Which DM do you like more - before 3.008 or after?

December 26, 2018

Honestly, who knows?

The devs refuse to give any meaningful insight on the damage model. It is either not possible to adjust the values accordingly or they think that the values are correct according to their, I hate to say it, secret documents/calculations.

December 26, 2018

51 minutes ago, Operation_Ivy said:

Honestly, who knows?

The devs refuse to give any meaningful insight on the damage model. It is either not possible to adjust the values accordingly or they think that the values are correct according to their, I hate to say it, secret documents/calculations.

and the Bug Report section is just a panic button.

December 26, 2018

9 hours ago, Sgt_Joch said:

Well no, you are misreading the data.

I'm not.

9 hours ago, Sgt_Joch said:

There is no proof that it took an average of 4 30 mm hits to kill a B-17, that is purely a German estimate based on apparently studying gun camera footage.

Incorrect. The German average is based on analysing a large amount of gun camera footage AND on firing trials, as clearly stated in the German report presenting the averages. In other words the German average is derived from damage observed during controlled testing and actual operational results. Crucially both the German & British estimates are based on results with ALL three types of 3cm shells (HE, HE-T & INC.). So it's probably not a coincidence that the British arrived at exactly the same average

Also just incase you're next going to start questioning the value of gun camera footage, whilst a US report regarding the assessment of gun cam footage rightly points out that not all hits will show in such footage, this mainly refers to smaller caliber hits and non explosive hits. Meanwhile you can be pretty confident that a hit by a 3cm HE(M) shells doesn't get missed easily as we're talking a giant flash every time, thus even in grainy 70+ old footage it is still quite easily visible. So this realistically wouldn't have had an impact on the accuracy in number of MK108 hits observed.

9 hours ago, Sgt_Joch said:

The estimate in the US study is actually that a 30 mm shell hit to a P47 has a 28% chance of resulting in a kill, so yes, it could take 3-4 30 mm hits to kill a P47 based on that estimate.

No, the US estimate was that a P-47 hit a single time at a random spot by the MK108 from the front & below at a range of 500 yards had a 29% chance of going down within the next 4 min, and a 42% chance of not making it home.

And the above really is a best case scenario for a P-47 in the event of a MK108 attack, as fire coming from that direction, angle & range provides the pilot, rear fuselage, tail & most of the engine's auxiliaries great protection via the big engine block infront. With an attack from the rear aspect instead (as well as within a more sensible range) said 1 hit kill probabilities would've gone up considerably - a clue as to this is provided with the results vs the engine alone where chances of a 1 hit knock out went from basically nil from the front (the HE shell simply spending itself on the thick engine block), to a 50/50 from the rear where the softer auxiliaries were exposed.

Finally, once again, the US estimates are for HE shells only (apparently specifically HE-T shells based on the 72 g filling), and no INC. shells, hence the typical chances of fires and damage dealt via deeper penetrations is simply not realistically represented in the US estimates.

7 hours ago, Sgt_Joch said:

No, my statement is correct. Look at the chart on page 39 of the report. 28 % of an "A" kill on a P47 from one 30 mm HE hit.

Not it isn't as a "B" kill is also a kill, meaning the aircraft doesn't make it home.

Furthermore this estimate is once again for a shot taken at 500 yards (right at the edge of the MK108's effective range in terms of KE), from the front & below only, a situation where the P-47's engine block screens a large amount of vitals. Hence in a far more common rear attack both the "A" & "B" probability estimates would've gone up considerably as now the cockpit, tail & engine auxiliaries are all well exposed. Add to this the effect of incendiary shells not at all considered in the US study, and suddenly you're looking at a very low probability of the P-47 coming home after a hit.

