Me410 Gunner is Far Too Accurate

[CUJO_1970]

2 hours ago, 357th_KW said:

Using a lower skilled AI (which is likely reducing the gunner’s performance) dialed back the aggressive maneuvers enough to where the 410 gunner got a few snap shots off, which were very dangerous - like you saw in the one engagement where the gunner actually fired on you,  it’s going to do some damage that would likely have prevented that Mustang from flying home from Berlin in the real world.

 

I have some more ideas for ways to test this that will prevent the AI pilot from pinning his poor gunner back in his seat.  To be continued ….

 

I'll have to try it with the AI toned down...I just thought it would be more of a challenge with them set to ace.

 

I wonder how human gunners online are doing in the backseat of the 410? Do human back-seaters have the physiology limitations? I don't know, I've never even played as a gunner online.

[357th_KW]

1 hour ago, CUJO_1970 said:

 

I'll have to try it with the AI toned down...I just thought it would be more of a challenge with them set to ace.

 

I wonder how human gunners online are doing in the backseat of the 410? Do human back-seaters have the physiology limitations? I don't know, I've never even played as a gunner online.

Those are good questions.  I’m also assuming pilot skill level changes the gunners skill as well.  And then there’s the question of what skill level gunner (if they are different levels after all) is given to a player?  You can’t assign an AI level to a MP spawn point aircraft in the mission editor AFAIK.

[6./ZG26_Custard]

2 hours ago, CUJO_1970 said:

I wonder how human gunners online are doing in the backseat of the 410? Do human back-seaters have the physiology limitations? I don't know, I've never even played as a gunner online.

If you enter the gunners seat You don't seem to have physiology applied to the gunner but of course if the pilot is pulling wild manoeuvres it makes it a damn sight more difficult to hit aircraft as you would expect in real life. However, if an aircraft is sitting nice and level on your six, it's easy to score hits, something the current AI seems to have a deal of trouble doing.

 

1 hour ago, 357th_KW said:

nd then there’s the question of what skill level gunner (if they are different levels after all) is given to a player?

If you are flying the aircraft and jump to the gunners positions the aircraft just stays in level flight, no fancy manoeuvres etc.    

 

Edit: The same AI is shared by all the gunners so I can only assume the turret guns of the 410 make that AI way more effective but I'm not sure to be honest 

Edited January 12, 2023 by 6./ZG26_Custard

[CountZero]

34 minutes ago, 6./ZG26_Custard said:

If you enter the gunners seat You don't seem to have physiology applied to the gunner but of course if the pilot is pulling wild manoeuvres it makes it a damn sight more difficult to hit aircraft as you would expect in real life. However, if an aircraft is sitting nice and level on your six, it's easy to score hits, something the current AI seems to have a deal of trouble doing.

 

If you are flying the aircraft and jump to the gunners positions the aircraft just stays in level flight, no fancy manoeuvres etc.    

 

Edit: The same AI is shared by all the gunners so I can only assume the turret guns of the 410 make that AI way more effective but I'm not sure to be honest 

2X13MM MG shooting with great angle is differance, AI is same on all airplanes, AI skill is selected by mission maker in spawn base for each airplane.

 

It is same thing like it was with Pe-2 for years, it was main complainer because it had big gunn on gunner position, every other airplane gunner could do same things but ppl mostly complained about pe-2 because airplane was fast and had biggest gun on gunner position that could one shot you anytime.

 

Back then we also had AI gunners shoting from all kined of weard positions, and tracking you even when they cant see you, and G limits not effecting gunners, we dont have that now. 

 

But nature of complain is same, good airplane with TWO big gunns shoting from gunner position, and no wonder players cry about it... this time its Axis airplane...

 

Its like wonting to nerf Hurricane IIc because in game it can do wonders it could not do in ww2, because in game it haz perfect 20mm x 4... when was Hurricane so good historicly as itz in game when flown online, its pocket battleship with thouse guns and +14lbs boost. And yet 410 gunners are to be behaving historicly LOL

Again fighters only problem, having standards for bombers/ground attackers they dont have for fighter airplanes.

 

 

[354thFG_Drewm3i-VR]

21 minutes ago, CountZero said:

 

 

Its like wonting to nerf Hurricane IIc because in game it can do wonders it could not do in ww2, because in game it haz perfect 20mm x 4... when was Hurricane so good historicly as itz in game when flown online, its pocket battleship with thouse guns and +14lbs boost. And yet 410 gunners are to be behaving historicly LOL

Again fighters only problem, having standards for bombers/ground attackers they dont have for fighter airplanes.

 

 

Red herrings like this should really be left out as they derail threads and we are all guilty of it from time to time. The rest of your post was good and reasonable, but just when we were finally getting somewhere and coming to a deeper understanding of the issue of bomber gunners in general and the 410 in particular, you go and bring up the completely unrelated Hurricane... ?

[69th_Mobile_BBQ]

 Neither a LaGG-3 or a Yak-7 can really catch a 410 on the deck.  They can creep up slowly or keep pace but, can't really catch up.  

The upside(?) is that you can get into the 410 gunner's long range and reliably dodge the tracers until they run out of ammo.  

