In the 400-year-or-so history of trading, traders have had to rely on themselves as computer, decision maker, researcher, fact-checker and, oh yeah, trader. Today, automation has increased traders’ productivity, just like it has for cashiers, sandwich artists, car manufacturers and so on, but it hasn’t eliminated the job. Automation, innovation and computerization may eliminate the need to hire a person, but it’s much rarer for a technology to come around that actually eliminates a job.
Quant trading strategies have progressed in much the same way. When quants began to show up in various forms toward the end of the 20^{th} Century, there were claims that they meant the end of trading, the end of markets, even. Traders would soon go the way of the stagecoach drivers and Latin professors.
But ultimately, all the quants and their quanty ways did was change the nature of trading. A quant is just one more type of trader, and quant strategies are just one more tool in the trader’s toolbox – and another potential source of misleading data.
The most successful quant strategy is actually a matter of real estate, not computers. High frequency traders perform trades faster than anyone else, allowing them to shave infinitesimal fractions of profit from a near-infinite number of trades. But the way they get so fast? Having shorter internet cables. No, seriously.
For an example, let’s say you place an order to buy TSLA at market: $600.00. You place the order and your brokerage says, “We bought it for 600.05.” This happens constantly when you place market orders: you get the best available price, which may be a little more or less than market price. But here’s what happened: in the time it took your broker to tell you “We bought it,” a HFT quant saw there was a TSLA order coming in at 600.00, bought it, drove up the price slightly, then turned around and resold it to your broker at 600.05.
The strategy’s simplicity is hard to deny. Just program an algorithm to detect those opportunities and make the trades, then conduct a few million of them, and you’d have a profit to write home about (instantaneously, of course.) Naturally, this simplicity has attracted plenty of competition, not unlike bitcoin mining, leading to an arms race of who can perform trades the fastest.
Statistical arbitrage (Stat Arb for short. Catchy, I know.) is a quant trading strategy that attempts to sus out patterns from stocks based on how they’ve always moved relative to each other in the past. The patterns must be so uncanny, so significant, that it’s as if an unknown factor (a statistician might call it a lurking variable) is influencing both stocks, but not the overall market.
Here’s a simplified example of how it works:
First, find two stocks that tend to move in tandem, like Ford (F) and General Motors (GM). The logic here is that these stocks are constantly fighting over market share, so if one sells more, the other sells less. At the same time, if there’s a major breakthrough in automotive technology, it will help both companies. Up or down, their stocks are inextricably linked.
Next, take a tiny leap of faith: if these stocks have always acted the same, then it stands to reason that if they’re acting different right now, they’re due to start acting similarly again. The tough part is figuring out how they’re going to start acting similarly. Is the one that’s overperforming going to revert or is the one that’s underperforming going to catch up? It’s hard to say.
To perform true statistical arbitrage, you would conduct this kind of analysis on hundreds or even thousands of stocks, alternating between long and short positions on each, determining which have recently moved in statistically significant ways – and which are about to.
If just reading that description sounds like a headache, imagine implementing it. Hence why it’s a quant strategy: conducted by computers on a mass scale. It’s not a strategy that “believes” in any of these stocks, or even the futures of the companies involved. Running on programmed assumptions, it’s simply a game of statistics, looking for one outlier after another.
When a strategy works, especially if it’s used by an entire hedge fund or investment firm, it tends to get noticed and copied. Fast. By definition, these are strategies that can be boiled down to 0s and 1s. They’re quantifiable and computerizable. So, if one firm can implement them, so can anyone with a processor and an internet connection.
Add to that the simple facts that millions of people have a vested interest in getting it right, and countless math PhDs have spent their careers chasing these opportunities – and succeeding. The easy quant solutions are now all but gone.
Fractalerts relies on algorithms, but not quant strategies. This may sound like splitting hairs, but the differences count. Think of quants and algos like two different measurements of a house. A quant approach might tell you the area of the house, while an algo tells you the height. Both numbers are based in the same science – geometry – and help you figure out what you really want to know: can I live in this dang house or not?
Algorithms and quant approaches are based in the same market mechanisms, but neither gives you a perfect, complete picture of the market. (You shouldn’t trust anything that claims to, by the way.) However, if you’re looking to make more sophisticated decisions about trading, going on than just your gut, algorithms can boil down lots of market information into simple, digestible directions.
Links: