Moneyball

Everyone treats this as the film where data won. To me it is a film about institutional resistance: the model was right and it was still losing in the roster meeting.

That pattern is familiar in software. The measurement can be correct and the chart you show can be unarguable, and the decision still comes out along the seniority order in the room. What Beane does is not better analysis but moving the decision somewhere nobody can undo it after the fact — he trades the players and ends the argument.

Architecture decisions need the same move: writing the document is not enough, you have to take a step that is expensive to reverse.