Besser gut schlafen, als gut essen.[1]
This is a book for intermediate (or higher) level investors--the type of investor who's lived through a major market cycle or two and wants a deeper understanding of investing's more abstract (though no less real) risks. If you're a beginner, most of this book will make little sense to you. But if you already have some experience under your belt, you will walk away from this book with a solid intuitive grasp of concepts like sequence of return risk, different types of tail risks, and making yourself "convex" rather than "concave" to different market scenarios.
It's unfortunate, but these risks and ideas simply do not compute for people until they've lived through a substantial drawdown--something on the order of a 2008 GFC-grade decline in their capital. Unfortunately for me, I learn very slowly, so it took me my entire career as an investment analyst plus the 2008 crisis plus reading all of Nassim Taleb's works[2] (uh, twice!) to sincerely and deeply understand how easy it can be to get blown right out of the game, and how catastrophic big drawdowns really can be to your lifetime investment returns.
Also, for something that may frustrate all readers: the author does not discuss any specifics of his actual investing strategy for managing these risks. I understand why this is necessary: an author can't know the specific financial situation of his readers, and furthermore, the instruments he likely uses[3] require deep experience and sophistication. But it does leave the reader in a sort of mental coitus interruptus (wait, when are we going to get to the good part?).
Then again, now that I've started my own Youtube channel where I discuss a range of investing themes, I get why you can't really get into specific stocks, specific options strategies, and so on. You simply cannot tailor specific investing ideas to viewers you don't know and will never meet, and so I find my videos likewise tend to be heavy on theory and concepts and light on nuts and bolts and practical application.
In any event, Safe Haven Investing helped me think through how to better protect myself, and so I walked away from this book a slightly smarter and savvier investor.
[Affiliate link to the book here: https://amzn.to/3hATJfD Note that you can support my work here by buying all your Amazon products via affiliate links from this site, or my sister site Casual Kitchen. THANK YOU!]
Notes/Thoughts:
Introduction:
1) I can't help but read this introduction by Nassim Taleb without thinking back to Taleb's glowing, highly pro-Bitcoin forward to the first edition of Saifedean Ammous's The Bitcoin Standard. Recall that within a bare few years Taleb and Ammous had a vicious falling out, complete with namecalling, Twitter fights, the whole nine yards (see this hilarious example from Ammous, who has a genuine gift for insult). Does this mean Spitznagel's days are numbered? 2) Quick discussion of ergodicity and finance's "mutua muli" (mutually respecting mules), basically a media-fostered circle jerk of pundits and analysts confident in their self-reinforced views; even sophisticated investors with complex financial models can fail to grasp the (often opaque) underlying reality of a marketplace.
Part 1: What Comes First
Chapter 1: At War With Luck
3) His hedge fund Universa, note that it put up a 4000% return during the 1Q2020, the COVID crash quarter.
4) The author kind of wants it both ways here: he thinks this book can help people, but he refuses to explain what he does (saying it should not be attempted by non-professionals, and it probably should not even be attempted by most professionals), "I will not be holding your hand and teaching you how to do it."
5) "...most investors don't think about the impact of the downside the way they need to; they just don't."
6) "Safe Haven isn't so much a thing or an asset. It is a payoff, one that can take many different forms. It might be a chunk of metal, a stock selection criterion, a cryptocurrency, or even a derivative portfolio."
7) "...risk mitigation is investing itself."
8) Risk mitigation must be cost-effective.
9) Sun Tzu's approach to lose the battle in order to win the war: "What will look like a bad idea for one roll of the dice strangely becomes the best idea over many."
10) On predictions: the investing industry way over-focuses on forecasting returns and making predictions; the author, correctly, considers the world of predictions pure bullshit.
11) Note also the paradox of having cost-effective safe havens doesn't just reduce your risk, it actually lets you take on more risk.
12) Investing is a multi-period problem where returns are iterative and multiplicative; compounding this is the "parlay" of our capital. This fundamentally influences the nature of investing in how we think about and interpret returns across our personal lifetime. Contrast this with a hedge fund manager whose incentive fee is based on an annual calendar year performance; see also behavioral finance people who label behavior "irrational" when it appears sub-optimal over a single period framework.
