fat tails

Every method in this series rests on one number: how close a system sits to its edge. Nassim Taleb spent a career arguing that this is exactly the number you cannot trust. For fat-tailed systems the data needed to pin down the tail converges too slowly, and being honest about your uncertainty fattens the tail further. This is the counterpunch, and where it leaves us.

The last essay left a loose end. Markets can sit near a critical edge where small shocks cascade, but critical points are usually finely tuned, so who keeps a market balanced there? The answer, found in a pile of sand, is that nobody does. Some systems walk to the brink on their own, and that is where their crashes come from.

Big market moves often show up with no news to explain them. A hundred-year-old model of magnets shows why. When people copy each other strongly enough, a market can hold two moods at once, and the smallest nudge tips it from one to the other. We build the model from scratch, with pictures, then turn it on markets.

We built fatcrash, a Rust+Python toolkit with 15 crash detection methods: LPPLS, DFA, EVT, Hill, Kappa, Hurst, GSADF, momentum/reversal, price velocity, and more. Tested on 96 drawdowns across BTC, SPY, Gold, 23 forex pairs, and equity crises with honest precision/recall/F1 metrics. Plus: which methods transfer to revenue and profit data.

We assembled forex-centuries, an open dataset of exchange rates, gold, silver, interest rates, commodity prices, GDP, sovereign debt, and more, across 27 sources spanning 1 CE to 2026 and covering 240 countries. Fat tails are universal. Pegged currencies are the most dangerous. Every currency loses against gold.