Mitsubishi Stuka and Thunderbolt

[JtD]

Gun convergence was set to 400 meters. Very typical for US fighters of the time. Quite possibly back then they knew a gun works beyond point blank range.

If you can manage a fifty meter radius around an F6F sized target, conservatively assuming linear distribution, one in 800 rounds will hit the F6F.
Statistically, 8 hits require 6400 rounds. Ten mission with eight enemy aircraft, every aircraft needs to fire exactly 80 rounds to achieve that outcome.
The ground tests from the worst mount gave a ten meter radius, the best mount gave a two meter radius at about that distance. That leaves room for somewhere between 40 and 48 meters of aiming error. That's like missing a disc 3 feet wide from 20 feet away.
Just as a ballpark.

As to the "regardless of speed or angle, you will be hit, period" - I shot down eight of them one flight without being hit at all, and I did it from close range. Period. It's called "proper tactics", and it works better than whining.

[horseback]

Originally Posted by JtD

Gun convergence was set to 400 meters. Very typical for US fighters of the time. Quite possibly back then they knew a gun works beyond point blank range.

Actually, the USN/USMC standard was 1000 ft, which works out to a bit under 275 meters. Against smaller targets that maneuvered energetically, many successful American pilots opted for shorter convergences, particularly in Europe, where some guys felt that a three second burst at 150 yards was more efficient and accurate than several bursts at longer ranges. If you had an assigned aircraft in the Army Air Forces, you could usually have your convergence range set to your preference. Some groups in the ETO were known to have a 'group' setting at some specified range, usually under 200yds/180m.

In the USN/USMC during WWII, the vast majority of fighters' guns were sighted in at 1000 ft, period. BuAer was pretty strict about it. The main exception was for night fighters, which were generally boresighted at about 200 yards.

If you can manage a fifty meter radius around an F6F sized target, conservatively assuming linear distribution, one in 800 rounds will hit the F6F.
Statistically, 8 hits require 6400 rounds. Ten mission with eight enemy aircraft, every aircraft needs to fire exactly 80 rounds to achieve that outcome.
The ground tests from the worst mount gave a ten meter radius, the best mount gave a two meter radius at about that distance. That leaves room for somewhere between 40 and 48 meters of aiming error. That's like missing a disc 3 feet wide from 20 feet away.
Just as a ballpark.

"Ground tests" means that the aircraft was parked, the target was not moving and that the gun was sighted in, strapped down at the handles and fired remotely.

Adding a human to fire the gun will increase the error significantly, and the error would vary from person to person, and day to day. A human element tears the statistical curve to shreds, but the error set with the gun strapped down will be an absolute minimum (and likely less than half the error of the best human result for an open mount gun on a pintle or scarff ring mount).

Applying those same accuracy standards to a moving platform firing at a target moving constantly and randomly relative to the firing platform is comparing apples to oranges.

When you compare the accuracy of a turret mount in a B-17 to a scarff ring mount in a G4M, that is comparing apples to watermelons. BIG watermelons, the kind that win prizes at the county fair. Clearly, you've never been in a ballpark any more than you've ever fired a real machine gun.

Unless you found a way to set a formation of 8 aircraft, you had two formations of 4 aircraft, generally separated by about 700-1000m apart, too far away to lend mutual support except in rare instances where you wandered in between them. So at any given time, the greatest probability is that you were being shot at by a maximum of 4 aircraft initially, and as you shot them down one at a time, that number decreased down to one before you went after the next formation of 4.

I tried your test a few times with an F6F-3 and a convergence of 500m, which allows me some accuracy beginning at 600 meters. Since you specified that you hung back at more than 500m, I consistently pulled away at 520 to 450 meters, and then made another approach. The Rookie Betties began firing intermittently when I got to 900m of them and their bursts increased in frequency as I got closer. I took as little evasive action as possible and kept my speed at approximately 190 kts, which allowed me to slowly approach and peck away. I took hits fairly regularly at 750 and closer, and lost my engine once, got a wing shredded twice, and on the third or fourth go-around actually got them all before I ran out of ammo. I took more than 8 hits by the second pass every time.