Edited December 26, 2018 by Panthera

December 27, 2018

11 hours ago, unreasonable said:

As for the mixture of results: that is also my expectation. What is at issue here, apart from the number of kills due to engine/fuel damage, is the number due to structural damage. If the engine/fuel numbers are way too high that suggests some combination of toning down splinter damage or increasing engine/fuel cell damage resistance is in order. If the number of losses to structural damage is way too low that suggests that either local blast is under-modeled or wings etc are too damage resistant.

Well we got a pretty big hint as to where a big part of the problem most likely originates in a post by the developers themselves not too many days ago where they stated that currently the 3cm HE(M) shell is 2.5 times as powerful as the 20mm shells ingame. We know for a fact however that the 3cm HE(M) shells were more than 2.5 times as powerful as 20mm HE shells, even when compared with the 20mm HE(M) ones. The dramatic difference visible in pictures of Spitfires hit in the same location by either round alone bares testimony to this.

Finally the fact that the ingame MK108 lacks its incendiary shells is also reducing its effectiveness as these were pretty much guaranteed to start a fire if they hit a fuel cell, and if they didn't they packed a good amount of penetration power, being capable of punching straight through a Spitfires wing chordwise.

December 27, 2018

Here's an interesting Aircraft Engineering article on structural repairs:

Spoiler

This Hawker Hurricane Mk I N 2427 was extensively damaged by at least two 20mm cannon shells and several 7.92 mm bullets.

December 27, 2018

4 hours ago, Panthera said:

Well we got a pretty big hint as to where a big part of the problem most likely originates in a post by the developers themselves not too many days ago where they stated that currently the 3cm HE(M) shell is 2.5 times as powerful as the 20mm shells ingame. We know for a fact however that the 3cm HE(M) shells were more than 2.5 times as powerful as 20mm HE shells, even when compared with the 20mm HE(M) ones. The dramatic difference visible in pictures of Spitfires hit in the same location by either round alone bares testimony to this.

Finally the fact that the ingame MK108 lacks its incendiary shells is also reducing its effectiveness as these were pretty much guaranteed to start a fire if they hit a fuel cell, and if they didn't they packed a good amount of penetration power, being capable of punching straight through a Spitfires wing chordwise.

Since my tests are of the 37mm round (and will be about the conventional 20mm round) I fail to see how this is relevant. Any discrepancies between test and US results that appear to be down to the DM will be entirely general. Not everything is about the mineshells.

As for incendiaries: it is true that the loadout pages do not list a belting and explicit incendiary round. This same issue has arisen in the FC discussion. However, the test of whether a round is an incendiary is whether it "incendiarizes", not what it says on the load out screen. FC "ball" ammunition is causing plenty of fires in Camels. What matters is the p of causing fires. I am certainly getting plenty of fires testing 37mm.

BTW, you make much of the one angle issue: it is actually easy enough to make an adjustment so that you can get a rough idea of what the table would look like from other angles. All you have to do is use the aspect ratio of components to recalculate. The simplifying assumption is that the p that a hit on a component (E,S,P,F) causes a kill, is the same from any angle. You should have no problem with this, as all of the "4", 25%, 6% etc numbers we have seen have to be used in this way too. Anyway, it can be varied if other information is available.

I am surprised you have not done this already to show the US report's implicit p that hits to structure will cause a kill.

Edited December 27, 2018 by unreasonable

December 27, 2018

19 hours ago, sevenless said:

I´m not so sure if that is really the case. The charts I quoted are of a 1945 report by the US Military Analysis Division and part of the study called United States Strategic Bombing Survey dating Sept 30, 1945. The study you quoted is from 1950. If the US had access to the Buehler survey from 9/1944 at the time the US charts were produced in 1945, I am actually unsure about. Furthermore keep in mind that in the 1950s study the Buehler 9/1944 report is only mentioned in passing as a citation.

If you check the list of references from your document on pdf page 29, you'll find the Frontfilmauswertung by LFA as number 2 and reports of the Lilienthal-Gesellschaft as number 6. I don't think it is coincidence that two reports based on the same sources arrive at the same results, but of course, I cannot say that with certainty.

