Disclaimer: The views and opinions expressed in this blog are entirely my own and do not necessarily reflect the views of my current or any previous employer. This blog may also contain links to other websites or resources. I am not responsible for the content on those external sites or any changes that may occur after the publication of my posts.
End Disclaimer
The Financial Super Friends
In 1993, in Greenwich, Connecticut, a type of arbitrage trading Super Friends was founded. These super heroes couldn’t fly, bend steel, or see through walls, but they were among the most decorated traders in their fields- alchemically creating money from their ideas. Together they would engage in esoteric types of global arbitrage, picking up nickels in front of proverbial steamrollers, but all the while safe because their bets remained uncorrelated. They engaged in trades between pairs of securities that showed a repeatable relationship. If that relationship stretched out of the norm, they would bet on the rubber band snapping back into place.
Their founder, John Merriweather, was by some accounts, responsible for over 80% of all Salomon Brothers profits while running their bond arbitrage desk. For this new supergroup, he recruited the trading Batmans and Green Lanterns of his day, some of whom, including Myron Scholes, had won the actual Nobel prize for advancements in finance.
Drawbacks to practical applications notwithstanding(assumes Gaussian distribution, no market jumps, etc., but hey, I’m cool with it), the Fischer Black and Myron Scholes options model was the formula on which whole volatility arbitrage trading desks were built in the 1980s and 90s. He was the man. It was as if you had hired Jeff Bezos to help you sell books in 1997, Willy Wonka to help you sell chocolate, Snoop Dog to help you sell…well you get it.
The Fall that happened after the Pride Goeth
Except that it didn't last- the fund known as Long Term Capital Management, cool enough to be known by its own acronym, LTCM, blew up. After the first few years of pantsing the market, other hedge funds caught wind and began adopting their types of trades, spreads got smaller. Trading alphas have a decay. Not all Australian swans are white. Russia defaulted on its debt, and uncorrelated trades ‘correlated’ as traders rushed to sell positions to raise capital. The Super Friends fund had levered themselves to turn penny trades into dime trades. They were over their skis, against a 1-in-whatever-year chance of things happening, caught swimming naked as the tide went out. To paraphrase Keynes, market irrationality was pushing LTCM towards insolvency. It happens.
The Financial Super Friends of their day didn't see it coming, but their business demise wasn’t completely in vain. We can learn from it.
What follows is an incomplete tour of learning how to fail from other people’s business collapses. Keep your hands and feet inside the tour vehicle at all times! We will be passing by the Hype Train exhibit, past the cognitive and behavior bias paddock ( be careful- you won’t see them coming), and spend some time looking over failure mosaics. Exit through the gift shop.
The Hype Train Part 1:
“I hear the Hype Train a-comin', it's rolling 'round the bend” ~Johnny Cash
Sometimes the force of the hype train is so powerful, the headlights (Deadlights?) so bright, that everyone involved acquires a kind of collective blindness around the the same time, let’s call it, a couple years. We are witness now, to the aftermath of the most recent Hype Train derailment.
The cryptocurrency exchange FTX was once valued at $32 billion dollars. Its founder, Sam Bankman Fried, was worth some portion of that. Time magazine wrote about him for their 100 most influential people list (in 2022!- whoops), “ In a crypto landscape ridden with scams, hedonism, and greed, Bankman-Fried offers a kinder and more impactful vision brought forth by the nascent technology.”(h/t Michael Lewis). Hmm, duly noted.
After its collapse, the newly installed CEO John Ray (former chairman of Enron Creditors Recovery Corp.) appeared before the U.S. House Financial Services Committee and revealed that FTX had been using the small business accounting software QuickBooks (coincidently also used by bankrupt crypto lender Celsius) to keep track of its finances. "Nothing against QuickBooks. Very nice tool, it's not for a multibillion dollar company."
Why didn’t anyone ask? Almaeda Research is rumored to have had a fake pool of reserve funds that was recalculated each day, to look like the funds fluctuated, with a randomly generated seed from a Python program. Holy cow. Those lights sure are bright.
But now, allow me a brief digression(elevator music plays)…
[Digression 1: A quick heuristic Hype Train Algo (HTA) to run at your ‘freakin awesome’ new company:
Every 3-6 months do a “pre-mortem” on what it would take for your business or company to fail fast. Check all that apply:
Your founder is flying towards the sun with a feather tracksuit stuck together with wax (how was that ever going to work, by the way?) All aboard! (Ozzy voice)
Your company has entered into its 4th indecipherable, off the books, joint venture this month
Your fund’s worst return in 215 months is -64 bps ( tough one -could be Madoff or Ren Tech- Peter, call me!)
Something called the Iphone comes out (but I love my full qwerty keyboard phone!)
Your boss tells you that they grow out their hair for a better bonus
and similar…
Digression 1 End ]
The Hype Train Part 2:
“Come on, ride the Hype Train, It's the choo choo” ~The Quad City DJ’s
During the early spring of 2016, in a building on Page Mill Road, in an area that is part of the Stanford Research Park, a different type of SuperFriends was being assembled. Their powers were not unique- the ability to not ask questions, to stay in their lane, and to be experts in fields other than the field for which they were being hired. They also shared a conspicuous brotherhood, being part of one of the “whitest, old man” boards of all time (Im sure there are plenty of contenders- to the discussion boards!) Elizabeth Holmes brought together a former Secretary of State, a power lawyer, and a future Secretary of Defense, among others, for her Theranos Board of Directors, but no medical CEOs or research scientists. Turn off the lights you’re blinding me of my ability to ask obvious questions!
