What Computer-Powered High Frequency Trading Looks and Sounds Like
With humans replaced by torrents of data, trading has become a strangely beautiful thing.
There was nothing very surprising about the news that the Kansas City Board of Trade will be shuttered this July, as the new owner of the 150-year-old trading floor--the central exchange for red winter wheat, used to make bread--folds it into the Chicago Mercantile Exchange. The once-raucous floor had been dead for years--traders had long since moved to desks with computer terminals.
But the announcement signaled another step into virtualization for the stock market, which has gone from being a real-life place where people shout orders at each other to bundles of high-speed wires and giant servers, a market built on speed and efficiency and ones and zeros. "The thing has come to an end," Morgan Shay, a trader at the exchange since 1971, told Reuters. "The handwriting was on the wall when the electronic [trading] started."
What is the stock market? Simply put, it’s a place where you can go and buy and sell stocks, or equities, which are shares of a company. In financial parlance, the stock market is an exchange. The New York Stock Exchange is an exchange where you can trade stocks listed by the NYSE. Same goes for the NASDAQ, with its own list.
But when we talk exchanges, we’re not just limited to stocks. At the Chicago Mercantile Exchange, you can buy and sell commodities like gold, oil, or corn. The previous exchanges used to be the main ones. Today, we’ve got about 60 exchanges where you can buy and sell things: stocks, commodities, or derivatives (contracts that give you the right to buy or sell stocks and commodities).
If you’ve watched the movie Trading Places, you’ll have a classic idea of what one of these exchanges used to look like. More specifically, you’ve seen what goes down on the trading floor, which is really just a bunch of dudes standing around.
Then, when something happens, everyone freaks out, trying to buy and sell with each other as everyone reacts to whatever new information they’ve just processed.
They’re freaking out because speed is of the essence. One of the few guaranteed formulas for investing--if you can call it that--is to be faster than the next guy. Being faster means you can buy lower and sell higher, either in terms of getting material information or being able to pull the trigger first. As such, we can visualize the stock market in linear, essentially 2D terms by looking at its price and the amount of volume being traded. The price either goes up or down based on supply and demand. Spikes in volume means something big must’ve happened that got everyone's attention.
The top image is what the NYSE trading floor looks like today. It remains symbolic in our minds and hearts, but in reality, it’s been deserted for practical purposes. As it turns out, human contact is both laborious and inefficient. Most pressingly, we’re really slow.
Enter high frequency trading. Where before we might have dealt in seconds, or half- or quarter-seconds, we’ve crossed the Rubicon, leapfrogging even milliseconds into the world of microseconds. That’s one millionth of a second. Today, the trading floor, and all of those specialists--the guys who are obligated to hang around and buy and sell to keep the market humming--with their rows and rows of servers, is not much different than the inner workings of Google or Facebook or Amazon.
So how does the stock market look when you’ve replaced people with algorithmic-based supercomputers? As Radiolab’s Andy Mills puts it, “strangely beautiful.” Forget Trading Places, we’ve turned the stock market into Tron.
A few seconds of high-frequency trading, visualized. Via Nanex.
Beyond the sheer volume--billions upon billions of transactions in a few seconds--the above GIF gives a sense of the complexity of today's trading landscape and the increasing intelligence of bank-programmed algorithms. The antithesis of 'buy and hold,' algorithms will flash buy and sell orders to parse the market or even fool other computers. In one instance, a particularly devious algorithm flooded the market with buy orders, driving the price up of its target stock and quickly sold before anyone, man or machine, realized what had happened. In mere seconds, the instigator made a cool half-mil. This is cyberwarfare for dollars and cents. Or rather, fractions of cents, multiplied many times over.
Nanex, the financial data technology research group that prepared the visualization, allegedly has a database now 20 times the size of NASA’s. That can help to create a microsecond play-by-play of the robo-market, especially helpful, one hopes, in the event of a cataclysmic 'flash crash,' like the one that sunk the market by 1000 points one day in 2010.
Because our ears are capable of processing information much faster than our eyes, Nanex founder Eric Hunsader also programmed a digital keyboard to strike a musical note every time the stock was bought or sold. Now we can hear high frequency trading, albeit slowed down considerably--a Glassian ode to capitalism composed by the Invisible Hand.
The actual music is less modernist sonata than chiptune cacophony. When time is money and with billions of dollars at stake, things are only getting faster. Outside of building smarter algorithms, the main strategy is to decrease physical distance. According to a report from Mother Jones, “every extra foot of fiber-optic cable adds about 1.5 nanoseconds of delay; each additional mile adds 8 microseconds.” That can mean the difference between making and losing millions, and explains the scramble among trading firms to set up shop closest to the exchange computers, which are based in warehouses in the wilds of New Jersey.
The rise in high-frequency trading in the stock market from January 2007 through January 2012. Via Nanex.
That initial land grab was the opening shot in the algo-electronic banking era's new battleground for speed. Recently, a firm named Spread spent hundreds of millions of dollars to build a straighter fiber optic cable between Chicago and New York so that signals that once took 15.5 milliseconds to bounce back and forth now took only 13.3, according to Radiolab. In the world of microseconds, 2.2 milliseconds is an eternity.
Traders now pay up to $300,000 a month to use Spread’s network, with the knowledge that they can make much more than that thanks to that 2 millisecond differential. In 2010, outfits like Spread invested an estimated $2 billion on super-fast infrastructure. Some are turning to even faster microwave antennae to transmit trades between Chicago and New York.
McKay Brothers is connecting the financial districts of New York and Chicago with a network that aims to execute the fastest trades in the country using microwaves.
Now that Chicago is hooked up, traders are looking across the Atlantic where underwater cables priced at $300 million each will shave off another five to six milliseconds by 2014 between New York and the London Stock Exchange. There’s only so many ways you can straighten out a cable and in fact, light travels through air faster than it does a fiber optic cable. One suggestion is a series of towers or even drones that flash data from one to the next. It might sound ridiculous but it makes cents.
So what does all of this mean? We don’t totally know, in part because it’s just too fast. Empirically, it’s been good in the sense that, when machines compete against each other, the cost of humans doing business goes down, in some cases tenfold in the last two decades. That means that if you wanted to sell your Apple stock, AAPL, the fees required are much less now than they were back in the '90s.
Workers installing trans-Atlantic fiber optic cable on British shores (via Daily Mail)
Still, there have been flashes of terror, notably the oft cited flash crash of 2010, when an errant algorithm bled over a trillion dollars out of the market in mere minutes. (Radiolab's fantastic episode, which is about "speed," includes an apocalyptic montage from the headlines on TV and radio that day.) The market would bounce back in the same amount of time, but the mood was set. The robots had taken over and the humans had little idea of what was going on.
It would eventually take the Securities and Exchange Commission three months to analyze what took place over the course of three minutes. Their analysis described "a market so fragmented and fragile that a single large trade could send stocks into a sudden spiral," the Wall Street Journal reported. Their final report was inconclusive, just one more theory among many. Three years later, we can still only guess what really happened.
That is really the central problem with our addiction to speed. It's nothing new. Before the invention of the telegraph--which ushered us into the quarter-second era of communications--Paul Reuter made a name for himself and his media company, Reuters, by famously employing carrier pigeons between Brussels and Paris since at the time, his birds traveled faster than any train. Machines have made the market far cheaper, far bigger, and far more efficient than ever before. But also far faster than we can comprehend, even when we try our best to watch it or listen to it go.