Mitsubishi Stuka and Thunderbolt

[JtD]

Originally Posted by Pursuivant

One possible fix for TD is to make Rookie gunners basically useless and Average gunners only slightly useful. Only Veteran or Ace gunners should be anything like a threat.

In solo flight, I attacked a formation of two flights of four rookie G4M each from dead 6, 10 times, in an F6F-3. I tried to maintain a position at their 6 at a distance between 500 and 700 m. I shot down 20, and got hit exactly 8 times, 3 times one flight, once in five flights, not at all in four flights. One of the hits was a 20mm hit, the rest came from the light machine guns. I always returned to base no problem.

To sum it up, the gunners were basically useless.
---
Got curious about how gunners will do when faced with proper tactics - so I engaged the same two flights in head on passes and shot down all eight of them without getting hit at all. Will probably not always work like that, but four kills for every hit received in return appears conservative.

I don't know what's "too good" about that abysmal performance.

[Pursuivant]

Originally Posted by JtD

In solo flight, I attacked a formation of two flights of four rookie G4M each from dead 6, 10 times, in an F6F-3. I tried to maintain a position at their 6 at a distance between 500 and 700 m. I shot down 20, and got hit exactly 8 times, 3 times one flight, once in five flights, not at all in four flights. One of the hits was a 20mm hit, the rest came from the light machine guns. I always returned to base no problem.

But, on the other hand, you were hanging out at 500-700 meters rather than getting any closer.

While I don't know what the Japanese gunnery doctrine was, or what they considered the effective range for machine guns/20 mm cannons to be for aerial gunnery, consider that USAAF doctrine held that 0.50 caliber MG were basically useless beyond 600 yards (~550 m).

So, that actually means that the "rookie" AI might have been pretty good.

A slightly more realistic scenario to test gunnery skill would be to fly a typical "pursuit curve" coming in from 2-4 or 8-10 o'clock high initially, making a diving turn as the bomber flies straight and level to attack from 5-7 o'clock level or high, and then continuing the dive beneath the bomber to the other side. On the the exit, drop at least 300 meters below the bomber, gain speed to get ahead of it bomber, zoom climb to gain altitude and repeat the attack. That would test the AI gunners' ability to hit an attacker traveling in two three dimensions and at very different ranges.

But, to test your theory about effectiveness of rookie AI gunnery, I set up a QMB mission - Midway map, 5000 meters altitude, no advantage, no situation, 2 flights of 4 Rookie G4M1 vs. 1 Ace F6F-3. 10 different flights. Autopilot for the F6F.

In sortie, the Hellcat got lucky and/or used good tactics and got 5 Betties before he broke off the attack. That was the exception, however.

In the other 9 sorties, despite jinking around as he bored in from the stern (obviously, in IL2 AI fixed gunnery school, they don't teach pilots beam or head-on attacks, nor diving attacks, against bombers), he ate a significant amount of lead while shooting down a maximum of 1 Betty per mission.

In 2 sorties he took enough damage to the engine or fuel tanks that his engine stopped in the air. In one, he burst into flames. In the other 6 sorties, he suffered enough engine or wing damage and fuel leaks to the point where he couldn't keep up with the bomber formation or felt the need to RTB.

NOT a realistic outcome, to say the least, if only due to the amazingly Stoopid Hellcat tactics.

And, while the rookie Japanese gunners weren't exactly wielding Radar-Guided Blaster Cannons of Death like their Ace counterparts, their gunnery was quite effective even at ranges well beyond 150 meters.

I hate to say it, since TD has done so much to improve AI behavior, but at least in this case the gunners ARE still too accurate, and fighter tactics vs. bombers, even for Ace AI, would get you washed out of fighter training.

[Pursuivant]

Just for fun, I did a variation on the scenario above.

QMB Midway map, 5000 meters, no advantage, no situation, 1 Ace G4M1 vs. 1 Ace F6F-3. Autopilot for the F6F, 10 sorties.

