Good observations.
They boil down to this:
Computer simulations have two parts: the "simulation" in the sense of the program that runs in the dark of the computer, and the "interface" that represents that simulation and provides input.
If we're talking about simulation, then it's easy to talk about realism. Here's a list:
A. Flight Physics.
B. Weapons Ballistics
C. Environmental Effects
D. Damage Modeling
For some of these, such as how much damage a 20mm shell does to a tailplane, there's room for interpretation, but most of the factors you can put down on a list and compare.
So, for example, while WB fans may think their guns do more "realistic" damage -- and we can't challenge that -- the fact remains that their guns only model 50% of the rounds fired out of them, whereas AH does 1:1.
Another common mistake is to confuse difficulty with realism. For example, how many of us have seen this argument moving from such a premise:
It's harder to land in Sim X than in AH.
Therefore Sim X is more realistic than AH.
Well, if Sim X doesn't model ground effect, but AH does, of course the landings will be rougher.
---
Now to the Interface. The problem with flightsims is that the physics model and the interface often overlap.
Take, for example, control reversal and similar phenomena.
Our controls sitting in a living room are spring loaded and plugged into a computer. If we're lucky, they've got tiny motors in 'em.
Flight Controls are hanging out in the wind. Trim is used to fix them to a position (not necessarily center). At certain extreme conditions, these controls on some planes will tend to move away from the center to the edges, often violently, and with unpleasant results.
So, how do we model trim? Currently it changes the relationship between the virtual stick and the real stick, not the position of the virtual stick.
But this nonsense with the controls moving on their own gets lost entirely. So how do you model a spin where the controls lock in one direction?
You don't.
And there we have a big problem. Take, for example, the DOA Sopwith Camel argument from a few years back. They released a F1 with nasty stall characteristics. A big stink ensued with one party insisting that the Camel stalled something fierce, and the other saying, "No, look at these reports from guys who flew the Camel: if it stalled, just center the controls and it'll pull out immediately."
For us at home, centering the controls means letting go of the stick, or at most overcoming a small amount of force feedback. But it must have been something very different in a crate whirling around in a spin with the controls locked in a corner.
Furthermore, you have the problem of representing a full field of view on a television screen, icons and the rest.
Here we get into further shortcomings of the interface, and the use of analogical means to overcome them. The success and validity of each analogy is open to debate.
Finally, there's the issue of "Realistic" situations, and none of the MMP sims do a good job here.
So to answer your question:
Realism is said in many ways.
In some ways, it's clear which FS is the more realistic.
In others, it's a matter of debate.
Regardless, flying AH is a pretty safe bet.
