9.2 Why a paired test is better than unpaired

Because you always groups things by observation. You assume the difference is due to the observation, but there may be confounders that you simply has not observed. The observation can be randomly perceiving only one factor but ignore other factors. Things are related, that one factor may change other factors therefore affect the subject. To control other factors, the best thing can do is use the same subject and control its environment, giving one factor (that one can perceive as one factor) at a time.