In EVERY sortie, the Hellcat did the same damned thing. 1) Fly past the Betty with about 300 meters of altitude and 1,000 meters of separation. 2) Split S about 1,500 meters behind the Betty. 3) Chase the Betty to within ~200 meters, SOMETIMES NOT EVEN SHOOTING!

If the Hellcat didn't shoot on its first pass, it then:

4) Climbed directly over the Betty to about 1,000 meters above. 5) Slowed down until it was about 1,500 meters behind the Betty. 6) Repeat steps 3 & 4 until Betty turns left (always left), watching the tracers fly past as you crawl into range. 7) Once the Betty finally turns, make a deflection shot at 20-45 degrees angle off.

Meanwhile, the Betty consistently opens fire just inside 1,200 meters (WAY beyond effective range), although it didn't start hitting consistently until 500 meters.

Not surprisingly, given its pathetic tactics, The Hellcat won this encounter just 50% of the time. 3 times, the Hellcat pilot was killed or his plane was set on fire. The other 2 times his engine was badly enough damaged that he RTB. In one case, the Hellcat got PK AND flamed in one shot from 300 meters while the Betty was in a 20 degree bank! There was obviously a serious muscle man hefting that 50 kilo cannon!

I think that the Ace gunnery is about right in accuracy in this situation, except for starting to fire way too soon and the lucky shot while the Betty was in a turn, but the fighter AI is just tragic.

[JtD]

Right, I guess 8 kills vs. 0 hits were too much of an argument to deal with, had to be avoided by starting some new "ace fighter AI is porked" red herring sideline. I take it the point now is that AI gunners mustn't hit anything even when faced with the most idiotic opposition. *facepalm* Nothing else to say, really.

Just out of curiosity - have you ever tried something bigger, like say 16 vs. 32 or something? Or is 1 vs. X the upper limit?

[horseback]

Originally Posted by JtD

Right, I guess 8 kills vs. 0 hits were too much of an argument to deal with, had to be avoided by starting some new "ace fighter AI is porked" red herring sideline. I take it the point now is that AI gunners mustn't hit anything even when faced with the most idiotic opposition. *facepalm* Nothing else to say, really.

Just out of curiosity - have you ever tried something bigger, like say 16 vs. 32 or something? Or is 1 vs. X the upper limit?

My reactions are a little different; at 500-700m, no human gunner behind a real machine gun or cannon could possibly have hit you on purpose from a moving aircraft. If you had been flying a real Hellcat against a real Betty formation and been hit at all from those ranges, you would have turned around immediately to inform your Task Force command that the Japanese have apparently developed some sort of automated fire control or aiming system that is accurate out to nearly a half mile, because more than one stray hit at those ranges would not be a coincidence.

The whole point of the ai is to mimic the capabilities and the historical behaviors of human beings in those aircraft, not to exceed them by several orders of magnitude. This is particularly true for the offliner who wants to re-create historical missions.

It is outrageous for the Rookie ai gunners to be able to hit a single Hellcat going the opposite direction from 780 meters while their aircraft is in a banking turn. It's just insulting when that hit results in the loss of rudder control in the bargain, which is what happened to me a couple of weeks ago.

If you want some sort of 'gameplay' difficulty (more Zombies!), then make the Veteran and Ace gunners your 'high difficulty' or online default, and make Rookies the single aircraft realism standard and 'Average' the 'formation multiplier' standard for offline historical players. And separate the pilots from the gunners, if you want to maintain some sort of bombing accuracy.

cheers

horseback

[JtD]

Funny, considering the several thousand rounds spent, I would say a hit once in a while is statistically prone to happen. After all, firing solutions don't get a lot easier than with a dead 6 approach.

Can you prove it is theoretically impossible to hit an aircraft from 500-700 m away? Otherwise, the chance for a hit is always there. That's just life.