If you have a friend that's above the 410 and waiting for the right moment to dive, just keep agitating the gunner to fire off all his bullets - and reloads - then let your friend whack him.  

[357th_KW]

Spent a little time using the Quick mission duel setup - took a 410 against an average AI P-51B.  As long as I dodged the first headon pass, and then just left the 410 in autolevel, the gunner was producing at least a "mission kill" (fuel/oil/radiator leak that would have been a kill in the real world) and in many cases an outright kill by hitting the pilot, setting the fighter on fire, even chopping its wing off before I was shot down.  On a number of runs the AI gunner actually got 2 fighters before I was shot down.  I tried about 5 times to see if I could do this from the gunner's seat, and was unable to even come close.  I might have landed a hit or two over 5 attempts - shifting between the 3 sights, and dealing with the limited visibility in that seat is a challenge.

[Tuna_Lead]

On 1/6/2023 at 2:37 PM, 357th_KW said:

 

None of which pertain to the topic of this thread.   Please stop trying to derail the discussion.

AI gunner effectiveness is actively being discussed, and people are comparing the ME410 gunner to the "legacy" options- this is showing what the comparison is. The gunners of the others are useless- those that claim the 410 in particular (thereby a standout) is too effective with a gunner have only these as a comparison- and don't complain about, as Custard put it, the situation being "perfectly acceptable to park behind bombers without a care in the world." I have only flown IL2 for a week, coming here after just shy of 3 years doing DCS WW2, where gunners on the ground and air are gods of lead, and found the AA to be refreshingly "moderate" for accuracy, especially on the non TAW servers. As for AI bomber gunners (bombers being one of the reasons I have given IL2 a try), I have found them in all but one of over a dozen encounters useless- even when diving to give them time, where my gunners are firing away for well over a minute, I eventually slam the earth in a wreck and only once- once- have I even seen the attacker leaking anything, much less defeated. Gunner behavior is abysmal. I know this is off because twice I have manually manned the guns on the JU88A4, and while one got away mostly unscathed the other belched fuel and coolant and was forced to pull off- I had no problem hitting him because he did not evade. I suppose he expected an AI to fire at him. 

[357th_KW]

18 hours ago, Vapor_12 said:

AI gunner effectiveness is actively being discussed, and people are comparing the ME410 gunner to the "legacy" options- this is showing what the comparison is. The gunners of the others are useless- those that claim the 410 in particular (thereby a standout) is too effective with a gunner have only these as a comparison- and don't complain about, as Custard put it, the situation being "perfectly acceptable to park behind bombers without a care in the world." I have only flown IL2 for a week, coming here after just shy of 3 years doing DCS WW2, where gunners on the ground and air are gods of lead, and found the AA to be refreshingly "moderate" for accuracy, especially on the non TAW servers. As for AI bomber gunners (bombers being one of the reasons I have given IL2 a try), I have found them in all but one of over a dozen encounters useless- even when diving to give them time, where my gunners are firing away for well over a minute, I eventually slam the earth in a wreck and only once- once- have I even seen the attacker leaking anything, much less defeated. Gunner behavior is abysmal. I know this is off because twice I have manually manned the guns on the JU88A4, and while one got away mostly unscathed the other belched fuel and coolant and was forced to pull off- I had no problem hitting him because he did not evade. I suppose he expected an AI to fire at him. 

The other AI gunners aren’t useless.  First of all they have PERFECt, superhuman SA.  You will never be surprised if the gunner is set to fire, and they can even spot targets that are out of their field of view, or in the sun or on the other side of clouds.  Second, they actually do score kills from time to time - this month on Combat Box it appears the 110G2 gunner has shot down around 50 aircraft.  The 110 is yet another example where we have a ton of historical gun camera footage and after action reports that tell us that USAAF pilots just flew up behind them, mostly ignored their gunners and shot them down.  And just like the 410, we know the 110’s we’re shot down in large numbers by USAAF fighters to the point where they were withdrawn from service due to the casualties.  And just like with the 410, we can see the losses of the USAAF fighter units, read the missing aircrew reports, and see that it was exceptionally rare for a gunner to even hit the attacking fighter, let alone shoot it down.  
 

 

 

Most of the gunners in the game produce a pretty realistic result - they are pretty ineffective, and generally won’t be enough to deter an enemy fighter, particularly if we’re talking about just one or two aircraft in formation.  The 410 gunner on the other hand consistently damages or kills the aircraft attacking it - an extremely unrealistic result.  Looking at Combat Box stats again for the same time frame, we see that the 410 gunner has shot down around 120 aircraft, in less then half the light time of the 110G.  Clearly something is amiss, and it’s producing ridiculous results.

[6./ZG26_5tuka]

2 xMG131 speak a different language when it comes to detering (usually straight and level approaching) fighters.

 

Also we don't know for sure the stats are showing gunner kills only. The HS-129 also has some pretty air to air scores (I wonder if this is backed by historical records, maybe  Hs129 is OP?) without any help of gunners undoubtetly.

 

So the stats you see could be literally anything, from airfield raid victims to main armament and/or gunner kills.

On 1/23/2023 at 8:29 PM, 357th_KW said:

  Looking at Combat Box stats again for the same time frame, we see that the 410 gunner has shot down around 120 aircraft, in less then half the light time of the 110G.  Clearly something is amiss, and it’s producing ridiculous results.

