Latency Tables

The formula for speed has three variables: speed equals distance divided by time. HFT familiarized themselves with the speed at which each brokerage firms’ trades traveled by researching the type of cables each used to send trades. HFT firms also had access to the time the trade took to reach the initial exchange via the receipt. Algorithmically inputting these two values into the formula allowed HFT firms to then solve for the distance from the exchange that the broker sent the order. They then referred to what is known as a latency table (a map of Manhattan containing hundreds of brokerage firm offices) to calculate which specific firm sent in the order on the receipt. 

An illustration of a frustrated broker

For example, if they knew Brokerage Firm XYZ usually bought stocks in quantities of 100 shares and the receipt only showed a purchase of 50 shares, the HFT firm could accurately anticipate that Brokerage Firm XYZ had already sent in an order to buy 50 more. However, the order was still in transit to another exchange. The HFT firm would then algorithmically send in an order to buy 50 shares and, because it invested so heavily in co-location and fiber optic wiring, the HFT firm would get to those fifty shares first. Brokerage Firm XYZ would suddenly receive a notification that its second trade did not go through. The final step in front running is the HFT firm re-offering those same 50 shares at a higher price. Because Brokerage Firm XYZ is obligated to buy the 100 shares requested by its client, it is essentially forced to do so at an increased price. The end result is a higher price paid by Brokerage Firm XYZ and millions of dollars to be made by the HFT firm.

This idea, Lewis emphasizes, is key for HFT firms' ability to anticipate demand and front-run by studying the historical amounts at which brokerage firms tended to buy stocks.