Charles Duhigg wrote a piece for the New York Times that described nothing good with the technology of the practices of HFT. Duhigg wrote, “Powerful computers housed right next to the machines that drive marketplaces like the New York Stock Exchange enable high-frequency traders to transmit millions of orders at lightning speed and, their detractors contend, reap billions at everyone else’s expense. These systems are so fast they can outsmart or outrun other investors, humans and computers alike. After growing in the shadows for years, they are generating lots of talk.”
On August 9, 2010, just three months after the flash crash, the New York Stock Exchange (NYSE) opened its Mahwah, N.J., data center.
“While the change won’t be noticeable to stock traders or investors, it’s a significant moment for the growing market for low-latency trading,” wrote Rich Miller for [the] Data Center Knowledge [Web site]. He continued, “The new data center features colocation space for trading firms seeking high-speed access to the matching engines. NYSE Euronext says it has sold out all the available colocation space in its first phase at Mahwah—reported to be at east two 20,000 square-foot pods.”
Staff reporter for the Wall Street Journal Scott Patterson explained the effect of Mahwah on the world of trading in his 2012 book “Dark Pools,” writing, “The (New York Stock) exchange was all but dead. Mahwah was the new floor, a powerful confluence of capitalism and state-of- the art computer technology. While tourists snapped photos of the exchange’s marble façade on the corner of Wall and Broad, the real trading was taking place 30 miles away in Mahwah’s vast air-conditioned floors of computer servers.”
Why did trading firms need to be co-located with the new hub of the NYSE? Because in the world of algo trading, the speed of light matters. There are few businesses in which the speed of light really matters, but in the world of high frequency trading, pushing up against the speed of light is an integral part of the entire industry. The thing that allows the algos to make so many trades so quickly is that they are hooked directly into the massive servers that now run the world’s trading markets. Like the NYSE’s Mahwah data center, trading no longer takes place as it had for most of the last two centuries. Gone are the days of frantic traders yelling ”Buy!” and ”Sell!” on the floor. Today, all trading happens via computer code in a server. It all happens in a black box.
Like the NYSE, most markets give some trading firms direct fiber optic access into the trading servers. In a fiber optic cable, data (the old ”Buy” and ”Sell” yells of the traders) travel via light. Light can only go so fast (also known as the speed of light). When you want to make a trade faster than your competitor, where you put your server matters, because the speed of light matters.
Let’s pretend you and I are competing to buy a stock on the NYSE at the cheapest price. You and I both send our “Buy” order at the same time. Whoever buys the stock first gets it at a lower price than the person who comes in second, because you and I are buying so much stock that the value of the stock will go up when either of us buys it. Your server is based in Chicago, and my server is cozied up next to the NYSE servers in New Jersey. When we hit “Buy” at the same time, I win, because both of our “Buy” commands travel at the speed of light but my “Buy” command has less distance to cross, so I get there first. I win. You lose. When you’re dealing with the speed of light (also known as latency), distance matters.
Technically what I’ve just told you is wrong, but it’s wrong at a micro level. Rarely does one order sweep out the liquidity—that’s one reason bids and asks are often in lots of 100. What happens is that I get it for a penny less than you do. This is assuming we’re talking generic highly liquid stocks. A hundred pennies means I paid a buck more on that trade. That doesn’t seem like a lot, but if I’m trying to move 100,000 shares that’s a thousand trades and I’m “losing” a buck on each which means it cost me a thousand dollars more to move my shares than I anticipated.