Historical results can't be reproduced in an unhistorical environment. The same goes for having "random" engine failures, toxic exhaust fumes, deafening noise, airframe build quality issues, fuel contaminations, supply shortages, effects of physical and mental stress and so much more.

 

All influencial factors when it came to the aerial war yet excluded from the game to make it appealing to the casual sim pilot.

 

So if you think that one technically much improved gunner station being more sucessfull when handled by the very same AI does not match your expectations of history you may think of the other dozens of things that you don't have to deal with to provide you with a comfortable fighter pilot expirience.

 

Maybe it isn't so bad to use tactics when engaging a not so much defenseless target for a change.

Edited January 25, 2023 by 6./ZG26_5tuka

[357th_KW]

On 1/25/2023 at 12:38 PM, 6./ZG26_5tuka said:

2 xMG131 speak a different language when it comes to detering (usually straight and level approaching) fighters.

 

Also we don't know for sure the stats are showing gunner kills only. The HS-129 also has some pretty air to air scores (I wonder if this is backed by historical records, maybe  Hs129 is OP?) without any help of gunners undoubtetly.

 

So the stats you see could be literally anything, from airfield raid victims to main armament and/or gunner kills.

Historical results can't be reproduced in an unhistorical environment. The same goes for having "random" engine failures, toxic exhaust fumes, deafening noise, airframe build quality issues, fuel contaminations, supply shortages, effects of physical and mental stress and so much more.

 

All influencial factors when it came to the aerial war yet excluded from the game to make it appealing to the casual sim pilot.

 

So if you think that one technically much improved gunner station being more sucessfull when handled by the very same AI does not match your expectations of history you may think of the other dozens of things that you don't have to deal with to provide you with a comfortable fighter pilot expirience.

 

Maybe it isn't so bad to use tactics when engaging a not so much defenseless target for a change.

 

I chatted with @=FEW=Revolves (who wrote the IL-2 stats mod) for a bit the other day, and he came to the same conclusion that I did.  The “total aircraft” listed on the main aircraft ranking page is showing the total kills for the aircraft.  The kills listed on the killboard page are just kills scored by the pilot (so forward guns, bombs etc ).  So if we look at the difference, that is showing us kills scored by the gunner.  The problem is that in the mission log, gunner kills are attributed to an ai gunner object rather than the player, and so the stats page doesn’t know how to handle it.  And unfortunately they seem to record slightly differently for all aircraft - so while they are separate like this for the 110 and 410 for instance, others like the Ju88A4 may not even be recording at all.  Revolves is going to try to add a “gunner kills” stat in a future version of the stats mod so that will make this much less time consuming to try and track and compare from one aircraft to another.

[Stonehouse]

Still think you are going about this discussion from the wrong direction. The weapon definition is the same across all AI users of the weapon and the calculation of aim error is the same across all AI gunners. So the only difference logically possible between individual aircraft would seem to be the actual turret position definitions.

That is,

• firing arcs available
• rotation speed of the turret - can the weapon be brought to bear quicker on target than the turret on your baseline plane of choice. eg Me410 turret rotation speed v's bf110 turret rotation speed.
• how many weapons can be brought to bear on the target.
• The caliber of the weapon and ammo type

The Me410 does have two turret positions defined (as they have different firing arcs) each with their own MG131 but both these turrets are controlled by a single gunner. I believe 1 gunner = one calculation of aim error. So there is no difference per se for aim error calc between the Me410 gunner and the B25 top gunner with 2 guns.

 

A turret position does not have any aiming error modifiers that can be passed into the aim error calc. So it isn't something like the Me410 gunner has some special gunsight making them more accurate via an aiming error modifier.

 

A turret position can have a barrel temp decrease modifier and a bullet dispersion angle modifier. These could influence accuracy.

 

I think the questions you should be trying to find the answer for to decide whether the Me410 gunner is too accurate are:

Do the Me410 turrets have better fields of fire?

How many weapons can overlap the firing arc of interest? eg attacker is in the six o'clock arc. Noting that each gunner is an aim error calc so 2 gunners with a single weapon each firing at a target will not be as effective as 1 gunner with 2 weapons.

How many rounds are in the magazine for the weapon? How many reloads and how fast is the reload time? Me410 has 500 rounds per gun & no reloads. Bf110e2 has 75 rounds in the mag and 10 mags in reserve, a g2 has 850 rounds in the mag and no reloads. When reloading the gunner obviously stops firing until reloaded. In the stock game the burst length in secs is the same regardless of weapon. So the 110e2 gunner will stop firing after 75 rounds and reload while likely the Me410 has about 400 rounds per gun still to fire. (Not sure which has the higher rate of fire off the top of my head)

Do they rotate quicker than say the flexible unpowered single gun in the Bf110 or whatever you feel is a good base line?

Does the extra weapon under a single gunner's control make a difference? Not many luftwaffe aircraft have more than a single weapon for each gunner bot.

 

Finally, are the firing arcs and rotation speeds and temp decrease modifiers and dispersion angle modifiers etc historically correct? I think finding an incorrect value is your best bet to get a change in the Me410.