13) Mitigating the risk of a remote extreme loss that never actually happened during an observation period (but still may happen in the future), which is why we look at a portfolio "over a sufficiently broad range of observable outcomes."
14) On the gratuitous mathematical formulas and "solutions in search of a problem" of standard modern portfolio theory. "Modern quantitative finance suffers from a certain science or physics envy." [Criticizing modern portfolio theory is like shooting fish in a barrel of course.]
15) The modus tollens syllogism (if H then O. Not O, therefore not H. Also called denying the consequent):
If my dog is good at catching groundhogs then I won't have a groundhog problem
I have a groundhog problem
Therefore my dog isn't good at catching groundhogs.
16) If a statement is true then so is its contrapositive. You can use it to falsify or eliminate a hypothesis but you cannot use it to verify a hypothesis is true; a type of empirical evidence, Popper's falsification principle is constructed around this.
17) Contrast this with the affirming the consequence argument, a fallacy: "If my theory is correct then we will observe this data. We observe this data, therefore, my theory is correct." The consequent is not proven true here, this is mere extrapolation/inductive logic. Likewise the logical fallacy of denying the antecedent, concluding I have a groundhog problem because I know that my dog is not good at hunting groundhogs.
18) Knowledge can only be falsified, never confirmed.
19) If you're N=1 you get one trial, but if your N is large you can think of success or failure based on many outcomes, many rolls of the dice; thus you then naturally care about the properties over many rules of the dice rather than just one. This gets to ensemble probabilities rather than one-off probabilities.
20) The Bernoulli brothers and their "brachistochrone problem" of identifying what path of rolling a marble down a slope will have quickest descent; discovering that the "straight shot ramp" was actually less fast than a curved, roundabout route; this "would ensnare naive options traders for eternity." [Good one!]
21) See also Bernoulli's work on the law of large numbers, where he inductively reasoned back to a generator by observing the properties of the outcomes, thus making formal statistical inferences, adding mathematical rigor to probability.
22) The St. Petersburg Paradox, via Daniel Bernoulli: what would you venture in a risky game with a payoff that's extremely high or even infinity?
23) One of Bernoulli's key insights was that profit and loss have to be scaled by someone's total wealth.
24) On geometric means versus arithmetic means/expected value. Geometric means are more useful for thinking about ending wealth outcomes. Bernouilli was the first to realize that you cannot look at expected value for an ending wealth outcome, you have to look at the geometric average. "We'll see in the chapters ahead that it literally changes everything, especially our understanding of risk mitigation."
25) Looking at different bet sizes relative to your full stack in the simplified Bernoulli wager. [Parallels here with blackjack, also the Kelly Criterion for bet/investment sizing, etc.]
26) The other St. Petersburg paradox: a way to think about diversification using ships of cargo traveling across the Baltic all of which will face possible piracy, this is what Spitznagel calls the Petersburg merchant trade, a risk mitigation strategy that happens to lose money on its own, yet raises the portfolio's rate of compounding, as well as raising its ending wealth.
27) "This is what's so decisive about the Petersburg merchant trade: the arithmetic cost of its risk mitigation is more than offset by its geometric effect--such that its net portfolio effect is positive." The key point here is that if you ever suffered a one period loss of 100% you're done, the game's over. A single zero makes the whole thing zero, since your ending wealth is essentially a series of parlays. But if you look at it from the standpoint of an arithmetic average, "including a zero" isn't catastrophic, it isn't so bad, it just pulls the average down. The problem is that arithmetic losses are an illusion existing only in a world without time periods. The problem is losses or gains are compounded! They don't get added in, they get multiplied in.
28) No also that large negative returns can pull your logarithmic functions curve downward so the returns get worse the more negative they are, and they end up being far larger than profits of the same (percentage) size can overcome.
29) [This next part of the book really rocked my world on a few levels] See the Nietzsche quote from "The Heaviest Weight":
What if some day or night a Demon were to steal into your loneliest loneliness and say to you: 'This life as you now live it and have lived it you will have to live once again and innumerable times again; and there will be nothing new in it, but every pain and every joy and every thought and sigh and everything unspeakably small or great in your life must return to you, all in the same succession and sequence--even this spider and this moonlight between the trees, and even this moment and I myself. The eternal hourglass of existence is turned over again and again, and you with it, speck of dust!' Would you not throw yourself down and gnash your teeth and curse the Demon who spoke thus? Or have you once experienced a tremendous moment when you would have answered him: 'You are a god, and never have I heard anything more divine.' If this thought gained power over you, as you are it would transform and possibly crush you; the question in each and every thing, 'Do you want this again and innumerable times again?' would lie on your actions as the heaviest weight!
[What's compelling here is when you think about the meta: Would you wish for your life to play on repeat--forever? But then if you knew your life would play on repeat, how would you then choose to live?]
30) Nietzsche's idea can be applied this way to investing: would you make the same investment return if it were an "eternal" return, would you make the same set of investments?
31) [Also worth nothing that this idea occurred in Blake Crouch's novel Recursion, where, interestingly, the people who had to relive a life path over and over again actually suffered, often grievously--which indicates perhaps the quality of life in our modern era.]
32) Now, imagine an opposite thought experiment: where you can run many trials, many alternate lives in a type of investment multiverse. The point of this thought experiment is to get you to think differently about probability; and for me, it also suggests it's a good idea to run multiple portfolios of your own assets using widely different strategies. You'll therefore have different buckets of geometric "parlays" and you'll be much less likely to be blown out of the game.
33) A good example here from the author of a hypothetical dice game with a geometric random walk, where you can see how significant losses destroy your long-term compounding wealth. And because you only get one path through this geometric random walk it gets a lot scarier. See photos of the game and chart of the "return cloud":
34) Note that when we only have one return path through this "return cloud" and it is path dependent with geometric returns (each year's returns are multiplied by last year's ending wealth total), we really care about the path that we're likely to get and we really, really care about avoiding the lower percentile paths.
35) This is the "volatility tax": big down years crush you, absolutely crush you, and it's all the harder to dig your way back out to where you were. This risk is also invisible for people "who live only an arithmetic space."
36) Further we live in a path dependent universe with "absorbing barriers" (if you blow up, you are out of the game, you're done, your fortune is gone); a non-ergodic environment where risks all too often can put you in the ground. "The frequentist perspective of the multiverse is an illusion... It will fool us into doing the wrong things."
37) In contrast, consider a risk-mitigated approach where you leave 60% of your money in cash (earning nothing) and bet with only 40% of your capital at any given time. "By simply betting less of your stack each roll, your arithmetic return falls, predictably... [but] your geometric average return jumps." [This is usually too counter-intuitive for many investors to really grasp, which is why you really have to live through a big drawdown for it to deeply make sense. The betting less of your stack approach makes it impossible to get fully blown out of the game, because at some point--you won't know when--you are going to experience a huge drawdown, and you need to survive it.]
38) This is basically an application of the Kelly Criterion, which sizes gambling bets based on the expected geometric average of ending wealth. See the works of John Kelly as well as the works of Henry Latané. Even Harry Markowitz, the creator of modern portfolio theory, became of proponent of geometric mean thinking but it was long after his MPT theories had taken off in the world of academic finance. See also Ed Thorpe, Ole Peters and also William Poundstone's book Fortune's Formula.
39) [This book has helped me reach an insight--an insight regarding what it's like to be on the other side of this risk trade and be "the casino" or be "the insurance" company. I'd always been confused about why many insurance companies and casinos do quite well when the gamblers themselves obviously don't do well over the long term... but then when you take the other side of this trade (like by selling put options, thus playing the "insurance company" for an equity holder) people still blow themselves up. But the reason "the house" or "the insurance company" earns a superior return is because they are running many discrete, small, simultaneous games, thus they're participating in an arithmetic average-type betting game; whereas you, as a single person (selling put options, buying stocks or whatever), have one path through your return landscape that is still geometric in nature. Thus the solution for this is to bet with a small portion of your stack; or even more ideally, to run multiple separate and totally hived-off portfolios.]
40) "You don't have a room full of Nietzschean demons to play simultaneous, independent games against. (You are not a casino.)" [But note: this is actually a sort of solution: make it so you DO have a room full of Nietzschean demons! Run multiple hived-off portfolios, etc.]
