I find that applying the minimum amount of abstraction necessary to reduce excessive duplication and keeping the call graph/class hierarchy relatively "flat" yields the best results. Making the design more table-driven, where most of the complexity/diversity of cases is data and the code is a simple interpreter, instead of directly encoding them in the code, also works well.
“Show me your flowcharts and conceal your tables, and I shall continue to be mystified. Show me your tables, and I won’t usually need your flowcharts; they’ll be obvious.” -- Fred Brooks
