SCV, long vs short duration bonds, commodities, TIPS, whatever. It's all overfit, particularly when you consider that every big economic crash had very unique causes and effects.
I've been diving deeper into this line of thought.
Each of the worst times to be a retiree investor have looked different from any other time. If you loaded up on bonds in the 1960s to hedge the risk of another Great Depression, you got destroyed by the stagflation of the 1970s. If in the 1980s you loaded up on gold to hedge the stagflation scenario, you had dead money for decades, harming your SWR. If you said f*** it I'm going YOLO with stocks in the late 1990s you got bashed by the massive 2000-2003 dot-com correction. And then you went into housing / mortgages in the mid-2000's, which had never before experienced a wave of defaults or housing deflation and
that seemed historically safe!
I'm sure at each step of the process, portfolio optimizers were thinking of all the ways things went sideways in the past, and considered this set of info to represent the full universe of ways things could go sideways in the future. Yet, each time, markets found a new way to destroy portfolios which had hedged heavily against the previous scenario.
So what are we doing today? Looking at the past 150+ years of stock and bond data to determine what can possibly go wrong, and applying the hedging strategies that would have worked in the past!
We're doing this even though the strategy has never worked before, no scenario has ever replicated itself exactly, and markets have always surprised us with a new way to fail that invalidated the earlier approach. We're calculating portfolio failure odds in the low single digit percentages with decimals, as if the entire universe of things-which-could-happen-but-haven't-yet doesn't exist.
For example, who is hedging against a 30%+ U.S. dollar devaluation? None of us.
If you don't have that one on your bingo card, that's because it's never happened. If you were a British, Turkish, Argentine, or Russian investor, OTOH, such an event is with living memory. How about the downfall of democracy, and the institution of a one-party state that limits your ability to invest internationally? What about a decade-long reversal in productivity? Again, it's never happened to U.S. investors, but it has happened multiple times to others, so maybe we all need to be thinking about currency diversification and offshore accounts/exchanges.
The problem when you start imagining a diversification strategy that spans countries, currencies, and exchanges, is finding a lot more failure modes in this broader historical dataset. More ways to fail leads to a perception that the SWR is much lower by following such a strategy. Yet what we're doing here is arguably uncovering more about what could go wrong, and what any portfolio's actual odds are. It doesn't make sense not to model a 30% currency depreciation in a totally USD account, but to model it if your account includes other currencies.
Expanding horizons is the opposite of fishing for a low SWR in a subset of data from one country (the most economically successful one), one currency (the one which became global reserve currency but did not lose that status during the sampling period, as all prior reserve currencies did), and with only the sorts of assets that were available long ago and also have long data series (i.e. not rent houses, black market precious metals, options or futures for hedging, cryptos, annuities, life insurance policies, iBonds, target date funds, or TIPS).
The counter-argument is that civilizations are slow to change, and the U.S's policies favoring business, technology, education, a strong currency, and social stability are unlikely to be modified in the next 30-40 years, so future crises will be comparable to the past.