset allocation. I made one change to the Invesco 60/40, I swapped out ICLO and added client holding JAAA so we could get a longer backtest. 

They all look very similar. There are a couple of instances of differentiation for Portfolio 3, it tends to go down less than the other two probably thanks to avoiding duration. 

The result of the Invesco model is fine but as we look at various models with so many moving parts, the Invesco model has 17 funds plus the cash. There's pretty much no differentiation compared to just buying VBAIX. The point here isn't necessarily to go as simple as buying only VBAIX but if a much simpler portfolio with three, four or five holdings gets essentially the same result with essentially the same volatility and the essentially the same path, I'm not sure why anyone would choose the model with 17 funds versus four. 

If you want differentiation from VBAIX' path (I do) and think you can get it, then taking on a little more complexity is warranted, I believe this to be the case. If you don't want differentiation (perfectly valid) then this might make even less sense.

The prompt for this post was an email from Finominal that did a review of the 0/100 version of their model, so just fixed income which has most of the same fixed income holdings as the 60/40, just proportionately larger weightings. The one difference is they add a small weighting to PCY.

Has the portfolio been thoughtfully constructed?
No, as many of the funds exhibit high correlations, offering practically zero diversification benefits. For example, the iShares Core US Aggregate Bond ETF features a 0.97 correlation with the Invesco Total Return Bond ETF and also 0.97 with the Invesco Equal Weight 0-30 Year Treasury ETF. 

Sort of a scathing review but eight funds to have a 0.97 correlation to AGG? This isn't about my aversion to AGG, plenty of people are just fine with AGG-like exposure but eight funds to replicate it doesn't make sense to me. Finomial says the model can be replaced with two funds; 60% AGG/40% FLOT which is a long time client holding. 

We've done this same exercise quite a few times and differentiation doesn't seem to be a priority very often in this realm.  I asked Copilot about this a few weeks and it very cynically said, model providers aren't trying to differentiate. 

The information, analysis and opinions expressed herein reflect our judgment and opinions as of the date of writing and are subject to change at any time without notice. They are not intended to constitute legal, tax, securities or investment advice or a recommended course of action in any given situation.

Adapting With All-Weather

Meb Faber tweeted out a link to an old Bridgewater paper that is provocatively titled The Biggest Mistake In Investing. Their premise is that the capital asset pricing model (CAPM) is built on return assumptions that don't work out as investors expect. Too frequently, return comes from unexpected sources. 

An example of this that resonates with me is that in 2022, the two most important decisions might have been avoiding duration and including managed futures. In late 2021, you wouldn't have found too many people arguing for those two trades. Maybe that's not what they have in mind but that is good context for me to dissect their point. 

From there, according to Bridgewater,  the answer to solve CAPM's flaws is balancing risk because we can't reliably count on where we will get our return from. Balancing risk then would allow for capturing better returns because we'd have exposure to whatever the unexpected return engine ends up being. This is pretty much what risk parity is and how it works. 

Risk parity is associated with Ray Dalio and Bridgewater, Cliff Asness is also a proponent of risk parity. The Dalio version is also referred to as the All-Weather portfolio which is light on stocks, heavy on duration with a good dose of commodities. The paper we are referencing today is a slight tweak on Dalio's All-Weather if you look that up, the paper is more like 20% in equities, 55% in bonds with duration, 15% in TIPS and 10% in commodities. 

If you do look up the Dalio all-weather, you'll see adaptations that use very plain vanilla funds, especially where bonds are concerned and that it hasn't done well in quite awhile. The idea dates well back into the 40 year bull market for bonds. It worked well as interest rates kept going down. The regime for interest rates ended a few years ago but I believe the concept is still valid even if treasury bond ETFs are not the answer. 

All-Weather with alts on the left and Plain Vanilla All-Weather on the right.

I don't mention SRDAX very often, it is a Stoneridge fund that diversifies several strategies that have no strategic overlap with plain vanilla fixed income. QDSIX is a fund of AQR funds with very low volatility with fairly steady returns. SPMO is in my ownership universe (along with SRDAX). Twenty percent in equities is pretty low, using SPMO dials up the exposure along the lines of what we talked about earlier this week by being more volatile than market cap weighted. SPMO can have the effect of increasing exposure a little bit but is heavy in tech though so if something hideous happens to markets and it starts in the tech sector then SPMO will probably feel it pretty hard. 

Below, we compare them to AQRIX which used to be AQR Risk Parity but is still influenced by risk parity. RPAR is the risk parity ETF and of course plain vanilla 60/40.