Anyway, if I had killed twenty Betties within just 10 missions in 1 vs. 8 situations, for 8 hits into my aircraft in return, folks would probably want to know my secret, not the secret of some loser gunners. I'd certainly not run to my superiors and whine, after all, I'd be the highest ranking US bomber destroyer ace.

[horseback]

Originally Posted by JtD

Funny, considering the several thousand rounds spent, I would say a hit once in a while is statistically prone to happen. After all, firing solutions don't get a lot easier than with a dead 6 approach.

There are none so blind as those who will not see. I can only assume that you've never fired an automatic weapon of any kind, much less an actual machine gun at a target even 100m away. 'Once in a while' is not three times in one flight; I would guess that considerably less than a thousand rounds per plane was fired at you during that flight, and that you were rarely if ever actually in a truly 'steady state' range/angle/angle from any of those gun positions.

Can you prove it is theoretically impossible to hit an aircraft from 500-700 m away? Otherwise, the chance for a hit is always there. That's just life.

Theory involves statistical probabilities; here the probabilities are not merely being violated--they are being gangraped.

If you are at a 'dead six' to any one of a formation of 4 Betties, there are at least 7 guns that can always be brought to bear on you at any time (the 4 tail gunners and the three top rear gunners of the aircraft you are not directly behind), there is one gun that can usually be brought to bear on you (the top rear gun of the aircraft you are most directly behind) and finally, there are maybe three guns that can sometimes be brought to bear on you (the three side guns of the flanking aircraft in the formation). At best, only one of those guns has an actual 'zero deflection' shot at you. The rest have to calculate some degree of deflection.

In my experience, the ai guns usually fire about three rounds every burst fired, and at ranges outside their 'guaranteed kill' ranges, fire random un-aimed bursts about 2/3rds of the time; the rest of the time, if you continue in precisely the same vector or curve you are going when the shots are fired, you are likely to be hit. Since it is nearly impossible for a human to 'fly' with that sort of consistency, you usually move a bit higher or lower or off to the side of where the burst is aimed, and as a function of distance from the point the shots are fired, the size of the miss is determined by how far you can move off-vector or curve between the time the burst is fired and when the rounds arrive at the targeted point.

Are you suggesting that you were the recipient of over 300 total bursts of ai fire from each aircraft in each mission? The 'chance' of a random hit is always there, but if you get hit more than once, it isn't chance.

In short, the closer you get and the less you react to or anticipate the gunners's fire (I don't recall seeing muzzle flashes 99% of the time, but I don't really look for it) the more likely you are to be hit and the difficulty of the angle and speed don't seem to me to be very great factors, even for 'Rookie' gunners once you come within about 200m. That is not quite a 'spray and pray' solution; once you enter their firing cones, regardless of speed or angle, you will be hit, period, very badly hit if more than one gun is involved.

That's not life. It's not even a reasonable imitation. From my own experiences, using high angle, high speed, high deflection attacks, I should hardly ever be hit at all. Instead, I get shredded more than half the time, and much less often take 'only a flesh wound.' I have also come to believe that what you are flying has more to do with your likelihood of taking a hit than how fast or at what angles you attack. I got hit a LOT more often at the same speeds and angles flying a P-39D than I did in an F6F, and I got hit a lot less in a P-40.

You don't even want to know what happens when you bring a Mustang within 700m of a Betty formation (and this is weird, because nobody can keep a Mustang absolutely straight and level in this game, but you can hear round after round banging into your nose and wings at an increasing tempo as you close to 500m or so).

Anyway, if I had killed twenty Betties within just 10 missions in 1 vs. 8 situations, for 8 hits into my aircraft in return, folks would probably want to know my secret, not the secret of some loser gunners. I'd certainly not run to my superiors and whine, after all, I'd be the highest ranking US bomber destroyer ace.

Standing off at ranges like that wouldn't be all that different from dropping the rounds on the Betty from a great height like so many little bombs. Setting your convergences for ranges more than double the accepted effective range of your guns makes them useless at any realistic range for the real targets of American fighters--other fighters. Unlike the ai, you can't adjust your convergences in mid-flight.

