Philosophy of science

Why can a statistical explanation fail to entail its explanandum while still being scientifically useful?