The backtest goes back to before the bond market regime changed. All-Weather w/Alts was up a little in 2022 which is a big reason why it is so far ahead. Most other years it is pretty close to 60% SPY/40% AGG in terms of growth but with much less volatility.

I wouldn't count on a repeat of 2022 in terms of so much bond carnage in one year but I still do not thing bonds with duration is the place to be and the SRDAX/QDSIX combo will continue to differentiate from treasuries with duration. If I were to actually implement this, I would use individual TIPS not an ETF, I don't think ETFs capture the effect as well as individual issues. If you put 10% in broad commodities, at times there would be regret not having gold and if you put 10% in gold, at times there would be regret not having broad commodities. Also, IRL splitting 55% between two liquid alts is something I would never do, I'd carve that up into smaller slices into more alts; diversify your diversifiers. 

Back to the paper and this chart that was included.

Again, getting anything close to this sort of outperformance is not realistic with a heavy dose of treasuries with duration. The portfolio we created for this post did outperform 60/40 which may or may not be repeatable but getting a similar return as 60/40 with a lot less volatility (I believe lower volatility is repeatable) is plausible. We're using Dalio's/Bridgewater's process, we could nudge up the equities a little bit if there wasn't enough equity market upcapture and repeating for emphasis, 55% split between two alts is just to keep the blog post simpler.

The information, analysis and opinions expressed herein reflect our judgment and opinions as of the date of writing and are subject to change at any time without notice. They are not intended to constitute legal, tax, securities or investment advice or a recommended course of action in any given situation.

We Want More Vol And Less Leverage

The ReturnStacked guys sent an email about having updated their model portfolios. These are always interesting to look at and fun to play around with. One the models is called ReturnStacked 60/40. They do a lot of volatility targeting with their work and I believe this particular model targets a volatility around 12.

Testfol.io currently has ReturnStacked 60/40 at a vol of 11 versus just under 10 for plain vanilla 60/40. The model is leveraged. The entire point is to test their thesis about using leverage and to support their funds. 

Since all of their models are behind a sign in, it's probably not ok to get too specific with all the moving parts but the notional exposure of ReturnStacked 60/40 is 173% with 62.5% in equities, 50% in fixed income and 60% in alts, 3/4 of the alts exposure is managed futures. This is all pulled together with various multi-asset and levered funds. The model owns RSST so a portion of the domestic equity exposure (S&P 500) comes from this fund as well as a portion of the portfolio's managed futures allocation. In other words they look through to the funds' holding and add up the various exposures. Most of the fixed income exposure is very basic with AGG-like exposure and intermediate treasuries. 

Sort of related to our conversation the other day about portfolio efficiency I wanted to try to replicate ReturnStacked 60/40 not with leverage but by dialing up the volatility. 

The weight to SPMO is pretty close to the S&P 500 exposure when adjusted for volatility. EEM isn't quite as close of a proxy for the foreign exposure but not ridiculously off. MFTNX has twice the volatility as AQMIX and RISR has almost twice as much vol as AGG. ReturnStacked 60/40 does not have overt negative convexity like BTAL but I wanted to throw it in anyway. Systematic macro is missing from my version. HFGM from Unlimited targets 2x volatility for global macro but its track record would shorten our backtest considerably. 

Portfolio 2 above tries to target the same volatility as ReturnStacked 60/40 while Portfolio 3 tries to target about the return by reducing each holding by 50% and then adding 50% in T-bills. Portfolio 3 is an example of leveraging down. We get a similar growth rate with less exposure to risk assets. 

Below, we remove the ReturnStacked 60/40 to allow for a slightly longer backtest that takes in all of the 2022 event.

The drawdown chart hovers over the Tariff Panic of 2025 and you can see the levered down version has a small drop. In 2022 the Replication With Volatility No Leverage version was up 14.41%, the Same CAGR With Less Volatility version was up 7.91 while Plain Vanilla was down almost 17%. It is important to note though that while 2022 looks pretty good, both of our versions lagged Plain Vanilla in 2023 by about 1200 basis points. 

Today's post was a useful exercise in taking someone else's process to create something more useful for, in this case, my approach but in refining your own process, borrowing bits of process from others is a great technique. I draw different conclusions about using leverage than the ReturnStacked guys do but I believe these versions we built mimic what they built and while our versions certainly have drawbacks there is validity to the outputs. 

The information, analysis and opinions expressed herein reflect our judgment and opinions as of the date of writing and are subject to change at any time without notice. They are not intended to constitute legal, tax, securities or investment advice or a recommended course of action in any given situation.