I’ve long thought on the subject of retreating in battle, and naval battles in particular as well. There are a couple of things that usually come into play in my mind on implementing this or not (focusing on naval combat here). I’ve never actually tried this, but something I’ve certainly thought of as well.
First of all, I generally agree with CWO Marc here. There were a lot of ways to track naval movement and in a way “predict” what another force may do. I’ve also thought that having a dice role to decide if navies find each other would be a good idea. I’ve thought that multiple things can affect this within the game. If we say that on a role of 1-3, as you propose above, the navies miss each other generally, we can have things alter this. Examples being: if navies are adjacent to land, you’d change the odds to a 1-2, showing land reconnaissance as spotters. And if you have a fighter and/or sub or something, you increase the odds even more.
The problem with this is using D6 in AA in my opinion. The odds become a bit difficult to calculate for me with a D6, a D12 would be better, but then you’d be adding a D12 for one component of the game, which could be fine.
The other big issue for me is realizing the time frame of a single turn. I feel like we all sometimes forget one one turn of an AA game can entail. If a single turn is multiple months, then a single move/battle is simulating all kinds of smaller things. Logistics, combat, repair, reconaissance, etc. Point being, I think we begin to venture into the realm of unrealistic to think that naval forces within the same sea area over a multiple month period wouldn’t at some point run into each other and engage in combat. I personally think that implementing a retreat option is a better option here.