-STL-, fcastro is talking about the odd behavior that allows Sensei to complete Toil Ferry
in the manner detailed in my guide.
After doing some testing, the vague and uncertain explanation that I gave in the guide appear to be wrong. At that time of its writing, I still wasn't quite sure why this would happen, but I believe I've more or less figured out the reason now. So, it's not quite just sending an extra APC/transport copter and infantry and "praying," though a small amount of luck may be involved. It's a special case since it was only a medium tank defending with two high-priority properties under duress. The same result could be reproduced with a neotank, anti-air, or bomber as the sole HQ defender depending on the CO - especially when the AI is not currently using a (Super) CO Power.
First off, a few restrictions to set up possible scenarios:
- The player's CO should not be Sami, and both the HQ raider and decoy should be at 10HP.
- The enemy property that the decoy must be situated on and/or capturing must be a deployment property (base, port, or airport).
- The HQ raider should almost definitely be an infantry while the decoy can be a mech or infantry.
- Of course, the sole enemy defender (medium tank, neotank, anti-air, or bomber) should be able to reach both the decoy and HQ raider.
The main flaws appear to be that priority of defense of deployment properties doesn't seem to be appropriately differentiated enough from the priority of HQ defense and that scoring a potential "crippling blow" (71% or higher against 10HP) on the deployment property can confuse those defense priorities if a (reasonably highly probable) crippling blow cannot be achieved on the HQ. When a CO Power isn't currently being used that day, medium tanks, neotanks, anti-air units, and bombers all pretty much fall into this scenario. If no crippling blows are involved, the defense priorities are not confused.
This behavior is rather reliably reproduced using a custom-made Design Map, and the enemy properties and enemy defender can actually be from different AI armies. (Strikingly, the decoy deployment property can belong to one army or even neutral with the HQ and defender under duress from another.) As long as this sweet spot is found, the desired behavior of the AI attacking the decoy instead of the HQ raider often occurs - e.g. an enemy tank controlled by a Max AI using Max Force/Blast as the HQ defender.