PacWar and WitP (the latter is the modernized/expanded/remake version of the former) are seen as highly realistic because they have a very high granularity when it comes to detail, and that granularity attempts to simulate actual naval battles and ship design and so on. So instead of an arbitrary mathematical formula to decide how a battle goes (like Paradox does as their battle algorithm has nothing to do with real battles) there are dozens of calculations on whether a shell X fired from cannon Y from a ship Z can penetrate the armour A on a ship B at a location C. Very few strategy games do that as it's cumbersome and requires massive amount of data, which is why it's mostly Steel Panthers, (old) Close Combats and tank/flight simulators which bother with it.
Of course, Hearts of Iron/Rule the Waves is more realistic than PacWar/WitP when it comes to the chain of command, or actual strategic level leadership, because the player cannot micro-manage their ships and task forces but rather has to issue overall direction.
What sloanjh suggests is actually a pretty decent compromise - it is bit jarring that research costs are always the same from game to game and that I can know down to the day when a particular bit of tech will become available, and I can fine-tune my lab assignments on the fly. I wouldn't want a total crapshoot either though, because having no power to mitigate obstacles and problems is just the game showing me the finger - and don't say "well build more research labs" because while that is a valid approach, it's also a really slow process because of the large disparity between strategic direction (industry, research, ship design/construction) that takes several months or years, and the tactical direction (survey, combat) that takes hours or days.