There is a different kind of financial services trader today. They do not work at stock exchanges or banks.
They work at hedge funds and trade at brakeneck speeds. These “high-frequency traders” (HFT) use computer algorithms—a.k.a., algobots—to squeeze out the most infinitesimal price discrepancies that only exist over the most miniscule time horizons. You can see just how small and how fast we’re talking about in the chart below from a recent paper by Eric Budish and John Shim of the University of Chicago and Peter Cramton of the University of Maryland. It uses 2011 data to show the stock market price difference between futures (blue) and exchange-traded funds (green) that both track the S&P 500. These should be perfectly correlated, and they are—at minute intervals. But this correlation disappears at 250 millisecond intervals, a little more than half the time it takes to blink your eyes.
Background
It wasn’t until the 1980s that a real breakthrough in the logistics of stock trading was actually achieved. The advent of the computer allowed traders to access data on a level never-before-seen or thought possible. With an initial investment of $30 million from Merrill Lynch, Bloomberg designed and built the first computer system to use real-time market data to quote stock prices and relay information.
By the late 90s, the SEC ruled in favor of creating electronic stock exchanges. This laid the groundwork for a new type of trading: high-frequency trading or HFT. In just a couple of years, nearly 10% of all trades were done using HFT with a clearing time of just a few seconds. HFT is actually a thousand times faster than traditional human-to-human stock trading.
High-Frequency Trading Since 2000
The advent of HFT at the turn of the millennium saw trades taking just a couple of seconds to clear and encompassing around 10% of all trade executions. Within just five years, HFT made up 35% of all stock transactions. From 2005 to 2009, high-frequency trading volume increased by 164%. By 2010, however, the first warning signs that this type of trading could be dangerous finally emerged.
HFT was officially responsible for more than half of all trade executions by 2010. In May that same year, computer-based trading sold off more than $4.1 billion in equity holdings, triggering a flash crash in which the Dow plunged 1,000 points in a single day. The initial sell-off triggered a wave of other sell-offs based on the designed algorithms, causing the extreme drop in values. While stocks quickly recovered from the error, the SEC became fully aware of the dangers associated with computer-driven trading platforms.
Trading times reached nanosecond speed in 2011 when a company named Fixnetix developed a microchip capable of processing trades at never-before-seen speeds. By 2012, nearly 70% of trades were accomplished using HFT. The same year brought a wave of HFT investments as well, including a transatlantic cable for the sole purpose of shaving 0.006 seconds off of trading time and a social media-based trading platform that executes trades based on social trending. The concept of micro-trends stemming from popular social media topics during the day helped amplify HFT successes.
As social media proliferated throughout the financial industry, a false tweet in 2013 triggered a brief panic sell-off, wiping 143 points off the Dow in a matter of minutes. The sheer speed and volume of HFT was once again seen with over $600 million traded in milliseconds before the news hit the mainstream media.
The Bottom Line
HFT has to spend money to make money. It’s an arms race, and there’s no silver medal for finishing second. That is because every HFT strategy depends on not only being faster than ordinary investors, but being faster than each other as well. Anytime somebody comes up with a new way to cut a few microseconds—that is, a millionth of a second—off of trading time, they have to spend whatever it takes to do it. Otherwise, they will lose out to their competitors who do. Check out the Intel® Xeon® Scalable Processors video below that explains how these types of speeds are now accomplished.
But it’s also an intellectual arms race, too. HFT isn’t just about the time it takes to send trades through tubes (or between microwaves). It’s also about how much time it takes your algorithms to crunch data. Due to this intellectual arms race, HFT has reduced the duration of arbitrage opportunities from 97 milliseconds in 2005 to 7 milliseconds in 2011.
In Conclusion
When it comes to high frequency trading (HFT), low latency, Intel® Xeon® Scalable Processors and other massive number-crunching analytics, we likely will see the continued push towards higher-density system architectures.
If you wish to further explore High Frequency Trading in depth or discuss how you can boost your high frequency server throughput. Talk to a Hypertec specialist.
This post is also available in: FR
