Using Rolling ADF Tests to Manage Pair Trades

Summary

The document discusses a pairs strategy that recalculates a hedge ratio from a rolling window of prices, forms a spread, and tests it for stationarity with the augmented Dickey–Fuller test. The central question is whether a failed test should force an exit from an open position.

One answer treats stationarity as a condition for initiating and managing the strategy: if the spread no longer passes the test, stop making new pair-rule decisions on the existing trade and use separate loss, trailing-stop, or profit-taking rules. It also suggests looking for another pair for future trades. A second answer clarifies that ADF output is usually a test statistic, where more negative values provide stronger evidence against a unit root; a change toward positive values may prompt reassessment. These responses offer differing interpretations of a negative result, and the post does not specify test thresholds, significance levels, or a validated exit rule.

Key ideas

• A rolling hedge ratio can be used to construct a spread for an ADF stationarity test.
• A failed stationarity test may signal that the pair strategy's assumptions no longer hold.
• An open trade can be managed with separate stop-loss or profit-taking rules when pair signals are no longer trusted.
• ADF test-statistic signs should be interpreted with the test's critical values and setup, rather than as a standalone exit trigger.

Tags

# pair trading - rolling adf test


# pair trading - rolling adf test












I am testing a pair trading strategy. Every day I recalculate the hedge ratio using the past N prices of the 2 underlying. With the hedge ratio, I calculate the past N spreads and do an ADF test to check stationarity. I am wondering how I should treat the case of a negative result. Assuming I was invested in the pair, should I consider that a negative ADF test should trigger an unwind of the position?

## Answer by alexprice (score 1)

https://quant.stackexchange.com/a/43322

For pair trading strategy to work (i.e. to start trade/end trade) the spread between your two underlyings should be stationary all the time. Failing ADF test would mean the spread is not stationary anymore.Assuming you are already invested in the pair by some pair strategy decision:

- you should not make further decisions on the trade by pair-strategy rules

- put simple stop-loss /trailing stop-loss / take profit rules on existing position.

- given that stationary of the spread is violated the best is to find another pair of underlyings for next trades.

## Answer by Ashish Garg (score 0)

https://quant.stackexchange.com/a/43316

ADF test are supposed to have negative values. More negative is the t-statistic the more stationary a time series is. It doesn't indicate unwinding the position.However, if the result start going positive then you may want to rethink your positions.

Shown in full with attribution under the source's licence. Licence: CC BY-SA 4.0 (Stack Exchange)

This summary was written by Stratmill's research agent from the original; it is not a copy of the source.