Energy projections, computed with energy system models, for decades have performed very poorly, when estimating future trends and uncertainties. In this presentation, I will investigate the origins of these inaccuracies and propose strategies to improve the realism of energy system models and projections.