Repeating the mission four times in a P-51C, I took hits to the engine every time, once at a range of over 750 meters which resulted in a runaway prop. Smaller target, less stable at that speed, which means that I should have been harder to hit, but the opposite was true. I was hit more often on the nose or engine compartment, and suffered greater damage sooner.

As to the "regardless of speed or angle, you will be hit, period" - I shot down eight of them one flight without being hit at all, and I did it from close range. Period. It's called "proper tactics", and it works better than whining.

Head on passes were rare in the PTO; they were not time effective--you only got one shot per pass, and after that one pass, the bombers were often into the task group's flak zone. I used the standard high deflection attacks that were standard USN doctrine at the outset of the war (and were spectacularly effective when applied). I begin with an altitude advantage of 1000m or more, steeply diving from at least a five o'clock position, and usually more like 3:30-4 o'clock, which generally puts me past my target's tail at speeds around 320 knots, or over 360 mph/580 kph, and puts my target between me and his wingmen.

I use my speed to zoom back up to a 3500 ft advantage, get back over them and to one side and lather-rinse-repeat... I still get hit often as I am in my dive or as I cross my target's tail which are both high deflection, tiny window shots for the top ring mount gunners and a nearly blind shot for the guys in the stingers as I blow through their cones of fire at 160 meters per second less than fifty meters away. The hard part should be avoiding his elevators.

The tactics are fine. They worked very well historically. But as long as the 'Rookie' ai gunners are more deadly than Arnold Schwarzenegger, Sylvester Stallone and Bruce Willis in their prime all wrapped into one, the game is unrealistic and artificial the moment you introduce an aircraft with guns that are pointing in more than one direction.

cheers

horseback

[Pursuivant]

Originally Posted by horseback

In the USN/USMC during WWII, the vast majority of fighters' guns were sighted in at 1000 ft, period. BuAer was pretty strict about it. The main exception was for night fighters, which were generally boresighted at about 200 yards.

Good info.

Originally Posted by horseback

When you compare the accuracy of a turret mount in a B-17 to a scarff ring mount in a G4M, that is comparing apples to watermelons.

On a related note, operational histories for the B-29 indicate that the long 20 mm cannon in the tail was removed from later models since it actually caused the tail of the plane to yaw when it was turned, due to slipstream effects. I have to wonder if there was something like that on the G4M, at least for certain models. In any case, it shows that even putting a "stinger" in the tail of a plane, where you'd expect that slipstream effects would be minimal wasn't always the case.

Originally Posted by horseback

Clearly, you've never been in a ballpark any more than you've ever fired a real machine gun.

Hey, play nice! You don't know where he's coming from.

Originally Posted by horseback

Unless you found a way to set a formation of 8 aircraft, you had two formations of 4 aircraft, generally separated by about 700-1000m apart, too far away to lend mutual support except in rare instances where you wandered in between them.

Additionally, the tail gun for the G4M1 only carried a limited number of rounds of ammunition and each drum of ammunition had to be manually changed (5 drums - one mounted, four stored, I forget if they were 20 or 50 rounds).

Originally Posted by horseback

I used the standard high deflection attacks that were standard USN doctrine at the outset of the war (and were spectacularly effective when applied). I begin with an altitude advantage of 1000m or more, steeply diving from at least a five o'clock position, and usually more like 3:30-4 o'clock, which generally puts me past my target's tail at speeds around 320 knots, or over 360 mph/580 kph, and puts my target between me and his wingmen.

This is the "pursuit curve" I mentioned in an earlier post. And, it wasn't just used by the USN.

[JtD]

So, I guess effective range means that beyond 600 yards, the bullets would just fly off to elsewhere, instead of continuing on their paths. No point in firing a gun at targets further out. Gunners wouldn't fire at anything too far away, because of their implemented radar, they knew to a foot how far the target was away. And of course, they were immune to psychological things, so they'd happily get fired at from 601 yards, without returning fire. Automatic fire with a mounted gun sort can't manage to stay within 3 feet over 20 feet distance anyway and gun dispersion changes if a human touches a gun instead of a remote control. Horseback can't set up a mission where formations support each other, so no one can. Someone programming the game adds if clauses to the AI gunners that make them behave differently depending on the targeted aircraft. 2% is an established figure for gunner accuracy, covering all conditions, because someone on the internet mentioned the figure. Even though 16 veteran Hellcats can wipe out 16 standard G4M with little loss to themselves, the historical results aren't there because 1 Hellcat can't do the same.

