I am an avid NBA fan. Aside from the game itself, I love the financial and human dynamics on how to build a successful team. For the uninitiated, the financial aspect is interesting because each team has a ‘salary cap’, you cannot solve your championship problems finding an Arab royal family with an infinite bankroll like in European football.
There is a well-grounded theory that you need a superstar player to win a championship. If you scroll the list of all the teams that won in the past, you can easily see empirical evidence on why this theory works in practice. The notable exception to the rule is the 2003-04 seasons when the superstar-less Detroit Pistons won against a team that had four (out of five starters) superstars, the Fab Four LA Lakers. You want a superstar, maybe more than one, but you do not want too many.
NBA star players come in all shapes and forms, from the sharpshooter Steph Curry to the my-hands-are-so-big-I-cannot-shoot Shaq. So each General Manager in the league that is lucky enough to have a star in his roster has to ‘build around’ that player, find other players that fit that player characteristics AND the general league playbook trend. Throughout the years, some unique players got into the spotlight because they were great at filling a specific need to make their team a championship team.
Take Golden State Draymond Green. He is phenomenal for the role he plays for his team and his salary reflects that value. But contrary to other players that have more general skills, Draymond is untradeable for GS because no other team will extract the same value out of him, especially at that salary level. Part of Draymond’s value resides in him being always opinionated; you can find him leading defense and offense on and off the court, directing players in the right spot and mentoring young teammates. The dark side of DG’s superpower is that there are as well people on the court that are not interested in his opinion or stand his temperament. During the 2016 Finals, he got suspended for a game after reaching four flagrant foul points: GS was leading the series 3-1 and his suspension open the gates for the Cavaliers’ historical come-back to win the Finals 4-3. Three years later, he had an argument with teammate Kevin Durant that most likely drove KD’s decision to leave the team at the end of the year. Would any other NBA team win a championship with Draymond on their roster? Probably not. Would GS have won three rings (and counting…) without Draymond? Probably neither.
Draymond value is context-based.
The best example is The Worm himself, Dennis Rodman. In a game where everyone has to attack and defend, score, pass, dribble, he had one talent but he was the world’s best at it: rebounds. Someone might see it as a pretty niche talent and yet, he brought three rings in Chicago and two in Detroit. While in Detroit everyone was fine with his persona in and out of the field, in Chicago things were different. Coach Phil Jackson and Michael Jordan had…a different work ethics; but Phil, in particular, was great at understanding people, he knew that to get the best out of Dennis on the court he had to leave him free outside. Dennis’s previous team, the San Antonio Spurs, ditched him thinking that his lifestyle was too much of a drag to have him on board. Instead of trying to change him where it did not matter (his private life), Phil embraced Dennis for what mattered: to stop opponents, reduce their second chances and give MJ more opportunities to have the ball in his hands.
Rodman was so unconventional that a lot of people hated him. They hated him because he did not look right. The hair, the piercings, the attitude. Does not matter if you are a great fit for your team if you did what you were supposed to in order for your team to win.
“How can he help the team to win if he does play only on one side of the court?”
“Yes yesterday he played fine but how long he can go on with that lifestyle?”
I can still hear the pundits. It is almost like, instead of The Worm, we should have called him…The Bond Market?
Yes, I am not 100% proud of this transition. Last week I read this wonderful post from Portfolio Charts (PC from now on) which exposes in a great way the peril of judging asset classes by themselves and not by the role they play in your asset allocation. How many times this year did you hear dismissive comments about bonds?
Yields are too low!
Negative real yields!
Bonds do not go to the mooooooon!
How many of these comments were context-based, i.e. what bonds can do in a specific asset allocation, the way bonds are employed by 90% of investors? I will go deeper into PC model and post later on but I would like to start with a couple of images that present well this point. PC Portfolio Finder calculates the risk and return of hundreds of different portfolios simultaneously and plots them all on the same chart. That lonely point in the bottom-right of the chart represents a portfolio of 100% gold, while all the other plots are any possible portfolios that include a portion of gold.
