Mean-reversion signals across Korean equity spreads. Three lenses on the same statistical engine: when does a structurally persistent spread sit far enough from its long-run mean that history says it should snap back?
Each strategy reduces a pair to a single time series — the spread. We then run the same standard treatment on every pair: empirical distribution stats, AR(1) regression for half-life, and an h-step normal-approximation forecast that yields a mean-reversion probability. No machine learning, no hidden parameters: every formula is in the open-source repository.
| Pair | A | B | Spread | z | Pct | MR 60d | Half-life | History |
|---|