 

Looking at kills statistically for say the BF110 or whatever and Me410 isn't really valid as the turrets defs and number of weapons and even weapon type (MG131 x 2 v's a single MG15 in a BF110e2 & single MG81 in a g2) involved are different and so you should expect a difference given that the aim error calcs are the same across the board. To do the statistical analysis you would need to have very similar situations to compare.

 

PS One other point about using stats - you don't say that I can see but do the stats break down to the level of kills by aircraft type/crew quality? If they don't then the results will not be conclusive for that factor alone as the aim error calc has a significant difference applied across the skill levels. You would probably also want the stats by gunner position as well rather than simply a total kills by gunners for each plane type.

Edited January 28, 2023 by Stonehouse

[Stonehouse]

Did some checking for you. The only other turrets I can see using the MG131 are the He111H16 top turret and the Ju88c6 top turret. The conclusion you are trying to prove is that the Me410 gunner is more accurate than either of these other gunners. However, the Me410 turrets have a much better field of fire than either of the other two turrets which cannot really fire below the aircraft.

 

Therefore, I think to use statistics to prove your conclusion you would need to create a test case that eliminated the differences in anything other than gunner accuracy. So, for example, the defending German aircraft would need to fly straight and level in all tests and not maneuver. In each case the defending crew skill level would need to be the same. The attacking aircraft would need to be the same in all instances and execute the same type of approach at exactly the same approach speed eg directly stern approach by say a P51. The approach chosen should not disadvantage any of the defending turrets nor advantage them. The attacking aircraft should not fire nor maneuver - it is simply a target drone. This eliminates the field of fire issue. You would not want any other turret on the He111 or Ju88 to be able to engage the attacking aircraft or else your results are skewed.

 

The He111 and Ju88 both carry 1000 rounds for the single MG131 while the Me410 carries 500 per gun and has 2 guns. I'd probably set up the test case so that the P51 approaches at slow overtake as the lower the velocity vector difference the less error there is. Stop the test once the P51 is out of action. This should make the only difference in the test the number of guns. Find out the number of hits on the P51 - I would suggest tacview (base version is free) as this gives easy access to such info.

 

You would have to repeat the test case for each defending aircraft type sufficient times to build up a decent statistical set of data. You would then need to extrapolate the He111 and Ju88 data to take account of the difference in number of guns and compare the Me410 results. Guessing it would be reasonable to double the number of He111/Ju88 hits and then compare to the Me410.

 

Honestly - the above is a lot of work, and I don't think it will conclusively prove anything because I already know that for the same gun using the same aim error calc with the same gunner skill in the same firing situation you are likely to get approx the same results and the greater the volume of data the more likely it will trend to be the same accuracy.

 

Assuming the turret definitions of the Me410 are correct I suspect that a likely root cause for the Me410's apparent over accuracy is that it has excellent fields of fire (much much better than the He111 or Ju88 top turret) and the guns reorientate very quickly. So there is not really any point except when the target is in the front hemisphere that at least one gun cannot bear on the target (on other German aircraft the aircraft structure often masks the target or field of fire limits stops the gunner firing as the bomber maneuvers). Coupled with the stock games 5G limit on gunners firing I think the Me410 is more maneuverable than say a He111 and even when pulling up to 5G can fire at a target pretty much anywhere in the rear hemisphere. I do also note that the turret definition bullet dispersion modifier is slightly different in the Me410 turrets to the Ju88 top turret but the me410 turrets are identical to the He111H16 top turret in this regard. Not sure why it is different for the Ju88 but I suspect possibly the mounting is not as good in the Ju88 so it gets a very slightly larger dispersion value. Barrel temp decrease modifier is the same across all 3 aircraft. 

 

So I believe it comes back to finding a flaw in the Me410 turret defs and providing that proof to the devs to get something changed.

Edited January 28, 2023 by Stonehouse
slight rewording for clarity. Ju88 Mg131 top turret has a slightly larger dispersion modifier than Me410 or He111H16 top turret

[357th_KW]

On 1/27/2023 at 4:38 PM, Stonehouse said:

Still think you are going about this discussion from the wrong direction ...

 

<SNIP>

 

PS One other point about using stats - you don't say that I can see but do the stats break down to the level of kills by aircraft type/crew quality? If they don't then the results will not be conclusive for that factor alone as the aim error calc has a significant difference applied across the skill levels. You would probably also want the stats by gunner position as well rather than simply a total kills by gunners for each plane type.

 

So, again - as this is a simulator, if we recreate a historical engagement and get results that vary wildly from what we would expect historically, that should be a big red flag.    But like you say, using an in game comparison to remove the variables is probably a better way to identify bugs, which I why I did that in my first post. 

