Rise of the Machines

Friday, March 27th, 2009

Cringely takes us back to the era of the day trader:

A successful day trader in the late 1990s could gain a following over Internet chat then use that following to make money by becoming an alpha trader. He’d say “I’m selling this” or “I’m buying that” and copycat day traders would do the same. If enough of them acted they could influence the price down or up and — since the leader was leading — he could almost always liquidate his position with a profit. The quickest of his acolytes would make profits, too. Those who didn’t profit weren’t seen as exposing the inherent flaws of this system, they were just viewed as too slow.

To a certain extent, the heirs of day trading have taken the lessons of that earlier era and applied them with devastating effect in the Twitter Age.

If a bunch of wealthy traders get together at Starbucks and agree to short-sell a company or a financial instrument, driving down that price ideally to the point where it never recovers, well that’s against the law. But with trading automation and the Internet as a platform it is possible to accomplish this same end without it being explicitly illegal.

Think about piranhas:

These little guys with their big teeth travel in large schools. They kill and eat their prey, which can be as large as cattle drinking in the river. Piranha, too, take advantage of force times acceleration. The trick is getting a lot of fish — hundreds of fish — to attack at exactly the same time.

How do they do it? How do the piranha know to attack? They don’t wait to bump into a cow leg under water. They don’t sniff for the smell of blood in water. Both of those responses are too slow and would lead to too many victims escaping. Force equals mass times acceleration, remember? And besides, piranhas have tiny little brains to go with their big teeth, so don’t look for any insight there. These are just violent little eating machines.

Piranhas hunt as a school and take all their cues from the fish beside them. Only one fish has to smell blood or bump into some food for the entire school to reflexively attack.

Now we’re back on Wall Street in today’s era of hedge funds and genetic trading algorithms. At any given moment in the market there is more than a $1 trillion in cash that can be brought to bear in seconds by computers that are functioning essentially like piranhas. The cash isn’t held in a few funds or hidden behind some mainframe interface — it is held by hundreds of workstations each operating independently yet as part of a global economic system — conscious or not.

These trading workstations are running in hundreds of offices, all scanning the same data. They have learned over time that certain signals lead to certain outcomes. They may be following an alpha trader but they don’t have to because at some point the market signal, itself, is going to be too strong to ignore.

Here it comes. An alpha trader makes a bold move against a firm or, more likely, against one or more of that firm’s financial products. Say the firm is big stupid AIG, an insurance company, and the instrument is a credit default swap sold by AIG.

Though AIG seems to have forgotten or ignores it, Credit Default Swaps act like insurance and are treated by the market like insurance, but they technically aren’t insurance. They are ultra-hyper-purified demonic risk and nothing else. That’s because CDS’s are not regulated (they are in fact immune to regulation — funny that), they can be shorted without having to ever actually own the underlying security (naked shorts of CDS’s are perfectly legal), because they don’t have to be owned the volume available to be shorted isn’t limited, and — here’s the best one of all — there’s no requirement that the trader have any causal, custodial, or familial relationship with the covered debt. In other words, while most credit default swaps are intended to hedge debt defaults, they don’t have to be. It’s like buying a life insurance policy on the guy down the hall because you hear him coughing at night. His death is meaningless to you so buying the policy is just a gamble, not insurance.

Here’s how it works in practice. The alpha trader senses, guesses, or maybe just wishes for weakness on the part of AIG and its particular CDS issue, so he shorts that mother. The signal from that short (it is big and aggressive, having as much force as possible) is detected by 500 trading workstations running genetic algorithms — workstations that are not regulated in any sense whatsoever. AIG’s CDS begins to glow in front of 500 junior traders. Some programs kick-in automatically and sell, too. The CDS glows even brighter and begins to throb as if its heart was beating. Traders pile-on like piranhas, sensing opportunity, smelling blood, until the CDS is oversold to nothing, until it is dead.

What we’ve accomplished here, through the miracle of synthetic derivatives, is buying a $1 billion insurance policy on a $10 million asset.

Leave a Reply