Pairs trading takes two instruments that usually move together, waits for them to diverge, and buys the laggard while selling the leader, expecting the gap to close. In forex it is applied to correlated pairs such as EUR/USD and GBP/USD, and in indices to related markets.
The claim is that correlations are stable enough for divergences to be mean-reverting.
We tested pairs and correlation-based strategies across 17 years of data with real trading costs, and found no edge. Correlations between currency pairs drifted enough that the divergences the strategy traded were often the start of a new relationship rather than a temporary gap, and the two-sided cost of a pairs trade doubled the spread.
The reason is that retail pairs have no economic mechanism forcing them back together, unlike two share classes of one company. A correlation is an average, not a tether.
If you trade it anyway, test the stability of the correlation itself over rolling windows before trusting any divergence in it.