 

But clearly that wasn't enough, so I did it again, this time with myself flying the Me410 and Ju88C (since it can be armed with 2 x 13mm's as well - though I doubt it shoots both at the same time) instead of the attacking fighter.  I did this in a quick mission, set my own group as "Ace" (although I'm not sure this modifier has any affect on the player aircraft, as there is no way to even set it for player spawns or MP spawnpoints in the mission builder - I think humans just get a set AI level for their rear gunner) with a "normal" AI P-51B attacking from above and behind, and just left the plane in auto-level on a medium throttle setting.  I then jumped in the external view on the attacker to watch the fun:

 

Me410:
1st engagement: P-51B made two passes, took damage to its radiator and fuel tank and disengaged - mission kill in game, outright kill in real life, ME410 surivived with damage and a pilot wound.
2nd: P-51B pilot killed on second pass, Me410 damaged.
3rd: P-51B set on fire on second pass, Me410 undamaged.
4th: P-51B took damage on the first pass and apparently the pilot was wounded as well, it disengaged.  Me410 undamaged.  Again, mission kill in game, likely an outright kill IRL.
5th: P-51B took damage on both passes, oil, fuel and possibly radiator and pilot and then disengaged.  Me410 might have taken a hit or two.

 

Ju88C-6:
1st: P-51B made 3 passes, finally took a few hits on the last pass.  Appeared to be superficial damage, but it disengaged. No damage to Ju88
2nd: P-51B made 5 passes, took 1 hit on pass 2.  Killed the Ju88 pilot on the last pass.
3rd: P-51B made 4 passes, took a hit on pass 2, took another hit or 2 on pass 4 - all superficial damage.  Flamed the Ju88s left engine on pass 4.
4th: P-51B made 4 passes, took a couple hits on pass 2 with no apparent damage, hit to the fuel tank on pass 3 causing a leak, and some more superficial damage on pass 4 and disengaged.  Ju88 suffered some engine/radiator/fuel tank damage.
5th: P-51B made 3 passes, took no hits, flamed the Ju88's left engine on the 3rd pass.

 

Five trials per aircraft isn't a lot, but the trends seemed pretty obvious, and are in agreement with what I'm seeing in my own gameplay, previous testing and MP stats.

 

Observations:

 

The "normal" AI is awful at attacking a static target like this, but at least makes a very consistent dead 6 approach - the upside was that I got to see a lot of gunner engagements with identical setups.  The gunners for both aircraft opened fire on every pass with short bursts, typically starting outside of 1000 meters.  The 410 gunner landed hits on EVERY pass.  And the P-51 never managed a 3rd pass, it was either dead or running away - this would have easily been 4 or 5 kills IRL due to the radiator/fuel tank hits.  The Ju88 gunner landed hits on 7 of 23 passes, and managed two mission kills one of which probably would have gotten home IRL.  The 410 potentially has more firepower as it would often be hitting with 2 guns instead of one, but again, the accuracy is just far better - not just hitting more often, but hitting center of target (and thus getting engine fires, pilot kills and radiator strikes) when the Ju88 gunner is landing hits out on the wings etc.

 

Here's a link to all the tracks.

 

Again, my biggest takeaway is that if the 410 gunner was anywhere near this effective in the real world, it would have shot down quite a few attacking fighters.  Pilots would have become wary of attacking it from behind, much like Axis pilots quickly realized that attacking a formation of US heavy bombers from the rear wasn't a great idea.  That isn't what happened in real life, and the results it achieves in the game are significantly better then any other gunners, and so I think there's a bug buried in there.  It may not appear in the files we can see, but there's definitely something off about it's performance.

[Stonehouse]

I'm not arguing about there being a bug. I've no opinion either way. All I am saying is there is a limited number of things that feed into accuracy and ability to hit targets in this game for AI gunners.

 

The weapon definition, the skill level, the aim error calculation and the turret definition. For a given weapon (eg MG131 in this case) the first three are common for all situations. So, the Ju88c6 and He111H16 top turret and the Me410 all use the same definitions and logic for the first 3.  Not sure what human gives your AI gunners but agree it would be one of the 4 values. Probably not novice.  <edit> confirmed as Ace skill for AI gunners on human aircraft in both SP and MP 

 

To the best of my experience digging around in the game files, the only thing that is left is the turret definitions.

 

So, I was suggesting to you that checking the Me410 turret definition for flaws/incorrect values is likely your best hope for giving the devs a reason to change the Me410. Probably you would have to do quite a bit of research to get the relevant values but likely the community would help. There are a lot of people with good reference libraries and material here.

 

Otherwise, if the first 3 are common and the Me410 turret is fine I'm struggling to see a way for you to get a change happening. I don't see that your statistical approach as you seem to be approaching it will prove anything conclusively. That's just my point of view and opinion having worked with statistical analysis in the past though.

 

That's all. I was trying to be helpful. I was not disputing or agreeing with your hypothesis that the Me410 is too good at hitting targets.

 

I guess you could be correct and that there is something hidden from view that makes the Me410 too good. I just don't see any evidence of anything in the aim error calc that does something special for the Me410. But if true I just don't know how you will prove it well enough to get a change instigated. 

 

Anyway, honestly hope you get the issue sorted one way or the other. I've attached the game logic used for calculating the aiming error for your interest. Each turret is assigned a bot definition, each turret gunner bot points back at this attached logic to calculate the aiming error. Note that this version of the file has comments both my own and others and translations from Russian. If you feel it is not a valid copy all you need to do is ungtp the file scripts.gtp and compare the actual file to the attached.

turretcontrollerai stock.txt

Edited February 7, 2023 by Stonehouse
Corrected comment about human/AI skill level. Attaching the aim error calc fyi

[6./ZG26_Custard]

2 hours ago, 357th_KW said:

Again, my biggest takeaway is that if the 410 gunner was anywhere near this effective in the real world,

It's already been pointed out to you that historical results can't be reproduced in an unhistorical environment. We are all flying in easy mode in GB's and there are so many things that are unrealistic and I wish could be changed.

