Why would you ever balance your test data? If 80/20 is the actual population distribution, the sample that forms your test set should conform to that. Balance all you want in train/validation sets, but never the test set.
Not balancing and using ROC is a terrible combo, but the metric is the problem, not the lack of artificial balance.
Not balancing and using ROC is a terrible combo, but the metric is the problem, not the lack of artificial balance.