It’s not that confusing actually, it will be easier to understand this rule when you think of this as a clinical researcher. One of the most important reasons is that there is always an overall objective you wish to confirm in the beginning, that will cause the observer-expectancy bias if you know which group is which. That means there will be some confounding factors such as the evaluation on the degree of the disease, the treatment and dosage control, and the knowledge of specific contradictions, that must have been artificially manipulated whether knowingly or unknowingly to distort the outcome in the process.