Home → Magazine Archive → October 2013 (Vol. 56, No. 10) → Online Algorithms in High-Frequency Trading → Abstract

Online Algorithms in High-Frequency Trading

By Jacob Loveless, Sasha Stoikov, Rolf Waeber

Communications of the ACM, Vol. 56 No. 10, Pages 50-56

[article image]

High-frequency trading (HFT) has emerged as a powerful force in modern financial markets. Only 20 years ago, most of the trading volume occurred in exchanges such as the New York Stock Exchange, where humans dressed in brightly colored outfits would gesticulate and scream their trading intentions. Today, trading occurs mostly in electronic servers in data centers, where computers communicate their trading intentions through network messages. This transition from physical exchanges to electronic platforms has been particularly profitable for HFT firms, which invested heavily in the infrastructure of this new environment.

Although the look of the venue and its participants has dramatically changed, the motivation of all traders, whether electronic or human, remains the same: to buy an asset from a location/trader and to sell it to another location/trader for a higher price. The defining difference between a human trader and an HFT is that the latter can react faster, more frequently, and has very short portfolio holding periods. A typical HFT algorithm operates at the sub-millisecond time scale, where human traders cannot compete, as the blink of a human eye takes approximately 300 milliseconds. As HFT algorithms compete with each other, they face two challenges:


CACM Administrator

The following letter was published in the Letters to the Editor of the February 2014 CACM (http://cacm.acm.org/magazines/2014/2/171678).
-- CACM Administrator

Jacob Loveless et al.'s article "Online Algorithms in High-Frequency Trading" (Oct. 2013) is an example of potentially valuable research misdirected. Ask any proponent of free-enterprise economics to explain its merits, and you will likely hear two themes: Profit motivates, and profit accrues by producing and selling valuable goods and services. The first buys the producer a bigger piece of the pie; the second increases the total size of the pie, thus raising, at least on average, the economic status of all. It works, most of the time, quite well.

Unfortunately, there are also many ways to profit while producing grossly inadequate, zero, or even negative economic value. Some of us are drawn to such schemes, so much so they work much more diligently at them than at a productive enterprise. To the extent this happens, free enterprise is undermined. Like printing counterfeit money, it works only if a minority does it, and even then, at the expense of everyone else.

Among the most serious such non-value-producing profit schemes is speculating in zero-sum derivative markets that produce no economic value at all, managing only to shuffle cash between winners and losers. Millisecond trading is just an escalation in vying for money this way. Even in financial markets like common stocks, where the original purpose is investment, and that do contribute to producing value, trading at sub-second time intervals is pure speculation or worse, as genuine investors could collectively be net losers to speculators. Putting effort into developing and using more successful speculation strategies is like going to a potluck dinner but bringing no food, just a bigger plate, while pushing more aggressively toward the front of the line.

Online and one-pass algorithm research can surely be redirected toward value-producing applications (such as robotics) where they can do more than just seize profits at someone else's expense.

Rodney M. Bates
Strong City, KS

Displaying 1 comment