giorgio10

correlation is only meaningful when measured over longer intervals. If correlation is high , it doesn’t mean prices always move in lock step over  every interval. There are always  short term  divergence happening . Still we can use correlation to judge early if this is long term divergence ( correction ) or merely short term fluctutaion which will revert to the mean once the event has passed.
The trick is to balance costs of rebalalncing with negative alpha from straying too far from the  design. If you react to every little event , yes you will have a good record tracking the design weights , but cost will be too high . You could use correlation as a guide to recommend when it is desirable and when it is not worthwhile.