Edited December 27, 2018 by JtD

December 27, 2018

4 hours ago, Raymondo77 said:

Here's an interesting Aircraft Engineering article on structural repairs:

Hide contents

This Hawker Hurricane Mk I N 2427 was extensively damaged by at least two 20mm cannon shells and several 7.92 mm bullets.

Interesting indeed: perhaps 3 cannon shells, or possibly 4 judging from the graphic. Prop, left outer wing, left inner wing, left elevator. Note that in the US criteria this was not even a B kill; it not only RTBed but did not even need to be written off.

Planes RTBing in this condition would be comparatively rare, but it would be a mistake to assume that this was necessarily because the single shot kill probability was high, or that the compound effects of previous hits is especially important. For a fighter pilot, the most important factor is that if you have been hit by a cannon shell once, your chances of being hit again rise greatly due to impaired handling and speed: or just the tactical situation. Many pilots would bail once hit with an enemy on their tail, rather than wait for more hits that would make bailing out impossible.

December 27, 2018

5 hours ago, unreasonable said:

Since my tests are of the 37mm round (and will be about the conventional 20mm round) I fail to see how this is relevant. Any discrepancies between test and US results that appear to be down to the DM will be entirely general. Not everything is about the mineshells.

No, the damage done can easily be anything but general as pretty much everything depends on the individual dmg values given to the specific rounds within the two categories blast & shrapnel. Hence the comparative destructive power between different rounds ingame can easily be off so that some do the damage they realistically should whilst others don't even come close, and the earlier statement by the developers really does hint greatly toward this being the case. Another factor is the toughness of the individual aircraft, these could also be off on a comparative basis.

Also I've concerned myself specifically with the 3cm HE(M) shells this entire time because they are the ones I've noticed clearly doing unrealistic types of damage ingame, and more importantly it's the shells we have by far the most detailed data on at this point. The other rounds ingame I haven't really experienced that many "well that was odd" moments with ingame yet, and infact I'm pretty satisfied with the damage dealt by for examples the Hispanos at this point.

5 hours ago, unreasonable said:

As for incendiaries: it is true that the loadout pages do not list a belting and explicit incendiary round. This same issue has arisen in the FC discussion. However, the test of whether a round is an incendiary is whether it "incendiarizes", not what it says on the load out screen. FC "ball" ammunition is causing plenty of fires in Camels. What matters is the p of causing fires. I am certainly getting plenty of fires testing 37mm.

Again though the HE shells were also quite capable of causing fires, as shown in the British tests, but nowhere near at the rate of the INC. shells which on top also acted as deep penetrators into structures. The ingame MK108 is entirely missing this type of damage potential, despite the INC. shells at least taking up 33% of the typical belt load out, and according to some papers 50%.

5 hours ago, unreasonable said:

BTW, you make much of the one angle issue: it is actually easy enough to make an adjustment so that you can get a rough idea of what the table would look like from other angles. All you have to do is use the aspect ratio of components to recalculate. The simplifying assumption is that the p that a hit on a component (E,S,P,F) causes a kill, is the same from any angle. You should have no problem with this, as all of the "4", 25%, 6% etc numbers we have seen have to be used in this way too. Anyway, it can be varied if other information is available.

I make much of it because it's a big factor, one not as easily adjusted as you describe, the reason for this being tests at other angles not having been carried out at all. Hence the actual vulnerability of the pilot, tail & rear fuselage (not to mention what difference an attack from the rear to the wings would make) there is zero data on. All we can say for sure is that the probability of lethal damage would be much worse from the rear as now suddenly a lot vital parts are exposed.

5 hours ago, unreasonable said:

I am surprised you have not done this already to show the US report's implicit p that hits to structure will cause a kill.

