Pairs trading with retail tools
Posted: Wed Oct 28, 2026 4:56 am
Pairs trading takes two related instruments and trades the spread between them, long one and short the other, expecting the relationship to revert. It's a staple of institutional quantitative trading. Retail traders can do versions of it in forex.
An example: EURUSD and GBPUSD tend to move together. When the ratio between them stretches beyond its normal range, buy the laggard and sell the leader, expecting convergence.
The practical hurdles include double spreads and commissions, the instability of correlations, and the risk that the relationship breaks permanently, perhaps because of a Brexit-style event or a central bank divergence. Position sizing is also awkward, since the two legs have different volatilities and pip values.
To implement it properly, the spread series should be calculated using a regression-based hedge ratio, and tested for mean reversion using statistical tools such as the Augmented Dickey-Fuller test or half-life estimation. Many apparent relationships fail these tests.
Costs matter greatly, because the expected profit per trade is small.
A first step is plotting the ratio of the two prices over time and looking at its distribution. If it wanders without returning to a mean, as is common in trending relationships, the method will fail. If it oscillates around a stable level, there's something to work with. Statistical tests on the spread, along with the half-life of reversion, tell how long positions typically take to pay off.
For intraday trading, the half-life matters hugely. A relationship that mean-reverts over several days is useless for a scalper. One that reverts in minutes might be tradeable but runs into cost issues, since the profit per trade is tiny.
Sizing the legs to be roughly volatility-neutral keeps the position from becoming a disguised directional bet.
To see whether a pair of instruments is even a candidate, a sequence of simple checks works well. First, plot the two price series over the last year and look for a stable relationship. Second, compute the correlation of their returns, not their prices. Third, regress one on the other to get a hedge ratio, and plot the residual series. Fourth, check whether the residual crosses zero often, and how long it takes to return after a large deviation.
Many pairs fail the fourth test. The residual wanders for weeks, making the trade a slow, risky bet on a relationship that may have broken. Those that pass often show half-lives of hours or days, not minutes, which doesn't suit scalping.
Costs complete the picture. Two legs mean two spreads, and if the typical deviation worth trading is a few pips, the net edge after costs can be negligible. A spreadsheet with realistic costs deducted from each simulated trade shows quickly whether the idea has any economic meaning.
Has a currency pairs-style strategy been attempted, and did the relationship hold long enough to pay for the costs?
An example: EURUSD and GBPUSD tend to move together. When the ratio between them stretches beyond its normal range, buy the laggard and sell the leader, expecting convergence.
The practical hurdles include double spreads and commissions, the instability of correlations, and the risk that the relationship breaks permanently, perhaps because of a Brexit-style event or a central bank divergence. Position sizing is also awkward, since the two legs have different volatilities and pip values.
To implement it properly, the spread series should be calculated using a regression-based hedge ratio, and tested for mean reversion using statistical tools such as the Augmented Dickey-Fuller test or half-life estimation. Many apparent relationships fail these tests.
Costs matter greatly, because the expected profit per trade is small.
A first step is plotting the ratio of the two prices over time and looking at its distribution. If it wanders without returning to a mean, as is common in trending relationships, the method will fail. If it oscillates around a stable level, there's something to work with. Statistical tests on the spread, along with the half-life of reversion, tell how long positions typically take to pay off.
For intraday trading, the half-life matters hugely. A relationship that mean-reverts over several days is useless for a scalper. One that reverts in minutes might be tradeable but runs into cost issues, since the profit per trade is tiny.
Sizing the legs to be roughly volatility-neutral keeps the position from becoming a disguised directional bet.
To see whether a pair of instruments is even a candidate, a sequence of simple checks works well. First, plot the two price series over the last year and look for a stable relationship. Second, compute the correlation of their returns, not their prices. Third, regress one on the other to get a hedge ratio, and plot the residual series. Fourth, check whether the residual crosses zero often, and how long it takes to return after a large deviation.
Many pairs fail the fourth test. The residual wanders for weeks, making the trade a slow, risky bet on a relationship that may have broken. Those that pass often show half-lives of hours or days, not minutes, which doesn't suit scalping.
Costs complete the picture. Two legs mean two spreads, and if the typical deviation worth trading is a few pips, the net edge after costs can be negligible. A spreadsheet with realistic costs deducted from each simulated trade shows quickly whether the idea has any economic meaning.
Has a currency pairs-style strategy been attempted, and did the relationship hold long enough to pay for the costs?