My point is that when models predict counterintuitive results (which they often correctly do; See e.g. Eurisco in the Traveller TCS championship, or the shape of the F-117 compared to contemporary stealth aircraft), it's worth double-checking.