Surprised? Unreasonable I've said right from the start that the US estimates are sensible, that when scrutinized do not really disagree with the UK or German conclusions or anything I've stated, and that they show quite clearly that structural damage is the main killer (against the B25 you've even got a case of a single hit to the wing causing an immediate kill, from 500 yards away). In other words I've never had a problem with the US figures, my problem is entirely about people wanting to use the US study to figure out the expected damage with hits to specific parts of an airframe, which simply isn't possible as the study provides no detailed data on this. What it does do is give you a good overall picture of chance with a random hit, making it quite useful for an airforce that wants to know overall how effective a gun is assuming random hit placement. For our purposes however it simply lacks the details on the damage with hits to specific areas such as the wings, tail, cockpit etc that we need for direct comparison in order to assess wether for example a hit or two to a wing (or tail, or rear fuselage etc.) should've realistically severed it or not. Here only the British & German firing trials provide data that can be directly compared with ingame results vs the same structural parts, as they contain the necessary details on hits to most of these parts & areas from a multitude of directions & angles.

Does that mean we can say with 100% certainty what the damage should be like vs every aircraft ingame? No, not all, but we got some pretty good points of reference to rely on if we were to make an educated guess on what is reasonable to expect - and 2.5 times as bad as a 20mm HE hit certainly aint it.

PS: One thing I haven't really mentioned yet is how much the range at which the US test were carried out is a likely cause of a difference in results, as at 500 yards the MK108 has lost a lot of its KE, likely enough to not always being able to penetrate the skin of the target irrespective of impact angle, which would lead to a lot of surface explosions instead of the more desirable & destructive internal ones that would've resulted had the engagement range been more in line with what is typical (150-300m). But with no details on each hit provided we are left guessing, which is the big problem with using the US study.

Edited December 27, 2018 by Panthera

December 27, 2018

2 hours ago, Panthera said:

I make much of it because it's a big factor, one not as easily adjusted as you describe, the reason for this being tests at other angles not having been carried out at all. Hence the actual vulnerability of the pilot, tail & rear fuselage (not to mention what difference an attack from the rear to the wings would make) there is zero data on. All we can say for sure is that the probability of lethal damage would be much worse from the rear as now suddenly a lot vital parts are exposed.

Surprised? Unreasonable I've said right from the start that the US estimates are sensible, that when scrutinized do not really disagree with the UK or German conclusions or anything I've stated, and that they show quite clearly that structural damage is the main killer (against the B25 you've even got a case of a single hit to the wing causing an immediate kill, from 500 yards away).

PS: One thing I haven't really mentioned yet is how much the range at which the US test were carried out is a likely cause of a difference in results, as at 500 yards the MK108 has lost a lot of its KE, likely enough to not always being able to penetrate the skin of the target irrespective of impact angle, which would lead to a lot of surface explosions instead of the more desirable & destructive internal ones that would've resulted had the engagement range been more in line with what is typical (150-300m). But with no details on each hit provided we are left guessing, which is the big problem with using the US study.

I cannot be bothered with any more of these walls of text full of factual inaccuracies, so just a few brief comments. We cannot at all be sure that the probability of lethal damage from the rear is worse: for the B-25 in the US report it is noted that from the rear engines are almost invulnerable to certain rounds. So there is data, you are just not reading it.

The US estimates do disagree with the German estimates sufficiently to require serious examination. Overall B kill p of 0.12- 0.16 on a B-25 is very different from 0.25 on a B-17 that has been mentioned. You can only get a mean of 4 to be consistent with a p of a kill of 0.25 if the probabilities are independent, BTW. In particular the idea that a wing hit on a P-47 should be nearly always lethal is very much at variance with the US report's findings.

The US report does not show that structural damage is the main killer for B-25s - fuel tank damage is the main killer, even for the 3cm. Read the chart. In the US analysis of German gun films, the analyst could only count fires, while pointing out that counting hits was impossible. But you seem to have no problem with no details on each hit being provided in the case of the German gun camera estimate.

