Did the AI forecasts hold up? Every row only scores predictions old enough that the real outcome is already known - a baseline is always shown alongside the learned model, and a bad result is never hidden.