Since I've been lighting up this thread talking about sniper bomber gunners, I decided to do some quick research.
Summary:
1) Bomber gunners tended to discourage or damage their targets and get them to break off much more than they actually shot down their attackers. This means that AI fighters might need to be more timid about attacking "boxes" of enemy bombers and more prone to break of an attack if they get hit by defensive fire.
2) Bomber defensive fire depends heavily on keeping a tight formation, which is related to pilot skill. This means that distance between planes, and ability to keep formation when attacked, is dependent on AI skill.
3) Turret gunners claim more kills than hand-held gunners, but this is probably due to arc of fire, additional guns, etc. rather than any inherent superiority of turrets over hand-held guns.
4) Bomber gunners (at least for the USAAF) overclaimed kills and damaged A/C by an order of magnitude.
5) Determining "ace" gunners is difficult due to multiple gunners firing at the same target, and the fact that records for kills by gunners weren't a carefully kept as those for fighter pilots. It's a bitch getting kill claims for dive bombers and attack aircraft. It's much easier to get data for the USAAF heavy and medium bombers, so my data is biased that way.
Data Points (to be taken with several pounds of salt).
Losses for RAF fighters between the dates 10 July-11 Aug, 1940
Destroyed
by cause;
BF 109, 87
Bf 110, 6
Bomber, 13
Collision, 4
Flak, 1
Friendly, 1
Unknown, 3
Accidents, 47
Grand Total, 162
This means 8% destroyed due to bombers, plus an unknown number that crashed on landing (accidents) as a result of damage from bombers.
Damaged
by cause;
Bf 109, 52
Bf 110, 10
Bomber, 38
Collision, 1
Flak, 1
Unknown, 1
Accidents, 68
Grand Total 106
This means 35% damaged due to bombers, plus an unknown number damaged due to a combination of bomber gunnery and accident. (Sadly, the data I got doesn't state degree of damage, so "damage" could mean anything from Category I - easily repaired, to Category IV - write-off.)
Top scoring US gunner in the ETO claimed 12 kills, 2nd highest had 7. Compare this to 34.5 for the top scoring fighter ace (Francis Gabrelski) in roughly the same number of missions.
8th AF gunners overclaimed by a factor of 10, fighter pilots by a factor of 2-3.
A list of known claims by individual gunners during WW2 and the Korean war (undifferentiated by Medium or Heavy bomber or by type):
74 upper turret 29% 30%
53 ball 21% 22%
46 waist 18% 19%
62 tail 24% 25%
12 cfc 5% n/a
13 nose 5% 5%
5 radio 2% 2%
256 total claims
244 total claims eliminating CFC position (only found on the B29, and reflecting some Korean war kills). The first list of percentages reflects all kills, the second column reflects kills after removing CFC claims. The total number of tail gunner kills should probably be reduced by 1% or so to reflect claims from the Korean war.
Waist gun claims are combined for L and R Waist positions, so total percentages should be halved to relfect each gun station. That is, 9% L Waist, 9% right waist.
The data showed that the nose and belly turrets slightly were more effective against Japanese fighters than German. This is because Japanese tended to atack more from the front and below (due to lack of radar vectoring, speed and high altitude performance).
My gut instinct is that if you factor in the number of guns, fields of fire, and available ammunition, is that, on a per-gun basis and figuring in firing arcs, each gun has about the same chance to hit.
Nothing to prove or disprove my contention that firing while the plane is rolling, pitching or yawing, or shooting to the aircraft's side (i.e., parallel to the slipstream) degrades accuracy. Given the fact that Oleg and Company generally did their homework when evaluating aircraft performance, I'd be inclined to not mess with the gunner AI that much. Certainly if you attack from 6 o'clock level you deserve everything you get.