Case Studies

Products explain what an instrument is; concepts explain how it works. These 40 pages explain what happened when it went wrong — mechanically, without moralising, because the mechanisms repeat. Each case links back into the products and concepts it used. Education only, as everywhere here — see the disclaimer.

The badges say how much a page assumes: Easy everyday words, nothing to look up · Medium one idea you may need to look up first · Hard written for people who work in the market.

The library

• Herstatt, 1974Mediuma bank closed between the two legs of a currency trade, in the hours when one side had paid and the other had not. Settlement risk acquired a surname.
• Northern Rock, 2007Mediumthe run happened in the funding market weeks before it happened in the queues, which is the part the photographs left out.
• The Flash Crash, 2010Mediuma market that fell and recovered inside half an hour, with shares printing at a cent. Liquidity turned out to be a service somebody chooses to provide.
• The Greek restructuring, 2012Hardclauses inserted into existing bonds by statute, and a hedge that argued for weeks about whether it had been triggered. Governing law as a term of the trade.
• Enron, 2001Mediumprofit on a twenty-year contract recognised the day it was signed, funded through entities the accounts did not consolidate. Both halves were disclosed; neither was legible.
• Madoff, 2008Easyone firm decided, held and valued. Every check a fund structure provides is one party checking another, so collapsing the three did not weaken them.
• The LIBOR settlements, 2012Mediuma rate that set payments worldwide, produced by asking banks a question about a market that had thinned out underneath it.
• Greensill, 2021Mediumshort, granular, self-liquidating, insured. Four properties that made the asset safe, and the marketing kept all four after the asset kept none.
• Evergrande, 2021Mediumpresales, supplier credit and offshore bonds, three funding sources with one common cause — and the bondholders in a different queue from the assets.
• Woodford, 2019Mediuma daily-dealing fund holding things that took months to sell. Meeting redemptions with the easy holdings concentrates the problem on whoever stays.
• Terra, 2022Mediuma peg defended by a token the system could print. The mechanism that held the peg is the mechanism that destroyed it.
• LTCM, 1998Mediumconvergence trades at thirty times leverage, and the discovery that dozens of "independent" positions were one bet on liquidity.
• Subprime & the CDO Machine, 2008Mediumhow a single national bet on house prices was tranched, relabelled AAA and multiplied synthetically.
• The UK LDI Crisis, 2022Hardpension funds hedged correctly and still nearly failed, because variation margin is due the same day.
• Archegos, 2021Mediumone family office, five prime brokers, and leverage that was invisible because it was fragmented.
• Volmageddon, 2018Mediumshort-volatility products that had to buy volatility precisely as it spiked.
• Negative Oil, April 2020Mediuma physical delivery constraint priced through zero in one afternoon.
• The Swiss Franc Floor, 2015Easythree years of suppressed volatility resolved in thirty per cent and a few minutes.
• FTX, 2022Easynot a market accident: custody that wasn't custody, and the run that revealed it.
• Barings, 1995Easyone trader running both his own desk and the office that checked it, and a 233-year-old bank sold for £1.
• Metallgesellschaft, 1993Mediuma hedge that was economically sound and financially fatal, because only one leg paid cash daily.
• Amaranth, 2006Medium$6.6bn on natural gas spreads, lost not to a wrong view but to owning half the market.
• The AT1 Write-Down, 2023MediumCHF 16bn of bank capital to zero while shareholders below it received stock. The hierarchy was a document all along.
• The Short Squeeze, 2021Easya crowded short, a coordinated crowd, and the week clearing margin became the story.
• Black Monday, 1987Easythe largest one-day fall in modern history with no news to explain it. Insurance that turned out to be a synchronised sell order.
• The Asian Crisis, 1997Easycurrency pegs, borrowing in a currency you do not earn, and the template every emerging-market crisis has followed since.
• The Dot-Com Bust, 2000Easythe technology prediction was correct and the prices were not. Why being right about the future is not being right about the price.
• Wirecard, 2020Easy€1.9bn of cash that never existed, inside a blue-chip index, audited for years. What index membership certifies, which is nothing.
• The Nickel Squeeze, 2022Mediuma producer hedge became the squeezed short, and the exchange cancelled the trades. What a cancelled market does to the meaning of a price.
• Silicon Valley Bank, 2023Easya bank that took no credit risk and failed in two days. Held-to-maturity accounting, a concentrated depositor base, and a run at the speed of a group chat.

What they have in common

• Leverage sets the clock. In LTCM, LDI and Archegos, the trades were defensible; the funding schedule was not.
• Correlation is a peacetime measurement. Every case where "diversified" positions failed together had one hidden common factor — liquidity, house prices, or a margin call.
• Mechanics beat opinions at the deadline. Expiry dates, rebalancing rules and margin schedules decided the outcome in 2018, 2020 and 2022 — all published in advance.
• Suppressed volatility is stored volatility. The calm before 2015 and 2018 was the position, not the environment.
• Controls decay as success grows. Barings and Amaranth both handed more capital and less scrutiny to whoever had made the most money last year.
• The legal claim outranks the economic story. FTX and the 2023 AT1 write-down both ended on what the documents said, not on what holders believed they owned.
• Nobody in these stories was doing anything unusual — which is exactly why the mechanisms are worth knowing.