Market-neutral z-score strategies on three cointegrated forex pairs. You don’t need to predict direction — you just need two related instruments to return to their normal relationship.
The classic analogy for cointegration: imagine a drunk person walking their dog. Both wander randomly, but they’re connected by a leash. The drunk might stumble left while the dog runs right, but they never get too far apart. Eventually, the leash pulls them back together.
That’s cointegration. Two assets might each move unpredictably on their own, but the spread between them stays stable over time. When it gets too wide, it snaps back. This is fundamentally different from correlation — correlated assets move together moment-to-moment, while cointegrated assets maintain a long-run equilibrium even when they diverge temporarily.
Take EUR/USD and USD/CHF. These two pairs are naturally linked — when the dollar strengthens, EUR/USD falls and USD/CHF rises. They move in opposite directions most of the time because both are primarily expressions of dollar strength.
Occasionally, one moves further than the other. Maybe EUR/USD drops sharply on European political news but USD/CHF barely budges because Switzerland isn’t directly affected. That gap is the opportunity. We go long on the pair that’s “too cheap” and short the one that’s “too expensive,” then wait for the spread to normalise.
The beauty of pairs trading is that you’re market-neutral. You don’t care if the dollar goes up or down — you’re only betting that two related instruments will return to their normal relationship. This makes the strategy resilient to broad market moves, news events, and central bank surprises.
Because you hold one long and one short position simultaneously, most systematic risk cancels out. What remains is the spread risk — the bet that the spread will mean-revert. This is a fundamentally different (and often lower) risk profile than directional trading.
We compute the z-score of the spread between the two pairs in real time. The z-score measures how many standard deviations the current spread is from its rolling mean:
|z| < 1.0 — Normal spread. No trade.
|z| > 2.0 — Spread is significantly stretched. Entry signal.
z returns to 0 — Spread has normalised. Exit signal.
The classic “dollar mirror” pair. Both are primarily dollar pairs with strong inverse relationship. The Swiss franc’s safe-haven status occasionally creates divergences that reliably revert.
Two commodity currencies from neighbouring economies with deeply intertwined trade relationships. Divergences are typically driven by country-specific news that doesn’t change the fundamental economic linkage.
Two European currencies with a shared economic zone (pre- and post-Brexit, the UK and EU remain deeply connected). Brexit-related divergences create trading opportunities that historical patterns suggest will normalise.
Each pair was selected because they show strong, stable cointegration over long periods. We continuously monitor the cointegration relationship and would remove a pair if the statistical linkage deteriorated.
Unlike directional strategies, pairs trading is largely regime-independent. Our GateKeeper regime classifier does not adjust Engine 3b — because market-neutral strategies don’t depend on whether the market is trending, ranging, or in crisis. The spread between cointegrated pairs tends to mean-revert regardless of the macro environment.
Pairs trading was first formalised by Nunzio Tartaglia’s quantitative group at Morgan Stanley in the 1980s. The statistical framework of cointegration was developed by Engle & Granger (1987), who received the Nobel Prize in Economics for this work. Vidyamurthy (2004) provides the standard reference for implementing cointegration-based pairs trading strategies.
For more on how regime detection interacts with strategy selection, see our Regime-Based Trading documentation.