Detailed numerical models of power markets with millions of variables and equations are often perceived as black boxes, because differences in results cannot be traced back to single equations or assumptions, respectively. We unravel parts of those black box by determining the impact of different investment cost specifications including the role of varying discount and interest rates. We further expand our analysis to the impact of simplifications strategies (foresight, spatial resolution, temporal resolution) that are applied to contain numerical feasibility of such models. The choice of investment cost modeling (and related discount and interest rates) has the highest impact on results. Increasing or decreasing, respectively, complexity in turn, has only minor impacts. Our findings questions the current focus of the literature on complexity of power market models neglecting the most relevant factor, which is the choice of handling investment costs.