Was Flash Boys a flash in the pan? Nearly a year since the Michael Lewis best-seller assailed high frequency trading as the root of all evil — or at least unfairness — in equity markets, advocates of algorithmic techniques may still be muttering about how their profession was tarred. But they have had to move on; unlike the printed pages of Flash Boys, markets, competition and technology are never static.
Typical of those active in the speed-trading world, John Bates believes the book has been “thoroughly debunked.” Bates, who currently leads the big-data analytics, cloud and industry solutions businesses of Darmstadt, Germany–based Software AG, developed the Apama complex-event-processing technology that became a component of algorithmic trading more than a decade ago. He contends that HFT plays a valuable role in providing liquidity to the markets and will continue to do so, albeit under tighter profit margins and regulatory scrutiny.
And that would be the case with or without the publication of Flash Boys.
Nor has there been any decline in demand for, and innovations in, low-latency data, analytics and trade execution. In-memory database companies McObject and Kx Systems continually ratchet up their performance and submit to the Securities Technology Analysis Center for benchmark validation.
Another case in point is SR Labs, a New York–based market data and exchange connectivity systems specialist. Richard Korhammer, who co-founded Lava Trading in the late 1990s and became CEO of SR Labs last September, notes that regardless of trading or investment strategy, “at any given speed level, someone is going to be at the exchange first.” There are operational differences when the relevant intervals are measured in seconds versus microseconds, Korhammer explains, but speed matters nonetheless — true both pre– and post–Flash Boys.
Lewis’s book brought attention to the upstart IEX Group platform and its order-handling methods designed to level out the advantages of ultralow-latency algorithms. Of course, IEX predated Flash Boys, and it likely would have gained prominence anyway because it offered something that institutional traders were open to and willing to pay for. Look for Aequitas NEO Exchange, a similarly motivated venture in Canada scheduled to launch March 27, to make an IEX-like impact without Flash Boys–like hype.
The overriding message of Flash Boys was that trading speeds had gotten out of hand and needed to be reined in through rule-making or IEX-type speed bumps. But the book did not take note of more nuanced and responsible attitudes in the industry toward the ever accelerating technology.
Back in October 2012, Ari Studnitzer, head of architecture and product management in CME Group’s technology division, wrote on the Chicago-based exchange operator’s website that customers “don’t just care about how many orders can be processed in the blink of an eye.” Instead, “efficiency, effectiveness and market integrity” were overtaking speed as “driving factors of electronic trading,” while CME focused on “increased capacity, consistency and predictability of our platform.”
Some of the hardware that helps push the speed envelope — notably, graphics processing units and field-programmable gate arrays (FPGAs) — are being applied to such constructive ends as operational reliability, real-time surveillance, compliance and risk management. With GPUs as part of its Adaptiv Analytics, for example, SunGard says firms can perform up to 100 million complex valuation calculations per second, nearly five times the rate of a stand-alone central processing unit.
NanoSpeed, a financial industry–focused FPGA company, “provides what is arguably the fastest technical trading solution,” says chief technology officer Sanjay Shah. Meanwhile, its Nano-Risk system checks for fat-finger errors and other risks in half a microsecond — “50 to 100 times faster than using software,” Shah notes.
Late in 2013, Metamako debuted a virtually-no-latency device that is pitched not just to improve trading performance but also to ensure consistency and a “level playing field” for deliveries of exchange data. Co-CTO David Snowdon says determinism is crucial. Whereas systems were traditionally “optimized for average latency,” the concern now is “worst-case latency, and it must be bounded.”
There will be no end to that race.
