The Digital Skeptic: Fast Traders Hit Speed Limit

Jonathan Blum

NEW YORK (MainStreet) -- In these nutty digital days, it's no longer what you know, who you know or when you know it.

Now it's where you know it.

"Geography is a new type of natural resource," said Alex Wissner-Gross, who explained the concept to me earlier this year. Wissner-Gross is an institute fellow at the Harvard University Institute for Applied Computational Science. He -- along with C.E. Freer, a similarly super-smart postdoctoral fellow at the Computer Science and Artificial Intelligence Laboratory at MIT -- are deep into the study of where traders trade in today's uber-fast digital financial networks -- as in where geographically, on the earth.

"Your geographic location might be quite financially valuable," Wissner-Gross says.

Back in 2011, the two published an interesting-to-read paper on how information travels on modern, global, so-called low-latency trading networks. These are the digital networks that connect major trading hubs such as the Nasdaq or the New York Stock Exchange to, say, the Chicago Merc or the London Stock Exchange. Trades on these networks supposedly happen in ever-faster millionths, billionths or, I guess, zillionths of a second.

But the more the two studied the reality of fast trading, the more they realized something profound: No matter how fast traders wanted to trade, our world has an absolute, dead-solid trading speed limit in the speed of light. So sure, investor can spend zillions on fast computers, super-duper optical cables and complex algorithms to trade a share of IBM(:IBM) at $196.91 in a bazillionth of second between the Nasdaq and London Stock Exchange. But the information that describes that $196.91 price can only move up to the speed limit of light between these trading centers.

The farther away trades are, the longer it takes for prices to travel from trading hub to trading hub. And the greater the chance for that price to drop out of sync with market reality.

"As that information is flying from New York to London," Wissner-Gross says, "market conditions can change dramatically."

This would all be so much theory if it were not for the fact the jig is finally up on high-frequency trading. Last week, The New York Times ran a front-page story confirming what everybody close to fast networks has been whispering for some time: Low-latency networks are about as fast as they are going to get. And exploiting their inefficiencies for profit is a fool's game.

And voila, blue-sky geeks such as Wissner-Gross and Freer are smack dab on the cutting edge of new approaches to trading.

What the two are arguing -- very coherently, by the by -- is that every fast trading strategy for every stock, bond or financial vehicle should include location as part of an optimal profitable trading approach. The factors in play include volume, price liquidity and the distance between where trades are executed. So say you're once again playing IBM between the Chicago Merc and the markets in northern New Jersey; somewhere in central Pennsylvania is probably the profitable location to trade. If you want to try to get paid with IBM between London and New York, the two say -- no kidding -- that a mid-ocean platform might make the most sense.  

"The fact is, there are specific, direct lines between these trading centers," Freer says. "And for companies willing to pay enough, where along that path you put a trading server gets you more efficient pricing."

"Pop-up Colo"

Wissner-Gross and Freer predict the future of fast trading will be based on so-called "pop-up colocation." (Think pop-up retail, except for location-based trading.) A fast trader will not colocate at a major center such as the Nasdaq. Rather, he or she will do business from an easy-to-install trading server module -- probably a portable shipping container stuffed with trading technology -- that can plopped down anywhere along a trading network route. Sure enough, these containerized server modules are hot items now. Silicon Graphics(:SGI), Cisco(:CSCO) and an operation called IOData all make such container servers. And the bellwether IT innovator, the U.S. Army, awarded a bunch of cloud data center contracts spec-ing these modules.

And here comes yet more grim news: As limits on the speed of trades push more volume outside major trading hubs, investors will soon realize they do not need to pay the freight for said hubs. They will see that, just like in media, publishing, and the legal profession, digital financial trading data turns simply too plentiful to be of much value. Which in turn means investors are looking at yet another bit of digital roadkill: the big, fancy centralized financial trading hub.

Exchanges such as the NYSE or the Merc will probably make excellent museums.