Ultrafast Trading Robots Hunt In Packs, Outside Of Human Control

High-frequency trading, a type of algorithmic trading, has taken over the global equity markets. These (almost) autonomous systems have also been blamed for many unusually violent swings in the stock market and for "flash freezes", computer glitches that abruptly brought stock market operations to a halt.

A recent paper, published in the journal Nature Scientific Reports, studied these glitches to understand the root cause of the problem. The study identified the emergence of an "ultrafast machine ecology" that is beyond the control of human traders!

The research team analyzed data (a high-throughput millisecond-resolution price stream of multiple stocks and exchanges) over a five-year period, from January, 2006 to February, 2011.

They also defined an ultrafast extreme event (UEE) - one where a stock price moved at least 10 consecutive times in the same direction, all within 1.5 seconds. The total magnitude of these mini-crashes and rises had to be at least 0.8 percent, representing 30 standard deviations from the normal run of trading.

Why did the researchers apply a threshold of 1.5 seconds? 

Because, at a minimum, a human being takes approximately one second to react to potential danger! HFT chips, on the other hand, can operate in a fraction of a millisecond (1 millisecond is 0.001 second)

Thus, by applying this threshold, the researchers ensured that there was no human intervention in all these extreme events.

The researchers identified 18,520 UEEs from the total data. As they narrowed the time window to below one second (the limit of human reaction time), the authors found a dramatic increase in the number of crashes and spikes.

Using mathematical modeling, the team concluded that this behavior was the product of ultrafast computer trading only.

What it means, in sensationalist terms, is that the ultrafast computer algorithms were exhibiting a collective predatory trading behavior, like wolves hunting in packs!

The reason for collective behavior is easy to understand - to enable faster processing, most of these ultrafast algorithms are relatively simple. As a result, they are more likely to start adopting the same behavior, and attack the same part of the market.

The team also observed that there was a definite linkage between UEEs and the financial crisis of 2008:
  • UEEs experienced a very significant growth spurt just as the financial crisis started in 2008
  • The 10 stocks that experienced the most UEEs were all banks

The research makes a compelling case that conventional market theories no longer apply in the new world of ultrafast robot algorithms.

What remains to be further investigated is:
  • What triggers these UEEs
  • Can changes in regulations or strategies eliminate these UEEs

Source Credits: