a process with "strong deviations" from the long term average is in no way  synonymous with more predictability unless the deviations can be predicted. Until we know one way or the other I suggest your "predictability" be replaced be "potential predictability"