The 100% gold portfolio is a shit portfolio. Comments you read around about this commodity are related to that even if no one employs it as such. Below graph highlights instead some great portfolios the includes gold:
Not as bad uh? The Golden Butterfly is a simple portfolio equally divided into Total Stock Market, Small Cap Value, Long Term Bonds, Short Term Bonds, and Gold. I found those images very powerful in demonstrating the concept. A single asset performance does not matter, the whole portfolio does.
Months ago I read that an asset manager created a portfolio combining some beta elements (say stocks) with a liquid alternative strategy (say a long-short sleeve). The reason was only cosmetic: they realized that trying to sell the alpha part of the portfolio alone, the long-short strategy, was way harder than the combination because the buyer, say a pension fund, would concentrate only on THAT strategy performance and not what that strategy could do next to the other assets that pension fund already had.
This cognitive (?) error is common between retail and institutional investors. I manage a fund of funds, let’s call it like this for simplicity, and every quarter I get questions about the fund that performed worst. The overall portfolio performance glanced over and rarely anyone is interested in understanding if the portfolio dynamic is acting as originally intended. Can’t see the forest for the trees or, even better, less is more: in this situation, reporting only the portfolio result without details on the single components would be beneficial for the target audience (this is also an impossible solution because the target audience will never admit to themselves that they are acting this way).
I grew up watching this cartoon:
The same storyline has been used for countless movies: the Avengers, the Justice League, Fast and Furious, Ocean 11 with men, Ocean 11 with women. How can it be so complicated to understand that teaming up will lead to achieving amazing goals?
Ah yes, Brexit. Sorry.
Now you are a portfolio believer, great. In the next section, I will do a deeper analysis of PC post, highlighting the parts I like the most and the ones I like less; unfortunately, you have to make some assumptions when building a portfolio and the quality of those assumptions will determine your real life, out of sample, returns.
Baseline Return vs Average Return
PC thinks there are three main issues in utilizing nominal average returns to rank asset classes. The first adjustment they propose is to use real returns, taking into account how inflation erodes purchase power. I see no issue with that. The other two can be summarised as ‘true compound growth that investors actually experience’ and ‘returns uncertainty’. To give you a simple example, in the ’00 decade the SPY ETF returned 0.35% annualized; in the ’10 decade, it returned 13.75%. There is a huge difference if you invested in the first 10 years period compared to the second one. This difference can be especially meaningful if you are living out of your portfolio returns because no one will compensate if you are so unlucky to live in the wrong decade. The average of the two periods is 7% and it is the value you normally see associated with ‘stock returns’. Experiencing those decades in that order or the reverse will strongly affect your compounded returns if you are investing a fixed amount every month.
In order to prepare the reader for what is called “sequencing luck”, PC effectively downgrades those assets that evidenced the highest volatility (here is the complete formula and explanation).
My opinion is that each investor lives two phases: the accumulation (when they save) and the distribution (retirement), or if you prefer the ‘get rich’ and ‘stay rich’ stages. In the accumulation phase, when the time horizon is some decades, each investor’s return tends to get closer and closer to those averages. Using leverage can diminish sequencing luck (see my previous posts about leverage) and rebalancing a 1/12th of your portfolio every month instead of 100% once per year smooth as well returns (listen to this podcast if you want to know more). That return volatility is one of the reasons some assets perform better than others in the long term. The risk here is to be too conservative and a lot of investors do not have that luxury. We have to learn how to manage risk, not to avoid it.
PC model works best in the distribution phase because investors are more interested in avoiding bad surprises than being exposed to good ones.
Ulcer Index
Standard deviation as a proxy for risk is a bad measure, I agree. PC proposes to utilize the Ulcer Index (here a description of what it is and how it is calculated):
The basic idea is to find a single number that can serve as a reference point for historical portfolio pain that 1) is far more informative than standard deviation; and 2) accounts for both the depth and length of a drawdown.
While the Ulcer Index can be used for single asset classes, it shines for portfolios: it is a single metric that considers the risky stuff you care about, like max drawdowns, correlations and downside volatility. As PC correctly says, you should not focus so much on the absolute value of the number, but instead use it as a reference point for comparing the relative pain of two portfolios.
If I have to find a defect, for the Ulcer Index it doesn’t really matter if the drawdown is deep and short or shallow and long, as both can potentially generate the same number. It is hard to say if this really represents investor preferences; 2020 was deep and short and the majority of investors came out pretty fine. On the other side, deep (short or long, does not matter) is way more dangerous if the investor employs leverage. All in, I think the Ulcer Index is a great upgrade compared to risk metrics used in standard finance parlance.
