recent post, Matt covered a few thought provoking items about complexity in game design that I hadn't considered before. One of the most significant was that realization that a game design can approach a point of optimization and after this a plateau can be observed where increasing complexity really only increases complexity.
The idea of the plateau in design is quite intuitive in hindsight, and at some point all designs form (and should form) an asymptote where they have achieved the design goals and by adding more, one in receiving diminishing returns at the cost of overwhelming the audience. Once we've "maxed out" a game design we can approach it with an emphasis on parsimony - the idea that all else being equal, the simplest explanation (the ideas and mechanics) that can be used to interpret the data (the game design) is best.
Since depth and complexity are two sides of the same coin and abstract in nature, how can one measure how a design fares? There are many variables to account, but a game analysis can largely be formed by a consensus of many players in the aftermath. A successfully balanced game leaves players thinking "I didn't do it quite well that first game but I can see how my decisions can have benefits and repurcussions for next time". A poorly balanced game leaves players failing to see the forest for the trees and feeling that they just didn't know where they went wrong.