WTF is HFT? Exploring the cowboy frontiers of algorithmic finance

In the time it takes you to read this sentence, a high-frequency trading (HFT) algorithm, connected to a stock exchange via “low latency” trading infrastructure, could make, perhaps, 1,000 trades.

I say ‘perhaps’, because it really depends on how long you pause on those commas I put in the sentence. If you’re an individual with great respect for commas you might give the algorithm a chance to throw in a few hundred more orders.

Let’s just clarify this. That means computers owned (or leased) by a firm somewhere can 1) suck in data from a stock exchange, 2) process it through a coded step-by-step rule system (algorithm) to make a decision about whether to trade or not, 3) send a message back to the exchange with an order for shares of ownership in a company – for example, a company that makes children’s toys – 4) get the order executed and confirmed, and 5) repeat this maybe 250 times a second.

Well, it could be more or less than that, too, and to be honest, few people seem to actually know how fast these algorithmic engines trade. But even if it’s only trading 50 times a second, or even a mere 10 times a second, it’s still inhumanly fast.

Having worked in financial trading markets – albeit in much slower over-the-counter swaps markets – and having worked on a variety of advocacy campaigns related to financial trading, this is a subject that fascinates me. The purpose of this piece, though, is not necessarily to convince you on whether or not HFT is a good or bad thing. Rather, it is to provide some frames through which to look at the phenomenon, and through which to understand the debates and news stories that will undoubtedly continue to be written about it in the years ahead.