As for the range issue: if you had looked at Appendix A: Completed firings, all the 3cm firings with the Mk108 listed therein were at 200 yards, compared to 500 for everything else. This appendix does not include firings at P-47 structures, but you can safely assume that the testers were well aware of the crappy ballistics of the Mk108. They were also aware of the fuze characteristics. They were not incompetents dabbling in matters beyond their comprehension, mistaking surface explosions for internal ones.

Edited December 27, 2018 by unreasonable

December 27, 2018

19 hours ago, Operation_Ivy said:

The devs refuse to give any meaningful insight on the damage model.

They are not obliged to.

December 27, 2018

2 hours ago, unreasonable said:

I cannot be bothered with any more of these walls of text full of factual inaccuracies, so just a few brief comments. We cannot at all be sure that the probability of lethal damage from the rear is worse: for the B-25 in the US report it is noted that from the rear engines are almost invulnerable to certain rounds. So there is data, you are just not reading it.

You cannot be bothered? No you're right, seemingly you cannot be bothered to conduct yourself tactfully when debating a subject, I've noticed that much. I've lost count of all the attempts at insults and attacks to my person, direct or indirect, at this point.

As for me not reading the data, the B-25's engines from the rear are listed with "A" kill probability ranging between 35-42% and a "B" kill probability of 45-70% from the MK108 HE-T shells only. For comparison from the front the mean probability figure is basically zero.

So if you're going to argue against a rear aspect attack by the MK108 being considerably more lethal than one from the front & below, then based on the engine vulnerability figures alone you're arguing against the US study you've been advocating for all this time.

2 hours ago, unreasonable said:

The US estimates do disagree with the German estimates sufficiently to require serious examination. Overall B kill p of 0.12- 0.16 on a B-25 is very different from 0.25 on a B-17 that has been mentioned. You can only get a mean of 4 to be consistent with a p of a kill of 0.25 if the probabilities are independent, BTW. In particular the idea that a wing hit on a P-47 should be nearly always lethal is very much at variance with the US report's findings.

No because the US estimates do not account for hits with all three types of shells fired during an attack, which alone greatly reduces the risks of fire and damage to components deeper into the structure, the types of damage which the 3cm INC. shells were meant to cause. Furthermore the US study doesn't try to arrive at an average nr. of hits to down a bomber, as the cumulative effect of several hits to the same target wasn't actually tested - if so they would've had to fire a lot more rounds at more examples of B-25's to assess the average nr. deemed lethal. Instead they combine their single shot probabilities later according to their formula. Futhermore the US figures are for a range of 500 yards, the very edge of the MK108's effective range.

Also once again, it's not just the German analysis which put the average at 3-4 hits to bring down a heavy bomber, it was the same conclusion of the British analysis after ground testing as well, which in addition noted that 3 hits to the wings with 3cm INC. would on average be enough to bring down a Lancaster.

It should also be noted that the British expected the Germans wanting to target the wings & engines as much as they could. A trend we also see in the gun camera films still left. The US study places no thoughts on this for the final probability estimate, concerning itself only with an even spread of fire.

2 hours ago, unreasonable said:

As for the range issue: if you had looked at Appendix A: Completed firings, all the 3cm firings with the Mk108 listed therein were at 200 yards, compared to 500 for everything else. This appendix does not include firings at P-47 structures, but you can safely assume that the testers were well aware of the crappy ballistics of the Mk108. They were also aware of the fuze characteristics. They were not incompetents dabbling in matters beyond their comprehension, mistaking surface explosions for internal ones.

I never said they mistook surface explosions for internal ones, you're once again trying to twist my words around. What I'm saying is that at 500 yards the chances of surface explosions goes up due to a significant loss of KE, hence when we're looking at a probability estimate that is for 500 yards then this same estimate can't apply to one from 200 yards away and vice versa without error. So unless the US study chose to use 200 yard results for a 500 yard estimate without adding in the effect of range, then we have to assume that the probability rating would be positively affected (even if only by a little) were they to do an estimate for 200y.

