Statistical arbitrage
Statistical arbitrage
(also stat arb) /stəˈtɪstɪkəl ˈɑːbɪtrɪʤ/ (n.)
A class of quantitative trading strategy short-term trading strategies that seek out temporary discrepancies in market prices of similar instruments, and bet upon them converging on the presumption of mean reversion and the basic soundness, over time, of the efficient markets hypothesis, supported by mathematical brainboxery, overwhelming computational horsepower, and lighting fast, and cheap, access to trading platforms.
|
Prime Brokerage Anatomy™
|
When it works, it works a charm: the top ranking stat arb hedge funds make extraordinary returns, especially when they drive hard bargains with their prime brokers to secure cheap, committed funding. When it doesn’t work it can be a disaster, as the Nobel laureates and masters of the universe who founded Long Term Capital Management, and then promptly blew it up, after the unseemly short term of four-and-a-half years.
Say you have spotted that because the most recently-issued (“on-the-run”) 10-year US Treasuries are used as a reference point for various benchmarks, and tend to be in hot demand for hedging and whatnot, while “off-the-run” Treasury bonds issued a couple of months ago — that were on-the-run until these fresh ones were issued — are in less demand and therefore trade at a bit of a discount. Presuming the unimpeachable full faith and credit of the United States Government, you know two things with relative certainty: One that the on the run bonds will, in short order, become off-the-run ones, and suffer the same sudden, jarring lack of popularity, and two, that both will, in a decade’s time, pay exactly the same amount of money.[1] Therefore selling (going “short”) the on-the-run bond, and buying (going “long”) the off-the-run bond, will capture that arbitrage in fairly, er, short order. This trade was a favourite of the most infamous stat arb firm in history, Short-Term Capital Demolition.
Another one is to pick a bunch of securities in the same sector, that have basically the same attributes, and making a quantitative decision among them which is overvalued — likely to underperform — and which is undervalued — likely to outperform, and shorting the former and buying the latter. Assuming the sector is going gangbusters, your performance will track it, but should improve on it, as this arbitrage works itself out.
These approaches in some ways feel like different things: at first we thought one was convergence and one was divergence, but really they are both about convergence, or mean reversion, to where stocks should trade, from where they are trading: they seek to capitalise on structural mispricing driven by arbitrary, but persistent market factors.
Of course the presence of arbitrageurs should reduce the opportunity, so these days if they open up, they close pretty quickly, so the quest for stat arb funds is to find and capitalise on these micro-anomalies as fast as they can. For that they need computer horsepower and fast connectivity to the trading venues where these anomalies play out.
This also means if you thought you might sign up to Robinhood and make a bit of easy money, you are going to have a disappointing time. Signing up to Robinhood can be fun, of course, as Redditors will tell you.