41) Another nuance on the Kelly criterion: consider a Kelly bet size optimized for the 50th percentile return over a return path versus a bet size for a 5th percentile return (at 5% optimized you bet with 10% of your stack versus 40% of your stack for the median optimized). "You can see why professional gamblers often employ a 'fractional' Kelly that size, which effectively maximizes arbitrarily lower percentile outcomes like this 5th percentile curve,.. at the expense of the median outcome." The point here is that the pro gamblers want to make sure they stay in the game.
42) The bet sizing Kelly method alters the nonlinear multiplicative dynamics of the game, it cuts off the left tail of the curve, eliminating the bad contingencies, and this is what raises the median contingency. The central point here is "it's really hard to recover from big losses." Effectively this shifts the entire distribution of return paths to the right. Quite interesting how it works.
43) "Holding back some of your stack" from betting is a store-of-value safe haven. (More on types of safe havens in Part 2: store of value safe have, alpha safe haven, insurance safe haven)
44) Thinking about side bets: insurance as a type of side bet that lets you act as a casino, playing many simultaneous games against fate, such that the really bad fates no longer matter because the other good fates will share the cost.
45) Note that setting aside a portion of capital in cash--since it raises the compound growth rate of the total over the course of your return path--must be providing some hidden benefit to overcome the apparent opportunity cost.
46) By adding an insurance contract the geometric payoff goes up further, even though the arithmetic payoff drops. See photo below from page 93 where insurance shifts the entire curve yet another step function to the right--and not only that, it also moves the fifth percentile line to where it is above zero.
47) Note also that you want to avoid too much insurance: you need only a pinch of salt, whereas more than a pinch ruins the return profile.
48) [Again, it's interesting thinking through how to apply this to a layperson's portfolio: one takeaway is carry much higher cash balances, but another takeaway is to buy some index puts (maybe up to a few percentage points of the portfolio value), or perhaps buy way out of the money put options or certain portfolio positions, something like this?]
49) "Insurance is most certainly not a zero-sum game."
50) "Our aim as investors is to maximize the rate at which we compound our wealth over time on our one and only realized outcome."
51) "Don't plunge off Bernoulli Falls!" We need to think about our geometric average of returns over time, not the arithmetic average, we want to cut off the left tail of the distribution of outcomes.
52) Taxonomy of safe haven types:
* Store of value safe havens
* Insurance safe havens
These are our two strategies for finding safe haven mechanisms. They can vary significantly in their cost effectiveness and we will get more into this in Part 2.
53) "Most importantly, we now have the deductive machinery in place to establish our safe haven hypothesis."
Part 2: What Comes After
54) A taxonomy of safe havens, also note the nuance of knowing something will be a safe haven rather than only knowing it "was" in the past (but possibly or probably won't be a safe haven in the future). On avoiding the "retrospective safe haven fallacy."
55) Safe Haven Phenotypes (photo below):
a) Store of value safe haven: low correlation, provides both a cushion and dry powder should a crash take place
b) Alpha safe haven: negative correlation
c) Insurance safe haven: offers convexity, possible significant upside
56) Most safe havens do more harm than good, or do not provide much portfolio protection when it really matters.
57) Strategic versus tactical safe havens: holding an asset with certain characteristics and letting the dynamic interplay in the portfolio's different parts (strategic) versus moving into cash or a "safe" asset at times when the cost of that safety is low/when times are good (tactical). The latter of course assumes forecasting skill which is not what we're looking for! If you had a crystal ball you wouldn't have to mitigate any risk in the first place. Don't predict.
58) Safe haven "imposters": note the very common sin of market pundits treating an existing or past crash with an "I knew it all along" attitude, and then spinning up a whole prospective risk mitigation strategy based on confidently stated grandiose forecasts. The "prospective safe haven fallacy" includes both arrogance in predicting the next crash as well as arrogance in the characteristics of that crash (what things will go up or down, what things can be used to hedge or offset risk etc). "Heuristic storytelling."
59) The "diworsifier" safe haven: see Buffett: "Wide diversification is only required when investors do not understand what they are doing." Likewise, it is "a confession that you don’t really understand the businesses that you own."