[horseback]

Originally Posted by JtD

So, I guess effective range means that beyond 600 yards, the bullets would just fly off to elsewhere, instead of continuing on their paths.

Words mean things. “Effective Range” is a specific term. Effective range means that the weapon in question is not reliably accurate for a given purpose beyond that point; in the case of a .50 caliber M2 heavy machine gun, mounted in an aircraft wing or turret mount, about 600 yards or 5 and one-half soccer fields is its maximum effective range. Other accepted definitions:
-Absolute maximum effective range: This the "this round is not considered lethal after crossing this threshold" distance. Neither of the other two common "maximum range" values will be greater than this. Purportedly, NATO defines this as the point at which the projectile's kinetic energy dips below 85 joules (62.7 foot-pounds). This is typically claimed when recounting that the P90's effective range is 400 meters on unarmored targets, as classified by NATO. It's worth noting that while the P90 looks neater than the civilian PS90, the extra barrel length increases the muzzle velocity and thus the civilian model actually has a longer absolute max effective range.
-Maximum effective range on a point target: This is the maximum range at which an average shooter can hit a human-sized target 50% of the time. "Point target" is basically a euphemism for hitting a human torso sized area in this context. If this range were greater than the absolute maximum, the absolute maximum would be quoted (a non-lethal hit may be accurate, but it's not effective).
- Maximum effective range on an area target: This is the maximum range at which an average shooter can hit a vehicle-sized target 50% of the time. In other words, this is the maximum distance at which it would make sense to open fire on a group or vehicle, etc. If this range were greater than the absolute maximum, the absolute maximum would be quoted (a non-lethal hit may be accurate, but it's not effective).

As I recall, the game makes rounds ‘disappear’ after they’ve traveled 1000 meters, which means that while the ai gunners in the Betty can shoot at me when I get within 900 meters and have a chance of the bullets hitting my aircraft, I have to get a bit closer before my bullets will reach them before disappearing.

No point in firing a gun at targets further out. Gunners wouldn't fire at anything too far away, because of their implemented radar, they knew to a foot how far the target was away.

see the above response. AI gunners don’t fire until they know that the bullets might reach their targets; human gunners in small formations and limited ammo supplies had to wait until they had a reasonable chance of (a), hitting their target, and (b) dissuading the attacker (by how close the tracers came, or actually hitting them) from getting closer. The AI routines absolutely make full use of precise knowledge attacking aircraft’s range and directional vectors, but the real life human beings they are supposedly representing could not possibly have done so.

And of course, they were immune to psychological things, so they'd happily get fired at from 601 yards, without returning fire.

Here, we have to keep in mind a few things: first, that the gunner has a limited supply of bullets, second if he’s firing 7.7mm LMG rounds, the farther away his target, the less accurate he will be AND the less damage he will do, and third, how does he figure out how far away his intended target is. Big WWII fighters were generally smaller than modern fighters are today; an F-16 dwarfs a P-38 (and I’ve seen both in close proximity to each other), and a Hellcat doesn’t begin to compare to the size and bulk of an A-10. A Hellcat’s fuselage viewed head-on in flight is about the size of a large SUV rolled onto its side (with the wings added to the roof and belly). From six soccer pitches away that is a pretty small target for the unaided eye.

Automatic fire with a mounted gun sort can't manage to stay within 3 feet over 20 feet distance anyway and gun dispersion changes if a human touches a gun instead of a remote control.

Here, you start with a false equivalency and go on to hysterics. First, the target is moving—constantly and in several random directions at once. Second, the platform the shooter is firing from is also moving, and that movement is also to a lesser degree random and unpredictable to the shooter. Third, you focus on objects 600 yards away somewhat differently than you do if you are focusing on something six and two thirds yards away. You simply don’t perceive these things in the same way, particularly if they are moving.

