Opening Cross: Hey, CERN, ‘Negative Latency’ Is Old News

The recent news that CERN, the European Organization for Nuclear Research, has apparently managed to transmit neutrino particles over a distance of more than 700 kilometers faster than the speed of light, has a lot of people in the data industry—where data latency is a big deal—very excited. However, while many are trying to achieve light speed (never mind even faster—after all, light isn’t necessarily the fastest thing; it’s simply the fastest that we know about), others are concentrating on something that can’t be achieved with faster hardware alone: “negative latency.”

For some time, I’ve been looking for a term that would encompass the new products and services that try to give traders an edge by creating information that pre-empts an event, rather than trying to just get information about an event to the trader as soon as possible after it occurs. And at last week’s European Financial Information Summit, I found just the catchphrase, courtesy of industry analyst Bob Giffords, who says he’s been banging on about the concept for years.

“With ultra-high-frequency trading, what you see is no longer what you get, even for co-located trading engines. So traders are now trying to anticipate market moves rather than just react to them. This creates a search for ‘negative latency’—the ability to take trading decisions in advance of the tape, even before you see a price move. In this scenario, the price feed is used increasingly to corroborate or calibrate the predictive model, rather than as the trigger for a decision,” Giffords says.

“The search for negative latency is leading to detailed analysis of tick data behavior in the sub-millisecond range and correlations between different asset classes. Over the past few years, research suggests this has become increasingly autocorrelated rather than random. This gives rise to inevitable micro-oscillations which are essentially noise. Some traders are even applying FPGA-based pattern-matching to recognize and respond to such complex patterns in real time,” Giffords adds.

However, I’d like to extend the concept beyond the high-performance trading space to encompass new types of data and analytics designed to act as an early proxy for traditional data. For example, industry messaging cooperative Swift recently announced the Global Swift Index, which will use message traffic over its payments network as a proxy for GDP data. The aim: predict economic figures before they are released by countries. Another example is MarketPsych’s upcoming Market Risk Index, which uses social media chatter to gauge stock sentiment, and apply that to market movements, to predict when prices will rise or fall.

The concept is also being used by long investors as a new component of research and commentary services, creating “ecosystems” that monitor everything around a company or sector to understand when something will affect that company before the rest of the market finds out—not by lowering technical latency after the fact, but by finding out before news breaks. For example, there is a lot of value to finding out about a fire that shut down a relatively unknown factory in China if that factory supplies minor yet crucial components to a major technology vendor—it could slow down production, impact quarterly sales, and ultimately cause its share price to drop. The “negative latency” component of this isn’t being the quickest to find out about the fire; it’s about knowing about the factory and what it produces in the first place.

Using intelligence to get ahead of the market in these ways potentially eliminates latency concerns of trying to be the first to capture and respond to price movements—though it may introduce a new wave of latency issues if everyone tries to identify and act on the same leading indicators. And then, CERN’s faster-than-light tests might take on new meaning for the markets.