Many motives spur astronomers toward numerical simulations. These computer experiments test well-defined theories, display outcomes of complex interactions, elicit quantitative comparisons with observations, and provoke new insights. Moreover, they are almost always guaranteed to lead to a publishable result. What could be finer and more delightful!?

Streams of simulations have therefore poured forth in abundance. They differ mainly in their assumptions about the amount, nature, and role of dark matter, and in their initial conditions. Most agree with some aspects of observations, but none with all. None, so far, are generally accepted as complete descriptions of galaxy clustering.

As computing power expands, each new generation essentially repeats these simulations with more complicated physical interactions, greater detail, higher resolution, and added parameters. While this development continues, it seems to me wiser not to discuss the latest examples here, for they will soon be as obsolete as their predecessors. Instead, we concentrate on the simplest case: the cosmological many-body problem. Even this reveals a richness of behavior that surpasses current understanding. Understanding is more than simulation, for it embeds the simulations in a much richer conceptual context.