Gun dispersion absolutely does change when a human being is controlling the handles. Machine guns have this thing called recoil and vibration or gun shake; it is modeled for the wing guns of the fighters—lose even one gun on a wing and see what happens. Even if you have that tiny distant point zeroed in where that Hellcat is going to be when the first bullet gets there, the gunshake will knock your aim askew; you’ve undoubtedly seen videos of some poor sap firing a shotgun or high-powered rifle for the first time and being knocked off their feet by the recoil, so you can imagine what would happen when the same poor sap pulls the trigger on an equally powerful weapon and fires three or six rounds in a split second. Even with the body of the weapon tied to a hard mount, a large portion of that energy still has to go somewhere.

Horseback can't set up a mission where formations support each other, so no one can.

On the contrary, you didn’t specify that the mission you set up was not a QMB; some of us don’t monkey around with the FMB in depth, and QMBs are the quickest and easiest way to go. In any case, my own experiment using QMB and your general conditions got me hit significantly more often by fewer defending aircraft than you report that you were.

Someone programming the game adds if clauses to the AI gunners that make them behave differently depending on the targeted aircraft.

I don’t think that there are necessarily ‘if clauses’ specifically added; I think that there is a clear hierarchy. If the Mustang is classed as being more fragile than the Hellcat, then it must take more damage sooner, and therefore must be hit. My results seem to reflect that; the P-51C tends to vary in altitude and direction more than the Hellcat, even with constant power and prop pitch settings and it is a much smaller target by any measure, but it gets hit more often at the same (ridiculous) distances.

2% is an established figure for gunner accuracy, covering all conditions, because someone on the internet mentioned the figure.

2% was quoted by one poster here from talks with a relative who actually trained as an aerial gunner during the WWII period. It is consistent with other known testimony. If you’re familiar with the actual process of shooting at a target sleeve from another aircraft with a flexible gun from 150-200 yards maximum distance, you would have to think that 2% is a ridiculously high baseline for targeting a maneuvering aircraft approaching rapidly and hitting it from approximately five soccer field lengths away.

Even so, it would be a vast improvement over the current model.

Even though 16 veteran Hellcats can wipe out 16 standard G4M with little loss to themselves, the historical results aren't there because 1 Hellcat can't do the same.

If the Player cannot obtain the same results of his AI wingmen when he is doing exactly the same thing, something is skewing the results. Which of these things is not like the other? Hierarchy.

cheers

horseback

[Pursuivant]

Originally Posted by JtD

So, I guess effective range means that beyond 600 yards, the bullets would just fly off to elsewhere, instead of continuing on their paths.

Or course not! Bullets beyond 600 yards would still have some chance of hitting something, but by that point the cone of dispersion is such that they have a very low chance of doing so.

Originally Posted by JtD

Gunners wouldn't fire at anything too far away, because of their implemented radar, they knew to a foot how far the target was away.

You're right that there should be some randomness built into the point where gunners start to shoot, but one of the big lessons that gunners got (at least later war US gunners) was target ranging.

It was pretty easy to "guesstimate" a target's approximate range using the gun's sight, and shoot only when the target got into the correct range.

But range estimation was the least of the gunner's problems. The bigger problems were estimating speed (both of the gunner's plane and the target plane), estimating proper lead for deflection shots, estimating bullet drop (especially for shots above and below 0 degrees of angle) and coping with all the "random factors" which made guns less accurate.

So, I've got absolutely no problem if rookie gunners start shooting at 1,500 yards distance (a common rookie mistake was to start shooting way too soon), but their shots shouldn't be at all accurate until the enemy gets much closer.

On the other hand, ace US gunners should only start shooting within, say +/-10% of 600 yards (or 1,000 yards prior to late 1943) and should generally have better fire discipline (e.g., shorter bursts, less risk of hitting friendly planes).

Originally Posted by JtD

gun dispersion changes if a human touches a gun instead of a remote control.

Actually, this is one of the issues we're complaining about. Human gunners should be about as accurate as AI gunners, and human gunners (and possibly some AI gunners) are too accurate in the game.

Originally Posted by JtD

Horseback can't set up a mission where formations support each other, so no one can.

Actually, it's pretty difficult to set up mutually supporting flights in the FMB, too.

