Pairs trading takes a long position in one asset and a short position in a closely related one, betting that the gap between them will close. It’s the simplest form of statistical arbitrage: one pair, one spread.
How it works
Find two assets that move together — same sector, same underlying driver. Model their normal relationship. When one outperforms the other by more than usual, short the outperformer and long the laggard. Exit when the spread returns to normal.
The trade is indifferent to whether both go up or both go down. It earns only if the gap between them narrows.
Choosing pairs
Correlation isn’t enough — two assets can be correlated and still drift apart permanently. The stronger test is cointegration: whether the spread between them is itself mean-reverting over time. Pairs that pass tend to share a fundamental link, not just a historical one.
Where it fails
When the relationship breaks. Two tokens that tracked each other for a year can decouple in a day if one gets listed on a major exchange, gets hacked, or changes its tokenomics. The strategy keeps adding to the position as the spread widens, because widening is exactly what it’s designed to buy — and the spread never comes back.
Stop-losses on spread trades exist for this reason, and they’re the hardest ones to honour because the model says the trade is getting better.