I know so many fighter pilots just want to stroll up behind bombers and take them out without breaking a sweat . What we currently have in a game that is "trying" to be a sim is AI that is quite frankly redundant when it comes to mounting almost any kind of defensive fire with the 410 being the exception (probably because of the turret FOF)  I'm sure that hurts the P-51, Spit and Tempest pilots no end. However, showing Gun cam footage (that would have been cherry-picked to show the best examples ) of gunners not returning fire or pilot reports doesn't give us the whole picture.

If folk really want to get that little bit  closer to realism for Normandy and late war scenarios the server admins need to reduce the player slots significantly for axis aircraft, but then of course the servers would be mostly empty.  

Personally, I think the AI, ammunition and damage modeling needs a complete overhaul and most certainly ballistics and other effects need to be added also. 

[C6_lefuneste]

4 hours ago, 357th_KW said:

 

So, again - as this is a simulator, if we recreate a historical engagement and get results that vary wildly from what we would expect historically, that should be a big red flag.    But like you say, using an in game comparison to remove the variables is probably a better way to identify bugs, which I why I did that in my first post. 

 

But clearly that wasn't enough, so I did it again, this time with myself flying the Me410 and Ju88C (since it can be armed with 2 x 13mm's as well - though I doubt it shoots both at the same time) instead of the attacking fighter.  I did this in a quick mission, set my own group as "Ace" (although I'm not sure this modifier has any affect on the player aircraft, as there is no way to even set it for player spawns or MP spawnpoints in the mission builder - I think humans just get a set AI level for their rear gunner) with a "normal" AI P-51B attacking from above and behind, and just left the plane in auto-level on a medium throttle setting.  I then jumped in the external view on the attacker to watch the fun:

 

Me410:
1st engagement: P-51B made two passes, took damage to its radiator and fuel tank and disengaged - mission kill in game, outright kill in real life, ME410 surivived with damage and a pilot wound.
2nd: P-51B pilot killed on second pass, Me410 damaged.
3rd: P-51B set on fire on second pass, Me410 undamaged.
4th: P-51B took damage on the first pass and apparently the pilot was wounded as well, it disengaged.  Me410 undamaged.  Again, mission kill in game, likely an outright kill IRL.
5th: P-51B took damage on both passes, oil, fuel and possibly radiator and pilot and then disengaged.  Me410 might have taken a hit or two.

 

Ju88C-6:
1st: P-51B made 3 passes, finally took a few hits on the last pass.  Appeared to be superficial damage, but it disengaged. No damage to Ju88
2nd: P-51B made 5 passes, took 1 hit on pass 2.  Killed the Ju88 pilot on the last pass.
3rd: P-51B made 4 passes, took a hit on pass 2, took another hit or 2 on pass 4 - all superficial damage.  Flamed the Ju88s left engine on pass 4.
4th: P-51B made 4 passes, took a couple hits on pass 2 with no apparent damage, hit to the fuel tank on pass 3 causing a leak, and some more superficial damage on pass 4 and disengaged.  Ju88 suffered some engine/radiator/fuel tank damage.
5th: P-51B made 3 passes, took no hits, flamed the Ju88's left engine on the 3rd pass.

 

Five trials per aircraft isn't a lot, but the trends seemed pretty obvious, and are in agreement with what I'm seeing in my own gameplay, previous testing and MP stats.

 

Observations:

 

The "normal" AI is awful at attacking a static target like this, but at least makes a very consistent dead 6 approach - the upside was that I got to see a lot of gunner engagements with identical setups.  The gunners for both aircraft opened fire on every pass with short bursts, typically starting outside of 1000 meters.  The 410 gunner landed hits on EVERY pass.  And the P-51 never managed a 3rd pass, it was either dead or running away - this would have easily been 4 or 5 kills IRL due to the radiator/fuel tank hits.  The Ju88 gunner landed hits on 7 of 23 passes, and managed two mission kills one of which probably would have gotten home IRL.  The 410 potentially has more firepower as it would often be hitting with 2 guns instead of one, but again, the accuracy is just far better - not just hitting more often, but hitting center of target (and thus getting engine fires, pilot kills and radiator strikes) when the Ju88 gunner is landing hits out on the wings etc.

 

Here's a link to all the tracks.

 

Again, my biggest takeaway is that if the 410 gunner was anywhere near this effective in the real world, it would have shot down quite a few attacking fighters.  Pilots would have become wary of attacking it from behind, much like Axis pilots quickly realized that attacking a formation of US heavy bombers from the rear wasn't a great idea.  That isn't what happened in real life, and the results it achieves in the game are significantly better then any other gunners, and so I think there's a bug buried in there.  It may not appear in the files we can see, but there's definitely something off about it's performance.

I'm not sure using IA to define if gunner are too deadly or not is a good way, because IA fighter usually sit in 6 of the target waiting to be shot down.

At the opposite, I never have been saved by my 410 gunner when flying online. Every meeting with an ennemy fighter ended as my plane beeing shot down.

