Wednesday, October 06, 2010

Wall Street is Googled

Bookmark and Share
There's a very interesting segment around 40 minutes into the following video on "quants" reviewing how the "black box algorithms" being used in high-frequency trading are literally googled into internet news streams to make instantaneous large-scale buy and sell decisions:

Here is what this can do when a simple false piece of information gets processed:

On September 8, 2008, the price of UAL shares fell by nearly 99% in fifteen minutes to $0.01 US amid rumors of another bankruptcy, before NASDAQ temporarily halted trading. The rumors were traced to an old story on the South Florida Sun-Sentinel website about the 2002 bankruptcy being picked up by Google News and subsequently presented by Bloomberg LPas breaking story. The share price subsequently recovered most of its value. [Wiki | More]

Price "value" can simply go *poof* in a matter of minutes no matter the worthiness of the effective information.

The panic decision-making system is in place as never before and is literally connected to the internet information stream.

With the "Flash Crash" in May we got to see just how dangerous the current state-of-affairs on Wall Street can be when the mood of the crowd shifts:

*YouTube playlist of the whole May mini-panic*

One of the principle indicators of the May flash crash was the TED spread. Monday it started to climb, leading to the sell-off in markets that day.

I believe this is key. The financial system might be overly-sensitized as recurring crises have been "modeled in". Thus, just the hint of trouble can trigger over-reactions that in and of themselves have an effect creating the potential for a self-reinforcing "falling dominoes" dynamic of extraordinary magnitude. Since quants seek to program in a reasonable automated response to the potential unreasonable responses of other market participants, they're collectively creating the basis for efficient insanity. It's as if the dynamics have formed for man as a species to suffer a collective psychotic break with little warning and lead-up. Even so, some "new information" would have to enter the "system" to trigger a given panic. Seasonality suggests we are at the end of the window of maximum vulnerability in 2010 to such a dynamic today, although between now and early-November remains a period of significant concern.

No comments:

Related Posts with Thumbnails