The No-Stress Tests
If you go to a hedge fund that does quantitative trading and try to sell your quant model, chances are (hopefully) they will run your results against a series of historically bad events to see how your model would have performed, a process known as stress-testing. The 1998 Russian default, 9/11, 2002 tech bubble, 2008 mortgage crisis- a test which, by its very nature, captures whatever the most-recent-worst-thing is. No trader or portfolio manager enters those meetings with a model that would have failed a historical stress test. The event or crisis that blows up your model or puts your company out of business hasn’t happened yet (Otakar G- delivering obvious insights since 2023!). This statement’s not meant to be a tautology. Model building and initial deployment should, by their nature, ensure that the models are bulletproof from the past. If you jerryrig this on purpose, however, then insuring your model against the past can be considered a form of ‘data dredging’ which itself is a subset of ‘p-hacking’ (and in scientific research- it’s often impossible to replicate the findings from research papers- this flavor is called ‘Publication Bias’). Traders tend to overfit to the past to get hired, but their lack of generalized risk modeling may doom them for what’s to come.
How Bad Can it Get?
In the (re)insurance industries, there’s a type of insurance that covers all manner of cybersecurity and data breaches that is generally called “cyber insurance”. Cyber insurance modeling companies now sell the ability to tell you how bad a coordinated and distributed global cyber attack could be, like the property catastrophe (CAT) modeling firms before them. This is a very hard thing to successfully parameterize in a complex and evolving sector. GenAI has made it even harder. There is no stochastic simulation, no roll of the Monte Carlo dice that can give you a more practical answer than “really bad, all of it”. That’s why the first rule of (re)insurance( any trading ) remains “don’t blow up”. There needs to be a cushion big enough that no answer from one of the stochastic roulette modeling firms is life-threatening. This leads practitioners to become comfortable with their stochastic results, deterministically. Set the number of iterations, level of severity, then feel comfortable that the answer is the worst it can get. Comfort and over reliance on standardized risk models is said to have been one of the factors that contributed to the quant trading blow up of 2007. As Nassim Taleb has said, “VaR & tools ‘modern’ portfolio theory have been blowing up funds & banks since first use in 1987.” Can I get an Amen? Any widely adopted and accepted risk system, whether stochastic or factor based, can lead disparate practitioners to collectively pool their bets in those places the model deems to be “less-risky”.
The Survivors
Every company that weathered the storm, every start up that made it- they serve as the deterministic model for what to do because they survived. I’m going to give you the granddaddy of all Data Science stories- the sine qua non of PowerPoint presentation anecdotes. During World War II, at Columbia University, mathematician Abraham Wald was tasked with helping lower the number of US bomber fatalities. While reviewing photos of returning bomber planes, one of his scientists suggested adding armor plating in the areas where the planes riddled with bullet holes. Wald suggested that, instead, they add more armor to areas on the airplanes completely free from bullets, like the cockpit, tail, and propeller housings. Those were the areas on the returning planes left relatively unscathed from machine gun fire. Queue the punchline. The returning planes were only telling part of the story- that of the survivors. Survivorship bias is one of the most prevalent behavioral biases that prohibit people from, in part, divining the split between luck and skill in success (or in actually surviving).
History is written by the victors who got the funding, the models that traded unscathed through the financial storms, by the people that “saw it coming”, but there is a lot to learn from historical failures despite cognitive and behavioral bias blind spots.
The Failures
Learning how others fail helps you avoid your own. That fact that history doesn’t repeat itself, but that it rhymes, is a well-worn aphorism. In business failures, it's often a lack of board or management oversight, a reticence to pivot, and a desire to continue doing things the way they've always been done. In trading, the failure function may be the use of leverage and the emergence of a previously unforeseen, correlating forced liquidity event. And in both business and trading, the cause is sometimes an underlying hubris or, in the most spectacular cases, outright fraud.
There is an applied utility in examining companies that have inadvertently optimized for ‘anti-success’. Quotidian failures and death by a thousand cuts seldom get enough attention, but gigantic financial explosions are combed over- Kaggle even offers a half a million of Enron’s emails to explore and model.
Piecing together a failure mosaic comprised of all the disparate tiles of past failures can help you avoid your own (or your company’s).
What’s past is prologue, but the rest is still unwritten (a Bill Shakespeare/Natasha Bedingfield combo). Learn from the past, but don’t be wedded to it. Don't blow up. You know the drill. Don’t slow down.
A Wholly Incomplete list for Learning how to Fail (please send me books recs for what you think I left out)
+ Something a little more technical
Dynamic Hedging:Managing Vanilla and Exotic Options, Nicholas Nassim Taleb
And 1 Paper…
Thank you for your kind comment! I'm glad you enjoyed reading the article.
Really enjoyed reading this not that incomplete article; well done - thank you!!