I think this is a confusion between “marginal” and “by and large”.

If I go to a coach to help me with running faster their advice won’t be “completely wrong” — it will be “by and large” correct.

But there are still opportunities for “marginal” improvements that can be found with better measurements. Eg in moneyball they weren’t showing that everything the scouts did was completely wrong. They showed that marginal improvements could give them an economic edge. Even a few percentage points of improvement of received wisdom v measurement meant their money was much better spent.