Thread by @gelmanisaac

Fraud topology and parting #RWRI stuff.

2008: We consolidated banks instead of breaking up banks.

Coordinated attacks on a small number of hubs can take out the entire network.

Biggest threats according to FSARC: maliciously executed trades, multi-day of a bank's wholesale payment system, multi-day SWIFT outage. Mostly talk unfortunately.

Moving to cashless society:
1. cost savings in not printing money and ease of transaction
2. crypto can dematerialize physical financial instruments
3. all money movement subject to surveillance and blocking

Today: one user one wallet
Future: one user/one wallet with banks in control of client private keys

Biggest indicator of blowup is high returns with low volatility

Parting stuff:

1+a to detect convexity
Better to be convex than wrong
Measure extremes if you have to measure
(Tail wags the dog under fat tails)
Mean absolute deviation is better than standard deviation
(Corrollary: standard deviations are for people who use Excel)

High return, no volatility is the path to bankruptcy
(Bob Rubin trade)
(Be niche, be a boutique boutique)
Cutting the tail is always cheap
1/N for Baal's sake: if you are in a fat tailed world you will always be under diversified.

Distance from worse is potent information. Under ergodicity, survive. What would bankrupt you? Then work backwards. Actionable risk estimation: know where the problem is coming from. "1% chance of crashing" vs. knowing your tires are naked and it will rain.

Oil futures in 2020: contract is not an option; you have an obligation to take the oil and store it without violating EPA rules. You have to pay somebody to not deliver you the oil.

Asymmetry in statement. Absence of evidence is not evidence of absence. Linked to Pascal's wager.

Cost of masks vs it's nonlinear payoff -- CDC and WHO did not get the asymmetry.

Options are convex to volatility.
The Lucretius problem: tallest mountain you have seen is the height of tallest mountain that exists (mistake of Fukushima).

Quantification of risk: use the Russian approach (inequality). Give the shape of what can happen and the dynamics of the uncertainty.

5 short term options better than 1 option for 5 years.
Shipping option: never intrapolate.
Don't optimize allocation. Optimization increases fragility.
The more precice the regulation, the more arbitrageable.

Shadow mean != observed mean, and the diff increases with fatter tails.

Ergodicity -- see prior thread.

Correlation is not correlation.
Starting a business.
The box cannot measure itself.
Facts aren't anecdotes: gives information about probability distribution and tails. Anyone know here to find @financequant's source on this?

In any transaction there is sucker and a suckee.
If it makes sense don't do it. You are the suckee.

Only the hyper paranoid survive.
Under fat tails, max estimate the mean better than an average.
Dynamics over statics. (Log is magical).
Computational irreducibility. No function to skip steps, making predictability very hard.

Island stabilization law: if you want to do GMOs do it on an island. If you want to escape COVID go to New Zealand.

You might have a bad speedometer but it can tell you if you are accelerating.
Drawdowns >> volatility. Drawdowns are ergodic.
Don't separate insurance from the product.

You are overallocated.
Never say "let's assume...". Protagoras/Thaler problem.
Multiscale survival.
Concentration and convexity of harm
Sleep test. Be able to sleep well. FU sleep.

Goat v Ferrari: goat can go faster than Ferrari in desert or mountain.
No energy constraint in the system=> fat tails.
#RWRI #RWRI14
Thank you kindly Nassim @nntaleb and all instructors!