Trading @ Lightspeed

But finally, the $300 million data pipe is sold at $10 million per cable pair, and Wall Street thinks it’s cheap at the price because it shaves a few milliseconds off the signal turnaround time between the Chicago Mercantile Exchange and Wall Street. Traders who own a bit of that fibre can pick up market data, interpret it a million times faster than human traders, and take and relinquish positions across widely dispersed trading floors quicker than the competition. They take advantage of price fluctuations across bourses and categories and make a killing from the arbitrage between cash and futures.

This is the essence of high frequency trading (HFT), a subset of “algo” trading, in which machines take the calls on the basis of algorithms, human-written but running without human oversight. Once upon a time in the late 20th century, US investors used to make their biggest margins from large trades placed through the executives of financial service companies like Morgan Stanley. But after Wall Street discredited itself in 2008, clients spread risk by dividing trades into small parcels handled by different agencies. As margins fell, agencies tried to make up with a time advantage, reacting to the market faster than human reflexes. They would have been trading at the speed of light, if they weren’t slowed a few milliseconds by the switches and amplifiers on the route.

Market professionals tend to downplay HFT, perhaps fearing that the sheer impersonality of the thing would unnerve retail investors, but it is surmised that half of all trades in developed markets run on it. It’s picked up in India, too, with stock exchanges offering co-location of members’ servers on their premises. It’s the same logic as the fibre optic cutting through the Alleghenies —  shortening the route. But perhaps it isn’t half as lucrative in India because of transaction taxes, smaller volumes and barriers between exchanges and asset classes.

But one day, growth will make HFT as profitable here as it has been in the developed markets. Earlier this year, an international conference organised by the Securities and Exchange Board of India (SEBI) addressed issues that would perplex most Indian traders —  server co-location, algo trading and HFT. Perhaps we should take this as a sign — algo trading may already account for a chunk of market activity, and some basic reforms can enable HFT.

In the US, Lewis writes, HFT was midwifed by the Regulation National Market System of 2005, which changed the duty of brokers from the fuzzy notion of “best execution” —  whose meaningcould be negotiated between broker and client —  to the absolute of best price. Prices were compiled from all markets into the National Best Bid and Offer database, bids and offers were matched by exchange servers, and brokers crossed the spread (the range between bid and offer) to close deals. The trouble was that the standards did not specify the minimum speed of this process. HFT firms stepped into the breach, setting up parallel hardware and transmission lines much faster than the exchanges’, producing information asymmetry. The official market ticker lags behind real time, the TV news ticker is even laggier, and traditional and small investors see faded, dated snapshots of the market milliseconds ago.

Lewis sees this as legalised cheating which helps institutional players to prosper at the expense of the margins of mutual funds, retail investors and pensioners. Here’s an example of HFT at its simplest: a client asks a broker to buy 1,000 shares of a stock below, say, Rs 300. He finds over 1000 shares offered on Exchange I at Rs 290 but on Exchange 2, 30 shares are offered at Rs 280. The human reaction would be to mop up 30 shares first from Exchange 2, and then 2,970 shares from Exchange 1. But the moment the broker moves on the first 30 shares, HFT machines sense a demand and bid for or even buy the 2,970, jacking up the stock. It’s not wholly unlike front-running, which is illegal.

HFT causes concerns about rogue algorithms and flash crashes like that of 2010, but Lewis addresses deeper ethical issues. He draws attention to ‘dark pools’ (Credit Suisse runs the biggest, he says), trading zones segregated from the traditional market to which HFT traders preferentially send their orders. He finds it odd that the orders generally find best prices within those pools, whose data is not wholly transparent — they flash rates too fast for a human to catch. He writes of ‘proximity services’ like server co-location on exchange premises, and of a specialist who tried to trademark proximity and sell it in a market hopped up on speed. “I felt like the getaway driver,” he said.

Indian markets are placid in comparison, but growth and its needs will step up the tempo. High frequency traders argue, for instance, that every bid represents liquidity, which sounds attractive in a fast-growing economy. And thus, along with achhe din, HFT could soon come to an exchange near you.