[Barnacles]

 

 

This twitter thread is a good read, regarding gunners. It seems there are things here to support both that the gunners are too effective, or not enough.

Dipl.-Ing Robert Lusser - Calum Douglas (calum-douglas.com)

Edited February 7, 2023 by Barnacles

[357th_KW]

57 minutes ago, Barnacles said:

 

 

This twitter thread is a good read, regarding gunners. It seems there are things here to support both that the gunners are too effective, or not enough.

Dipl.-Ing Robert Lusser - Calum Douglas (calum-douglas.com)

Here’s the relevant passage from Herr Lusser’s letter (though the whole thing is quite interesting and worth a read):

 

11.) Neglect of on-board weapon development.

The need for strong defensive armament, for the bomber and reconnaissance aircraft, has been unrecognized for years. Until the outbreak of the war, the position, which could easily be proved to be wrong, was adopted that our German bombers were so fast that they were very difficult to catch and attack by enemy fighters. The Heinkel company in particular represented this point of view. You only need to stick out a “broomstick” or a “tranquilizer” to drive away the opponent!

Since 1936, on the occasion of the armament of the Me 110, I have repeatedly drawn attention to the fact that our bombers with the drum-fed MG 15 guns have a very weak defense and finally in January 1939 I proved in an investigation submitted to the RLM that a normal fighter ( Me 109) was superior to the normal bomber (He111) not by a factor of 7, as was believed, but by a factor of more than 500 in aerial combat. This investigation remained unnoticed for 4 months and was only retrieved from the files after my personal inquiry. No consequences were drawn from this.

It was only when our bombers were shot down by the English fighters in frightening numbers that it became apparent that our defensive armament had been seriously neglected.

At the beginning of the war, the English had equipped their bombers (Wellington) with very modern power-controlled twin and later quadruple machine guns with belt feeds, which gave our fighters a hard time.

[Stonehouse]

On 1/29/2023 at 6:19 PM, 357th_KW said:

I think humans just get a set AI level for their rear gunner) with a "normal" AI P-51B attacking from above and behind, and just left the plane in auto-level on a medium throttle setting.  I then jumped in the external view on the attacker to watch the fun:

FYI. This came up again recently across in the AI gunnery mod thread. It was confirmed by a dev that AI gunners on human piloted aircraft are set to Ace in both SP and MP.

[150_GIAP-Red_Dragon]

It's a shame that it hasn't been fixed yet. Recently, I attacked the 410th on the Combat Box. He used jerking almost constantly! I couldn't aim because I was constantly maneuvering to avoid being hit. Imagine my amazement when I saw that the accuracy of his shooting was 31%! 31 hellish unreal percentage in conditions when I was afraid that I would get a black screen because of the G-LOC! And this despite the fact that the A-20 rear gunners generally prefer not to shoot. I did tests for myself, where I just flew with Bf109 (almost level flight)  on six of A-20 and there was no shooting at all. I have a strong suspicion that the A-20 gunner have increased the dead zone around rudder.. Anyway it's all a shame

Edited September 4, 2023 by -332FG-Red_Pilot

[Props]

Yeah, those 410 AI gunners are insanely accurate and deadly. Combined with the 13mm cartridge it's very difficult to attack from anywhere in an arc around the stern. I adopted a high, or low, side approach or if possible a high deflection shot and preferably from the front at an angle. Sometimes it takes a while to setup a good shot this way, but I am more likely to survive. Even then the 410 seems to absorb an ungodly amount of damage compared to other twins.

[Roland_HUNter]

The developers should investigate this problem.
The ACE gunner could hit my plane from 1.24 km with 1 flick shot, and when in close range, the 410 lost his wing, in the spin the gunner damaged my engine.
In the past the Pe-2 had this problem.

[LukeFF]

49 minutes ago, Roland_HUNter said:

The developers should investigate this problem.
The ACE gunner could hit my plane from 1.24 km with 1 flick shot, and when in close range, the 410 lost his wing, in the spin the gunner damaged my engine.
In the past the Pe-2 had this problem.

 

Track files, please. ?

[Roland_HUNter]

20 hours ago, LukeFF said:

 

Track files, please. ?

I'll try to remake it.

[Roland_HUNter]

Done.
https://easyupload.io/zin60y

 

I have no idea what triggering the AI  to do flick shots, sometimes they just fly and do nothing.
Sadly, I couldn't recreate the wingloss-flickshot.

[LukeFF]

1 hour ago, Roland_HUNter said:

Done.
https://easyupload.io/zin60y

 

I have no idea what triggering the AI  to do flick shots, sometimes they just fly and do nothing.
Sadly, I couldn't recreate the wingloss-flickshot.

 

Thank you, I will pass along the report.

[superion]

I'm sad to see a post like this, would a video of a player shooting down a me410 without being damaged by it in the current game be a valid argument?????????????? Many want to turn this into a "warthunder with perfume" after this happens they will enjoy playing offline because a lot of people are stopping playing because of this.

[Stonehouse]

@LukeFF One difference I did notice while updating the AI Gunnery mod between another MG131 turret and the Me410 is the targeting script used. No idea whether it is relevant or significant to the issue, but it is a difference. 