60) "I would add that diversification is a confession that you don't care about cost-effectiveness in your risk mitigation. You just want less risk, no matter the cost." [I should clearly spend some time thinking about this problem, I am definitely guilty of it!]
61) Diversification is "a dilution of risk, not a solution to risk." Also volatility is conflated with risk. Further, since correlations are always higher than they appear when it really matters (during market crashes, times of distress), diversification doesn't work as well as it ought to when everybody wants out of markets (using the book's piracy analogy from the St. Petersburg paradox "the pirates sieze all the ships). Worse, these metrics change every time with every cycle and every crash. "Strategies that were once uncorrelated, stable, and liquid become the opposite of all those things, as investors are forced to sell what they are able to sell, all at the same time."
62) Switching from traditional six-sided dice to a d120 (120-sided) die to approach realistic probabilities. Each one of the sides of the d120 die represents one historical calendar year return from the S&P500; we can then we can do a frequency distribution for S&P500 payoffs. "An inductive die."
63) On emergent properties: the properties of a thing are not the properties of the things that make up that thing. The emergent properties only arise when the whole thing is viewed holistically. "The whole of the wager has properties--specifically its geometric average or median wealth--that the aggregate component wagers do not. The holistic properties emerge from the interactions of those component wagers as they are iteratively rebalanced and compounded."
64) The safe haven or cash on the sidelines provide capital for the next dice roll by resetting or rebalancing the size of the wager after each roll: "Safe havens can thus top up or feed the wagers in the main game, particularly if the previous role resulted in a big loss, without really costing the frequent positive wagers enough to matter.... They actually transform the dice game."
65) Therefore there's a strong emergent property in the investing path over multiple years in the value of the safe haven bet, which as a loss turns out to produce much more significant gains in the aggregate bets over time. This is contrary to all the assumptions of traditional finance, modern portfolio theory. etc., which are more reductionist models and don't address any emergent properties.
66) "Dwell on the past and lose an eye.
Forget the past and lose both eyes."
A Russian proverb that illustrates the dangers of naive extrapolation of the past, but also illustrates the importance of remembering that history rhymes.
67) "Cost effective risk mitigation doesn't just slash risk or systematic exposure; it actually lets you take on more exposure at the same time."
68) "There is only one scoreboard, and that is the rate at which we compound capital on this path--the one that happened. And so we need to get this path right--whether it is the 50th percentile or the 5th percentile or whatever. It's not about getting the expected path right. And, lest we forget, our N always equals 1."
69) Pit stops in auto racing as a similar metaphor: You directly give up time in the race in order to change your tires, such that you can drive faster--yet at lower risk--around the track. The value in the decision comes down to the cost-effectiveness trade-off of the pit stop, but it doesn't come down to the explicit cost of what it feels like to be in the pit stop not moving for ten seconds as your competition flies by. The reason we don't see the fallacy of narrow framing here and "eliminate all pit stops" is because we're focused on the scoreboard (which is the finish line), not each individual lap.
70) Don't forget: any investment strategy or safe haven strategy can only be provisionally successful, they have to prove their mettle in future periods.
71) Note also the risks of naive falsificationism: assuming Newtonian mechanics were wrong because Uranus had a warped orbit (such an orbit in naive terms would "falsify" his theory), rather than speculatively conjecturing the existence of Neptune. All as yet unfalsified theories are provisional, but so too are falsified theories, because experiment conditions can be misleading, or change, or be wrong. Likewise we can be wrong about a payoff profile.
72) Mechanical payoffs versus statistical payoffs, an option that goes into the money versus a "flight to quality" effect that generally drives treasury prices higher. The latter tends to be a type of "payoff" but it may not necessarily be so in future periods.
73) Comparing the safe havens of:
* Three month treasuries (clear drag on performance)
* 10 year or 20 year treasuries (some positive gains from 20 year treasuries but with a lot of statistical noise and payoff variability, this a "hopeful haven" or potentially a diworsifier haven)
* Commodity trading strategy (drag)
* Gold (the effect is positive but very noisy, also it has no yield, note also goal's return has been far flatter and recent decades--gold really only worked well in the 1970s--also, perhaps something is warping its payoff)
74) The author mentions Bitcoin and cryptocurrencies on page 185: it's clear he doesn't understand Bitcoin at all: he "likes blockchain technology" (classic tell of cluelessness) and ultimately concludes that it's a symptom of the "liquidity-fueled environment that created it." Thus it's too speculative. But on the next page, literally the next page, he goes on to explain how his fund Universa's "safe haven payoff profile has been much more explosive than a tenbagger" having no idea that had he understood the value of Bitcoin many years ago he would think totally differently about it if it had "tenbagged" in his portfolio.