It's an artificial comparison. Killing 20 Betties at those ranges is not quite as unrealistic, impractical and improbable as the phony accuracy levels of the ai gunners, but it's in the ballpark.

cheers

horseback

[Pursuivant]

Originally Posted by JtD

Funny, considering the several thousand rounds spent, I would say a hit once in a while is statistically prone to happen. After all, firing solutions don't get a lot easier than with a dead 6 approach.

Alright, take 2% accuracy against a relatively non-maneuvering target as your baseline hit percentage overall. It's historically documented by several people on this forum.

Park your fighter at 4-5 o'clock or 7-8 o'clock off a bomber's beam at about 300-400 meters and fly in formation. Determine hit percentages from that. Anything much about 2% means that gunnery is too good.

Originally Posted by JtD

Can you prove it is theoretically impossible to hit an aircraft from 500-700 m away?

Of course not; the guns used were hypothetically accurate out to several miles, and lucky hits could happen at any range. But, as Horseback and I have been arguing, there were many factors that made that level of accuracy statistically highly unlikely. And, if you fly enough missions offline, the stats start to add up and they're historically wrong.

Originally Posted by JtD

Anyway, if I had killed twenty Betties within just 10 missions in 1 vs. 8 situations, for 8 hits into my aircraft in return, folks would probably want to know my secret, not the secret of some loser gunners.

That's not the point. You achieved your kills by hanging out at the extreme limits of effective range for any gun and used the superior stability of the Hellcat's guns to pick off some of the most flammable aircraft in the game. That's not a typical situation, nor is it realistic tactics, although it is a testament to some pretty impressive shooting on your part.

I'm more interested in what happens over dozens of missions where fighter pilots make more realistic attacks (pursuit curves, like I posted above) with their guns set at historic convergence distances (typically 300 meters for U.S. planes, although pairs of guns could be set to converge at different distances).

Due to a combination of inept AI bomber interception tactics and AI gunnery which is perhaps too good in some situations, I'm getting a situation where the gunnery accuracy by bombers is much better than it was historically. This makes it impossible for me to recreate very typical missions, such as late war German fighters intercepting U.S. heavy bombers.

[Pursuivant]

Originally Posted by horseback

And separate the pilots from the gunners, if you want to maintain some sort of bombing accuracy.

Actually, there should be seven axes for crew skill:

Level Bombing
Dive Bombing
Torpedoes
Rockets/Ground Attack
Fixed Aerial Gunnery
Flexible Gunnery
Piloting

And, different crewmen should have different skill levels. For example, Ace tailgunner and Rookie ball gunner, or Average bombardier and Veteran pilot.

[Pursuivant]

Originally Posted by JtD

Right, I guess 8 kills vs. 0 hits were too much of an argument to deal with, had to be avoided by starting some new "ace fighter AI is porked" red herring sideline. I take it the point now is that AI gunners mustn't hit anything even when faced with the most idiotic opposition. *facepalm* Nothing else to say, really.

That's NOT what I said.

I think that given the tactics used by the AI Hellcat in the 1 vs. 1 and 1 vs. 8 missions, the Japanese Ace gunnery is about right.

Rookie gunnery might be a bit too effective, however.

Gunnery might still be too effective at all levels when swinging a heavy gun - like a 20 mm cannon - while a plane is banking.

At least for Japanese gunners on the G4M1, they start shooting at extreme ranges - 1000 meters or greater.

I also previously said that hanging out at the limits of effective range for most air guns, even at 6 o'clock, and taking sniper shots is a good way to get lots of kills and minimal damage.

A better test of AI accuracy would be to take maneuvering shots from the rear quarter but within the effective range of the AI guns.

Originally Posted by JtD

Just out of curiosity - have you ever tried something bigger, like say 16 vs. 32 or something? Or is 1 vs. X the upper limit?

Yes.