EDIT: Conversely in the case that they chose to use 200y results for a 500y estimate without any adjustment for range effects (an odd sudden lack of thuroughness), then the US estimate can't be applied to a 500 y scenario as it would then reflect results where shells were hitting with greater KE than they would infact be doing at 500 yards.

Edited December 27, 2018 by Panthera

December 27, 2018

1 hour ago, Brano said:

They are not obliged to.

quote me where i said they are? i am just stating facts.

December 27, 2018

1 hour ago, unreasonable said:

The US estimates do disagree with the German estimates sufficiently to require serious examination.

Personally I'm not convinced that the apparent differences could not be down to details in the evaluation that for now escape us.

December 27, 2018

9 minutes ago, JtD said:

Personally I'm not convinced that the apparent differences could not be down to details in the evaluation that for now escape us.

Me neither, and I have feeling that active targeting of specific parts could be a factor, the Germans & British focusing a lot on the vulnerability of the wings assuming this would be the most frequently targeted part - something gun camera film seems to support. Thus this could very well be a factor behind how they arrived at their averages.

Edited December 27, 2018 by Panthera

December 27, 2018

1 hour ago, JtD said:

Personally I'm not convinced that the apparent differences could not be down to details in the evaluation that for now escape us.

Neither am I: looking at the available sources for kill probabilities we are not exactly comparing like with like. But if you have to make a decision about what you expect the game to churn out in a given situation, you have still to quantify some expected probability.

What sort of details could count as accounting for the difference between the sources? Potentially and not exclusively:

1) Some are not counting at all

2) They are not counting the same things

2) They are not counting with the same degree of accuracy due to observational difficulties

3) The probabilities being generated from the counts do not have the same reliability due to sample size issues, and we often do not know what the sample size is.

Just to expand on that last point: as I am sure you know, suppose you have an average 25% kill/hit ratio for some munition/target combination: if you had got that from 25 first shot shoot downs out of 100 hit targets, you would have a 90% confidence range of ~18%-33%.

For the same ratio, if the sample was smaller: say 5 from 20, the 90% confidence range is now much wider: ~10% - 46%

While in either case the observed figure is still the best estimate without other corroborating data, the uncertainty is so large that hanging your hat on one single source and number is extremely unwise, especially if it is the worst documented. It is possible that the differences between the US tests and the German estimates, apart from being for different targets, are also merely an artifact of the sample sizes.

December 27, 2018

14 hours ago, Raymondo77 said:

Here's an interesting Aircraft Engineering article on structural repairs:

Hide contents

This Hawker Hurricane Mk I N 2427 was extensively damaged by at least two 20mm cannon shells and several 7.92 mm bullets.

Indeed interesting. Those hits were by only two/three 20mm FF Mine Shells. Now imagine it would have been hit by ten 20 mm Mine shells or two 30mm Mine shells.

3 hours ago, Panthera said:

the Germans & British focusing a lot on the vulnerability of the wings assuming this would be the most frequently targeted part - something gun camera film seems to support. Thus this could very well be a factor behind how they arrived at their averages.

The germans knew that near the No.3 engine were the central hydraulic system parts of the bomber and thus being most vulnerable. Destroying No3. and hydraulics was a most certain way to bring the bomber down. They tried whenever possible to hit this area, after killing the tailgunners/ball turett gunners first.

Edited December 27, 2018 by sevenless

December 27, 2018

7 minutes ago, unreasonable said:

What sort of details could count as accounting for the difference between the sources?