75) At the end he considers the insurance safe haven (essentially way out of the money put options), and gold as the only useful and cost-effective safe havens, with 20-year treasuries right on the margin of showing usefulness. And furthermore gold appears useful only because of how great it was the seventies, it may not be very useful going forward.
76) "The message from this is that you don't raise your geometric average or median wealth by taking greater risk. You raise it specifically by lowering risk--the right risk, the worst outcomes."
77) The author gives an example of the difference between an expected outcome and his median outcome by analogy of winning the principal french horn chair of a major symphony (a very unlikely outcome but one that would mean he would love his fate), while recognizing that a "median" outcome would be unacceptable to him (he would curse his fate), thus he threw in the towel on being a professional horn player. This definitely resonates with me: I sort of saw a similar calculus--albeit in a much more limited way--I thought my disappointing outcome if I became a pro musician would be not being able to feed myself, much less ever become financially successful, and so I gave up music too.
78) We get one and only one path. Our "potential paths" are "wondrous to contemplate" but will we love the path we chose (the path we ended up with)? Will we curse or kiss the Nietzschean demon if we're required to live that path over and over again innumerable times? Likewise what would we say about our investing path in this context?
79) The irony is that once-in-a-century stock market crashes happen every several years, so if you can be concave to these crashes (meaning you cut off a large downside and actually have potential exposure to significant upside from these crashes) a very small amount of exposure to these insurance-like products (put options) for example could dramatically increase your returns over your investment path.
80) "It is unsettling to think about randomly drawing our fate from a vast distribution of possible fates. But thinking about it any other way is an illusion."
81) The St. Petersburg paradox basically shows that a large loss destroys so much of your capital, leaving you with a far lower stake to reinvest and compound going forward, that just a pinch of insurance--even though it has an arithmetic cost--would have cut off this left tail experience and increased your geometric return over time. "Multiplicative compounding is a most powerful force (the most powerful force, as some would say). It is to be harnessed and used. But when it throws you over Bernoulli Falls, then it becomes the most destructive force."
82) Returning to Nietzsche's eternal return, he didn't say we have to "appreciate our one and only path--we must love it!" Amor fati: the love of one's fate, an idea Nietzsche co-opted from the Stoics.
83) Per Nietzsche: "My formula for greatness in a human being is amor fati: that one wants nothing to be different, not forward, not backward, not in all eternity. Not merely bear what is necessary, still less conceal it...but love it."
84) We need to evaluate every one period investment as if we were to parlay that investment over and over, compounded many times. "A big loss today will impact your ending wealth decades from now, just as if it happened decades from now."
85) "We need to be robust to the realized path"
To Read:
Mark Spitznagel: The Dao of Capital
David Sklansky: The Theory of Poker
Friedrich Nietzsche: Thus Spake Zarathustra, also The Gay Science, 1882
William Poundstone: Fortune's Formula
Books of George Soros
Books of Ludwig von Mises
Caitlin Zaloom: Out of the Pits: Traders and Technology from Chicago to London
Everett Klipp: Eek (autobiography)
Milan Kundera: The Unbearable Lightness of Being
Footnotes:
[1] "Better to sleep well than to eat well." This quote points to protecting yourself from severe downside first, rather than greedily reaching for stellar returns.
[2] If you're a beginner investor (and big kudos you've made it this far) and you still want to put yourself in a position to wrap your mind around this book, I recommend reading Nassim Taleb's Fooled By Randomness, followed by The Black Swan, and then Antifragile.
[3] If I had to guess--and it's only a guess--he buys way-out-of-the-money put options (likely mostly index puts, but possibly also WOTM puts on specific stocks or ETFs): this would yield a portfolio that violently explodes upward in value during a crash, but will lose a small, knowable amount of money in other periods.