Past Returns
Some days ago I found this Twit that explains (part of) my thinking really well:
Past returns are useful…up to a certain point. Take our beloved Gold. Before the advent of GLD, retail capital had few, inefficient options to invest in gold; futures require a large capital base and some experience in managing margins and rollovers, coins or gold bars have high ‘trading’ costs and cannot be stored as easily as a stock with your online broker. When new capital gets access to an asset class, it is reasonable to assume that new patterns will emerge.
Gold was also considered the go-to asset for inflation protection or sort of purchase power hedge; Bitcoin might have taken that role for a new generation of investors: will gold past returns remain valid in the future considering this?
If a new pattern materializes, it could invalidate the portfolio dynamics shown in PC model. In their latest quarterly letter, GMO provides further examples on how past average returns for an asset class might be misleading in assessing future potential returns. Without delving into the components of that return, they argue, it is impossible to come up with a reasonable estimate of what returns might be expected going forward. Look at the following example for bonds:
The yield component contributed to the lion-share of bond returns. With Treasuries now yielding 1.6%, it is hard to imagine how in the coming 10 years returns will regress towards their historical mean. That would be a wrong assumption.
Past Correlations
A crucial aspect to build an effective portfolio is to use assets that are uncorrelated or, even better, inversely correlated. As for past returns, using a too large timeframe to assess correlations might include periods that had outdated circumstances and have no information value for the future. A timeframe too short might give too much relevance to the inevitable noise in the dataset.
Correlations might fluctuate and mean-revert: an asset combination that worked really well in the recent past can be an indication to start to position your portfolio in the opposite direction. Should your model ‘sin’ (as Cliff Asness might call it) this way?
These are just a few of the reasons why building a risk-parity portfolio, where the correlation matrix is a key input of the model, is more art than science.
Factors
PC model identifies Small Cap Value as one of the assets three MVPs. Unfortunately, we are all well aware of the current existential crisis the value factor is experiencing. While the theory behind the value factor, buying the cheapest stocks in relative terms, should stand the test of time, translating this concept into rules for a passive ETF is harder than slapping the word ‘value’ on a fund name. Various ETF sponsors use different rules to manage their funds or they might decide to change the fund rules while you are invested; building a backtest considering these aspects introduces further complications.
The Vanguard Small-Cap Value ETF, for example, follows an Index calculated by the Center for Research in Security Prices; CRSP classifies value securities using the following factors: book to price, forward earnings to price, historic earnings to price, dividend-to-price ratio, and sales-to-price ratio. The growth in the importance of intangibles has forced quants (index providers) to move away from measures like Price-to-Book into Free-Cash-Flows type of proxies to correctly gauge a firm’s value; if you are interested in understanding more on this concept, this podcast with Michael Mauboussin is great.
QVAL, an active value ETF managed by Alpha Architects, is an example of a fund with a less traditional approach. Here you can see the comparison provided by AA of their fund against the traditional CRSP Index and in the small print how AA defines ‘value’:
The solution adopted by some professional asset allocators is the multi-manager approach. If your total target allocation to Small Cap Value is 20%, invest 10% in the Vanguard fund and 10% in QVAL. This solution avoids the total-ruin scenario of investing in a single fund that goes rogue but introduces other problems, like having to monitor more funds.
Currencies
As indicated by PC, their model results are for USD-based investors. If your functional currency is not USD, you should either expect more volatility (positive and negative) or apply a currency overlay hedge. Whatever you decide to do, just expect your returns to be different. [personally, if you have a long investing horizon I would not bother to hedge, especially the bond component of your portfolio]
Conclusion
PC work is a blast and I am definitely a better investor since I discovered their website. Unfortunately, once you decided on your preferred portfolio/asset allocation, the job is not done yet. It will still require the right amount of oversight, especially for the assumptions made in building it. BUT it is a great starting point!
What I am reading now:
Follow me on Twitter @nprotasoni
If you liked this post, please consider subscribing to the mailing list. If you did not like this post, but you are a good soul, please leave a comment: it is all about learning for me!