I voted get rid of it, because there's no 'drastic changes' option.
I could live with AI gunning if the following changes were made:
Vision and situational awareness
1. AI gunner is only aware of enemies that he can see through the sights (i.e. not behind cover and not off to the side of where he is currently aiming).
2. AI gunner is only aware of enemies that are behind cover if he saw them move into cover.
3. AI gunner constantly scans his sights left and right to pick up targets unless directed onto a target by the player-in-tank-commander-seat.
The AI should be no more able to locate and engage targets than a human player would be in that seat. I can't see targets unless they're visible in my sights, so why should my AI gunner be able to see targets on the far left when his gun is pointing to the far right?
Ammo expenditure and targeting
1. AI gunner fires MG in bursts, not in single shot precision-mode.
2. The AI gunner targets the centre of visible mass of the target.
This will be more believable for the human getting shot. Getting 1-shot headshotted by the AI MG nearly every time is just ridiculous. It's also how a human would tend to play the role. A human player would tend to go for the centre of visible mass of the target and would fire bursts, so you'd get mostly torso hits with some limbs and a few headshots and you'd see higher ammo consumption with more frequent reloads.
Accuracy
1. AI gunner's accuracy should be limited to a level roughly linked to the ability of the human player in the tank.
Whatever variables control the ability of the AI to hit a target should be tweaked so that in practice the AI roughly achieves a hit percentage across all ranges that is determined by 3 factors. Those factors would be:
When using hull MG or coax MG:
1 - the Machine-Gunner class level the human player in the tank has achieved (e.g. if it was me, I'm a Veteran Machine-Gunner, so L4).
2 - the DP-28 / MG-34 level the human player in the tank has achieved (e.g. if it was me, in a German tank, it'd be L13; if I was in a Russian tank, L40).
3 - the average accuracy human machine gunners at that level have achieved using the MG found in the tank (e.g. if it's a Russian tank and I'm playing, it'd be the average accuracy of all human L4 machine gunners who have achieved L40 on the DP-28).
When using main gun rounds:
1 - the Rifleman class level the human player in the tank has achieved (e.g. if it was me, I'm a Recruit Rifleman, so L1).
2 - the MN91/30 / KAR98 level the human player in the tank has achieved (e.g. if it was me, in a German tank, it'd be L1; if I was in a Russian tank, L3).
3 - the average accuracy human riflemen at that level have achieved using the MG found in the tank (e.g. if it's a Russian tank and I'm playing, it'd be the average accuracy of all human L1 riflemen who have achieved L3 on the MN91/30).
I don't know what the average accuracy for each class- and weapon-level combination is, but I'd guess it's probably around 20% at Veteran (L4) with an L40 DP-28 and probably around 70% at Recruit (L1) with an L3 MN91/30. Should be possible for TWI to run a database query on the player stats to find out the real average figures though.
There's several advantages of this:
* AI gunner performance will be believable compared to human performance.
* Using average of all players at the same class/weapon level instead of the individual players prevents gaming the system - otherwise you could ensure your AI gunner got 100% accuracy by taking your DP-28, firing a single shot at a sure target to get 100% accuracy, and then never using it again.
* There is an incentive for people to use and level up some of the less popular weapons - the LMGs and bolt-action rifles - in order to improve their tank AI gunner's performance.