Alex,
Please examine these "AI makes a poor choice" scenarios:
- When receiving a retreat order especially (but it also happens on just about any order requiring a significant reposition): AI is doing this bizarre "long U-turn" behavior that's making them swing far out into enemy territory (often getting destroyed) when they could have easily, immediately, begun retreating in the correct direction/path which was far safer.
- AI is pretty bad about getting trapped against map boundaries because they are waiting too long to move in a better direction.
- especially when executing a retreat-type order- even with fast ships (speed > 100)- ships are very often kissing enemy capital ships on the way by (passing very close, right in front of the main guns) when they could have easily passed by at a safe distance: there weren't encroaching ships that forced the choice of that extremely risky path.
- Generally, AI is very bad about choosing an escape route when executing retreat-type orders: they just "go down" w/o regard for known enemy distribution even when they are visible on sensors.
- AI retreating in good condition (flux/hull/low-no malfunction chance) isn't weighing "obstacle" ships in a sane way: if the escape path is obstructed by ships we could turn on the shield and drive right by no sweat - do that. Questions to answer: "What's their DPS vs my flux cap/dissipation?"; "How likely am I to get flanked considering the speed diff between me and attackers and how much shield coverage I have?"
Honestly, I wouldn't mind micromanaging all this...except as it stands AI is PRETTY LAX about immediately starting to move toward my order, command points are very limited, and the control experience is very unlike an RTS that demands kind of micro. Humbly, I hope that you'd fix these AI behaviors themselves rather than punting to player micro - so that enemy AI can behave in a more smart/slippery way too - which I welcome.