What is statistical arbitrage? (And why hedge funds have been doing it since the 1980s)
Arbitrage, classically, is the simultaneous buying and selling of the same thing in two different markets to profit from a price discrepancy. Buy gold in London for £1,000, sell it in New York for £1,010, pocket the difference. Risk-free money.
Real arbitrage barely exists anymore. Markets are too fast, too connected, too full of people with the same idea.
Statistical arbitrage is different. It's not risk-free. It's based on probability. And it's been one of the most reliable systematic strategies in the industry for nearly forty years.
Here's how it actually works.
The basic idea
Some stocks move together. Not perfectly, not always, but reliably over time. They're correlated because the underlying forces that drive them are the same: same customers, same suppliers, same interest rate sensitivity, same macro exposure.
Coca-Cola and PepsiCo. Both consumer staples. Both live and die by consumer spending, commodity prices (corn syrup, aluminium), and sentiment towards fizzy drinks.
Visa and Mastercard. Both payment networks. Both benefit when people spend money. Both get hurt when credit tightens.
Goldman Sachs and JPMorgan. Both banks. Both exposed to credit, rates, and the general health of capital markets.
Over time, these pairs tend to trade at a relatively stable ratio to each other. Not a fixed price, but a stable relationship.
When that relationship stretches — when one stock gets ahead of the other by more than usual — it tends to snap back.
That snap-back is the trade.
What a z-score actually means
The z-score is just a way of saying "how unusual is this, in standard deviation terms?"
If Visa and Mastercard have traded at a ratio of, say, 1.2x over the last year (Visa costs 20% more than Mastercard, historically), and today that ratio is 1.45x, you can calculate how many standard deviations away from normal that is.
A z-score of 2.0 means the gap is two standard deviations from its historical average. For a normally-distributed variable, that happens about 5% of the time.
A z-score of 2.5 means it happens about 1% of the time.
The higher the z-score, the more stretched the pair, and the more likely it is to revert to normal.
In plain English: the further the gap has stretched, the better the odds it closes.
Why the edge exists
Markets are good at pricing individual stocks. Millions of analysts, thousands of funds, all staring at the same companies, all trying to find an edge. Any publicly available information gets priced in almost immediately.
But the relationship between stocks is less efficiently priced. Nobody is specifically paid to maintain the spread between Visa and Mastercard. Index funds buy both. Quant funds might exploit the spread, but only at massive scale where they face their own constraints.
The edge exists because:
- Noise moves individual stocks — earnings misses, analyst upgrades, random Twitter controversy — without changing the fundamental relationship between correlated companies
- Crowding effects at scale — big funds can't quietly exploit a £50k spread. By the time they've built a position worth trading, they've moved the price
- Mean reversion is real — correlated stocks have economic reasons to stay correlated, and those reasons persist
Why it's still working in 2025
The Goldman/Morgan Stanley pairs traders of the 1980s were operating with a clear edge: nobody else was doing it. That edge has compressed over decades. The low-hanging fruit got picked.
But the strategy didn't die. It adapted. Faster execution, more pairs, better statistical models, more sophisticated risk management.
And for retail traders working at small scale, the dynamics are actually quite favourable. The strategy works best when you can slip in and out without moving the market. At £5k-£50k per trade, you're invisible.
The hedge funds haven't killed this. They've changed how you run it. And for small operators, that turns out to be fine.
Next up: why being small might actually be the biggest advantage you have.