Skin in the Game, Nassim Taleb
Note the cover: an elephant, which never forgets, and which has a broken tusk ... i.e. it has taken some risk and it has survived. It has, in a word, skin in the game.
Go and read this book. Or if you can't find a copy of it, at least go over to Taleb's Medium site, Incerto, and start reading; a fair proportion of the book is contained in a series of articles at that site. Jesus,
here.
That said, the entirety of the book isn't in the articles, and for some reason I can't explain - maybe it's just the difference between reading on the Internet and reading a text - there were lots of little connectors and insights to be found in this.
The most salient stuff I can think of right now, which I haven't otherwise touched on in posts here and elsewhere, is this:
The deepest insight to be understood in the book here is the concept of
ergodicity, an example of which is: if you base your probability of winning at the casino on some idiot of a social scientist who concludes out of a sample of 100 punters going to a casino once that 40% of them will come out ahead, you are asking for trouble. This is an
ensemble sample, deciding a probability based on a group of people
in single transactions. It is also how most social science assesses the subject of probability. And what it misses is the other way of sampling:
time sampling, which is to just take one of those punters and see what his likelihood of winning is when you run
him through the casino 100 times. In his case, bankruptcy will arrive much faster than the 100 visits to the wheel. If he's wiped out on visit 28, there will be no 29. Which makes his actual probability of wipeout 100%.
That is, most probability in this area doesn't account for what the risky action you're doing does to you over time. Ergodicity, in essence, is to take into account the probability of (1) ruin, and (2) the risks inherent in repeated exposure. If there is even the possibility of ruin in an act, all cost-benefit analyses are essentially worthless.
Put it another way: suppose you are to win 1 million dollars on Russian Roulette. Naive statistics say your notional expected return from any single round of Russian Roulette is about $833,000, since if six people play Russian Roulette at once, five out of six of those people will get a payout. The moment you as an individual play more than once, though, your expected return is "not computable", i.e. it breaks down entirely since
your odds of getting the bullet increase each time.
Cost-benefit analyses conclude that it's fine to smoke any single cigarette, since your odds of getting lung cancer from any one cigarette are tiny. But that's not the point: the risk from lung cancer comes from repeated exposures to the risk.
Taleb reduces it to one line: "
If one claimed there is 'statistical evidence that a plane is safe', with a 98 percent confidence level (statistics are meaningless without such confidence bands) and acted on it, practically no experienced pilot would be alive today."
Makes you think again about Superman's joke line that flying is statistically speaking still the safest way to travel. Too bad the statistics don't account for the fact that ruin is essentially inevitable if the numbers don't go your way: not a lot of people survive plane crashes. Ergo, the statistics are meaningless. Your paranoia about getting on planes is entirely justified, very much so if you're a frequent traveller and thus expose yourself repeatedly to the risk.
It's when you start applying this form of logic to stuff that "in small doses" apparently does no harm to you -- for example, half of Monsanto's back catalogue of hits -- that you start to think about a lot of things like that, and a bit of a chill runs down your spine. Often, the probabilities are spun this way precisely because of the time sampling problem. Of course the bonus glysophate in the Roundup won't hurt you, the probability is, like, 0.00001% or so that a single dose of the stuff will do you any harm. We won't talk about the farmer who uses it every season, or when it builds up in his system, or the fact he exposes himself to it every time he walks out to his crops and takes a big breath.
This idea is really another facet of the idea of black swans, which has been part of the Incerto since the first book: there is too much opacity in real world probabilities for arrogant science or arrogant statisticians to accurately predict it. Thus, the Black Swan. And thus the two results of it, depending on whether you live in a Mediocristan or an Extremistan: recoverable harm, or catastrophic damage to the system, respectively.
This book is like a keystone to a lot of Taleb's concepts, as with
Antifragile it's a real paradigm-changer, although probably less so to those who have read some Taleb already. Either way, the main thing to recommend Taleb is the fact that he has the runs on the board to verify his point of view on life: 20 years as a trader, made his money against the market in 2008, based off the ideas and insights he'd already used and propounded in this series. Read it, read it, read it. I'll go to the Taleb thread a bit later to expand on some other stuff that was already in the book.