Pair Analytics

Cointegration means two listings' log prices historically shared a long-run relationship: a hedge-ratio-weighted combination of them (the spread) tended to revert toward its own mean instead of wandering. The badge reports an Engle-Granger residual ADF test at the 5% level over the stored window.

Z-score measures how far today's spread sits from its trailing 60-session mean, in units of that window's standard deviation. Half-life estimates how many sessions a spread deviation historically took to decay halfway back; it is null when the fitted process was not mean-reverting.

Many tests were run. The scan tests every candidate pair in a sector separately at the 5% level, which for a 40-symbol sector is several hundred simultaneous tests. A 5% test applied that many times is expected to produce dozens of spurious passes — pairs that earn the "cointegrated (5%)" badge purely by chance. The table sorts cointegrated pairs to the top, so those chance passes land where you look first. The exact counts for the scan you are viewing are shown above the table. This scan applies no false-discovery correction, so treat the badge as evidence about one pair rather than as a count of real relationships in the sector.

All of this is historical statistical evidence, not a recommendation. A relationship that held in the stored window can break at any time, and nothing on this page predicts that it will hold. Method version and as-of timestamps are shown beside every result.