You're spot on. Another point I like to raise is that predictors are often extremely redundant in terms of information. The implication being that nearly as good predictions might be made with three predictors vs thousands.