Originally Posted by JtD

2% is an established figure for gunner accuracy, covering all conditions, because someone on the internet mentioned the figure.

I never suggested that 2% accuracy is the perfect percentage for gunner accuracy in all conditions.

I mentioned it as a historically documented (by a WW2 veteran gunner and by the son of a veteran based on his dad's service books) acceptable standard for rookie gunners shooting at target drogues in order to graduate from USAAF/RCAF flexible gunnery school.

That means that 2% is a "ballpark figure" for what rookie AI should be able to against a maneuvering target under more or less ideal conditions.

In any case, the 2% figure wasn't meant as a challenge, it was meant as a suggestion for a starting point for calibrating AI gunnery skill.

If you were to take that 2% figure, rework AI gunners flying from a plane flying straight and level in Clear weather, so that they got about 2% hits on average against fighters maneuvering in the plane's 4-8 o'clock arc within 600 yards, flying at about 200-250 mph (about the speed of most target towing planes), I'd be a very happy man.

And, if that 2% average included higher hit percentages for shots directly to 6 o'clock, and a lower percentage of hits as the target fighters got out to 600 (or 1,000) yards, I'd be ecstatic.

From there, it would be easy to calibrate accuracy upwards or downwards for skill, poor visibility, turbulence, etc.

Originally Posted by JtD

Even though 16 veteran Hellcats can wipe out 16 standard G4M with little loss to themselves, the historical results aren't there because 1 Hellcat can't do the same.

Again, the test missions are only used as a way of generating statistics on AI air gunner performance (and AI fighter behavior) to compare against historical averages. They're not meant to refute actual historical statistical outliers. I can also accept outliers within the game, as long as on average the game mirrors historical performance.

[Igo kyu]

Originally Posted by Pursuivant

I can also accept outliers within the game, as long as on average the game mirrors historical performance.

Yeah, agreed, but the outliers have to be included in the cases that generate the average, which should mean that the mode goes down relative to the mean when the outliers get horrendously better than the mean. When that doesn't happen, you get a broken sim, as this one is in this aspect.

I believe that the fighters score a lot better than they ought to, and the bombers do too, the bomber gunners are more obviously wrong but both need fixing. A lot more hits ought to do insignificant damage on both sides, pilot kills at 500 metres ought to be a rare event, not one flight in twenty. A machine gun is effectively a very large bore slow acting shotgun, it's not like a rifle at all, even if it fires rifle calibre bullets, but the AI bomber gunners get results snipers would be proud of.

[Woke Up Dead]

Originally Posted by Igo kyu

I believe that the fighters score a lot better than they ought to...

Right, a lot has been said about what should limit the AI gunners in bombers, but our gunnery is not limited by whatever real fighter pilots faced.

[Pursuivant]

Originally Posted by JtD

As to the "regardless of speed or angle, you will be hit, period" - I shot down eight of them one flight without being hit at all, and I did it from close range. Period. It's called "proper tactics", and it works better than whining.

I'm not whining about MY tactics. I can engage bombers and do pretty well.

Instead, I'm looking at AI performance overall. I want to be able to set up missions where bombers and fighters behave like they did historically, and where the results of AI combat roughly match expected historical results.

Objectively:

* Fighters, even Ace AI, don't use proper bomber interception tactics.

* Gunners, even Ace AI, start shooting far beyond effective range.

* We still don't have much data on just how good WW2 era flexible gunners were in the air against maneuvering targets. What data we do have suggests that 2% accuracy against a relatively easy target was the accepted standard for rookie gunners, 600 yards (~550 meters) was the accepted range at which a 0.50 caliber MG had any chance of hitting, and that different gun mounts had different levels of inherent accuracy (ignoring other factors).

Subjectively,

* AI Gunners might still be too effective, particularly for shots made while the aircraft is maneuvering, for larger guns, for manually-turned guns, and for shots made to anything other than the plane's 6 o'clock.

* Gunnery by human players might be unrealistically accurate because the game engine doesn't model things like airframe vibration, turbulence, slipstream effects on guns and bullets, physical requirements of slewing guns around, g-forces, illusory effects of tracer rounds, and so forth.