The He111 H16 top turret also uses the MG131 and a similar mix of AP/HE ammo and has:

Spoiler

TargetingScript = "LuaScripts/WorldObjects/Ballistics/Projectiles/BULLET_GER_13x64_AP.bin"

 

The Me410 has:

Spoiler

TargetingScript = "LuaScripts/WorldObjects/Ballistics/Projectiles/BULLET_GER_7-92x57_APsmk.bin"

 

 

[LukeFF]

5 hours ago, Stonehouse said:

@LukeFF One difference I did notice while updating the AI Gunnery mod between another MG131 turret and the Me410 is the targeting script used. No idea whether it is relevant or significant to the issue, but it is a difference. 

The He111 H16 top turret also uses the MG131 and a similar mix of AP/HE ammo and has:

  Hide contents

TargetingScript = "LuaScripts/WorldObjects/Ballistics/Projectiles/BULLET_GER_13x64_AP.bin"

 

The Me410 has:

  Hide contents

TargetingScript = "LuaScripts/WorldObjects/Ballistics/Projectiles/BULLET_GER_7-92x57_APsmk.bin"

 

 

 

@Regingrave would be the better person to comment on this, as I've never really dug into the targeting script files.

[Yogiflight]

@Stonehouse are you sure the 410 ammunition is the turret ammunition, because 7.92x57mm is the rifle caliber of the MG 17s in the nose?

[Stonehouse]

3 hours ago, Yogiflight said:

@Stonehouse are you sure the 410 ammunition is the turret ammunition, because 7.92x57mm is the rifle caliber of the MG 17s in the nose?

It's a line in the turret definition file so it would be a reasonable assumption that it is referring to the turret MG131s. I can't be 100% sure because I can't see the code behind it. There are separate references/config for the nose weapons in the plane file, so I believe the info in the turret definition is specifically for the turret weapons.

I did build a little test mod that overrode the reference to 7.92 ammo to BULLET_GER_13x64_AP.bin. The change caused no issues and while I think from very limited testing (I don't have much spare time at the moment due to house reno's) that there was a difference it wasn't earth shattering.

 

In the end, as I am of the opinion that it is a bug/typo I updated the AI Gunnery mod to include the change of targeting script bin file. I uploaded the new version yesterday.

 

My guess is that just as I've done 100s of times at work to build a new pgm by taking an existing one that does most of what I want and cloning it and tweaking the clone because it saves time, the devs have taken an existing turret definition (e.g. a Bf110 perhaps) and tweaked it to make it into a Me410 turret definition rather than build the definition from scratch but missed changing the targeting script reference. It doesn't cause any obvious issues during testing so there would be a little chance of it being spotted unless there was a code review/walkthrough by a peer/colleague/other dev. In other words, a person different from the one that built it does a visual inspection of the turret definition looking for issues prior to release. It happens at work as it is a complex financial system with real world impacts if mistakes happen in the production application along with regulatory requirements that the development methodology is very robust with lots of checks. 

 

Spoiler

 

Edited October 30, 2023 by Stonehouse

[LukeFF]

On 10/27/2023 at 7:29 PM, Stonehouse said:

@LukeFF One difference I did notice while updating the AI Gunnery mod between another MG131 turret and the Me410 is the targeting script used. No idea whether it is relevant or significant to the issue, but it is a difference. 

The He111 H16 top turret also uses the MG131 and a similar mix of AP/HE ammo and has:

  Reveal hidden contents

TargetingScript = "LuaScripts/WorldObjects/Ballistics/Projectiles/BULLET_GER_13x64_AP.bin"

 

The Me410 has:

  Reveal hidden contents

TargetingScript = "LuaScripts/WorldObjects/Ballistics/Projectiles/BULLET_GER_7-92x57_APsmk.bin"

 

 

 

I just asked about this, and the reply was: "No, that doesn't affect accuracy. TargetingScript parameter is a different thing, not related to AI."

[Stonehouse]

4 hours ago, LukeFF said:

 

I just asked about this, and the reply was: "No, that doesn't affect accuracy. TargetingScript parameter is a different thing, not related to AI."

Thanks LukeFF. That's in line with what I saw when I overrode it with a small test mod, it didn't seem to change the accuracy of the AIs shooting. 

[354thFG_Drewm3i-VR]

On 1/12/2023 at 2:01 PM, CUJO_1970 said:

 

I'll have to try it with the AI toned down...I just thought it would be more of a challenge with them set to ace.

 

I wonder how human gunners online are doing in the backseat of the 410? Do human back-seaters have the physiology limitations? I don't know, I've never even played as a gunner online.

The 410 is probably the best dogfighter on Combat Box with low fuel (~30%) in skilled hands. Between the overpowered gunner, ability to "float" making slow speed handling godlike--in addition to its ability to pull lead and fire out in front--it is almost unbeatable. Add a comptent wingman and a slight altitudd advantage and it is game over for even a coordinated squad of P-47s/P-51s. With the types of schenanigans seen online (full rudder head on reversals), it is hars to image the 110/410 are moddled accurately. 

 

Additionally--and all of these years later--the 410 gunner is still, by far, the most effective in the game.

Edited April 28 by 356thFS_Drewm3i-VR