Well, mathematics are one thing, my thinking revolves more around technical open points. Basically: why would different people judge the same damage differently in terms of lethality? Structurally, things are load depended (air speed, weight, g load ...), aerodynamically, pilots skills play a huge role, fires are effected by fire suppressing systems, self sealing or even fuel running out. A P-47 and a Spitfire are quite different as far as structure is concerned, maybe a B-25 and a B-17 are too. A Lancaster is different from a B-25 for sure. Maybe the impact is considerable? And so on.

Only then we need to add difficulties in observation, probability of effective hits or mathematical models.

But yes, even if we only go with what you wrote, two studies giving 20% and 35% respectively could still mean the same thing with just 7.5% error each, which would be pretty good all things considered.

December 27, 2018

If the gap was small then yes, you could probably chalk it up to a normal margin of error.

But that does not apply when you have a German report stating you can take down a B17 with an average of 4 x 30 mm hits and a US report stating you need as much as between 10-17 x 30 mm hits to take down a B25 ( i.e 6 to 10 % probability).

Yes, we might be able to account for the discrepancy if we had access to the German report and see what parameters they used, but without knowing what the Germans based their conclusions on, it is futile to try to argue it is a more accurate conclusion.

Edited December 27, 2018 by Sgt_Joch

December 27, 2018

2 hours ago, Sgt_Joch said:

If the gap was small then yes, you could probably chalk it up to a normal margin of error.

But that does not apply when you have a German report stating you can take down a B17 with an average of 4 x 30 mm hits and a US report stating you need as much as between 10-17 x 30 mm hits to take down a B25 ( i.e 6 to 10 % probability).

Yes, we might be able to account for the discrepancy if we had access to the German report and see what parameters they used, but without knowing what the Germans based their conclusions on, it is futile to try to argue it is a more accurate conclusion.

The US report doesn't really try to arrive at an average, it combines the 1 hit probabilities, in which case it takes 9 of them to reach the average of 50%. But again this is without data on the cumulative effects of several hits to the same aircraft, and crucially without at all considering 33-50% of the rounds fired at the bomber, i.e. the Brandgranate (INC).

Furthermore I just found out that the single shell type which appears to have been used in the US tests, the 3cm M geschoß leuchtspur mit zeleger (HE-T, 73 g RDX+tracer), only came with a burst on impact detonator, and not the VC70 duplex detonator designed to make the shell burst only once inside the target. (disregard, turned out not to be the case)

So there's a good chance that the US on average didn't get more than half as good results as compared with the Germans & British who took into account all three of the major types of shells fired within a typical belt load out.

Edited December 27, 2018 by Panthera

December 27, 2018

33 minutes ago, Sgt_Joch said:

If the gap was small then yes, you could probably chalk it up to a normal margin of error.

We can chalk it up to margin of error even though it is a large gap, because the margin of error in this case is also very large, for instance structural damage on a B-25:

B kill probability: They are 95% confident that a hit will lead with a 1 to 17% probability to a B kill, estimated value: 6.4%.

It's obvious that with a that large confidence range basically any outcome can be covered by uncertainty of prediction. Just go with the 17% and you'll instead of an about 80% find a less than 50% survival chance for 4 hits just based on structural damage, not accounting for fires, fuel or pilot kills.

All possible, but as stated above, I suppose there are technical reasons for it, too.

16 minutes ago, Panthera said:

I just found out that the single shell type which appears to have been used in the US tests, the 3cm M geschoß leuchtspur mit zeleger

Where'd you find that? Quickly browsing the report I only find mentioning of "HE".

December 27, 2018

What I'd really like to see made right ingame is not limited to the "probability of shooting down x aircraft with a single random hit" or "the average number of random hits to down x aircraft", AKA the general effectiveness of the weapon system, as that number can be achieved whilst at the same time the aircraft are going down for completely unrealistic reasons, such as in our current case where a loss by MK108 is almost guaranteed to be due to critical engine damage. Hence I'm quite focused on also getting the cause behind the loss right, and here esp. the British trial reports are invaluable.

In short I'd like to see realistic outcomes with hits to specific parts & areas on the airframe, as otherwise aiming for real life weakspots (such as the wings in the case of the MK108) becomes pointless. And in order to achieve this there are number of things we need to get right from the beginning, such as, but not limited to, realistic belt load outs, main types of damage caused by each round and the comparative difference in destructive power between these rounds.

3 hours ago, sevenless said:

The germans knew that near the No.3 engine were the central hydraulic system parts of the bomber and thus being most vulnerable. Destroying No3. and hydraulics was a most certain way to bring the bomber down. They tried whenever possible to hit this area, after killing the tailgunners/ball turett gunners first.

With shells as destructive as the 3cm M shells it's likely that they often didn't even bother with the gunners and instead just focused on hitting the wing(s) as quickly as possible, as a few hits here would end things very quickly according to the British findings.

Edited December 27, 2018 by Panthera

December 27, 2018

1 hour ago, JtD said:

Where'd you find that? Quickly browsing the report I only find mentioning of "HE".

IIRC the shells fired are described as having a 72 g explosive filler. But I can't check now as Im once again on my phone. It's possible I could've mixed up two reports though, in which case we can't say what type was used ofcourse as "HE" doesn't really tell us much.

EDIT: Managed to locate the pdf online and find the page listing the weapon details, and I can see I must've misread "overall length 72" " as filler weight because of the small size of the page & text Actual filler weight appears to be 0.19 lbs (86 g), so US tests were undoubtedly with VC70 fuzes, albeit some would've been without. Thus this can't help explain some of the difference in results.

In short any difference has to come down to a combination of missing the effect from incendiary shells, not accounting for active targeting, cumulative effects of several hits & difference in sample size. That's all I can think of at least.

Edited December 27, 2018 by Panthera

December 27, 2018

1 hour ago, Panthera said:

With shells as destructive as the 3cm M shells it's likely that they often didn't even bother with the gunners and instead just focused on hitting the wing(s) as quickly as possible, as a few hits here would end things very quickly according to the British findings.

Yep might be. What could have happened if those 190 A8/R2s armed with MK 108s were able to attack in formation can be read here:

8th May 1944. After first being vectored to an assembly point in the Hamburg region, IV.( Sturm)/ JG 3 was then ordered to fly south to intercept the approaching bombers. Shortly after 1000 hrs, they sighted several boxes of B-24s on a south-easterly course heading in the direction of Brunswick. By chance, no US escort fighters were flying close attendance on the bombers. In less than ten minutes the Gruppe had claimed no fewer than 19 of the Liberators. The 12 destroyed tally closely with the 11 B-24 combat losses suffered by the Americans on this mission, while the seven Herausschüsse may have accounted for at least some of the additional seven written-off in crash-landings back in the UK.

Weal, John. Luftwaffe Sturmgruppen (Aviation Elite Units) (Kindle-Position588). Osprey Publishing. Kindle-Version.

19 minutes ago, Panthera said:

IIRC the shells fired are described as having a 72 g explosive filler. But I can't check now as Im once again on my phone. It's possible I could've mixed up two reports though, in which case we can't say what type was used ofcourse as "HE" doesn't really tell us much.

72g HTA + 3g Nitropenta = 75g HE Filler

HTA = Hexogen, TNT, Aluminium

Edited December 27, 2018 by sevenless

December 27, 2018

1 hour ago, sevenless said:

HTA = Hexogen, TNT, Aluminium

Yes, but I misread the details, the actual shell used in the US tests was either the "3cm M Geschoß 108 ausführung A mit zerleger" or "ohne zerleger" (with or without self destruct fuze), as the filler weight is listed as 0.19 lbs (86 grams), with either the regular impact duplex detonator or the delayed duplex VC70. So I'm quite certain they couldn't have missed testing shells with the VC70 as these were supposedly more numerous in a belt than HE with the regular duplex detonator.

Also interestingly that sheet above lists the VC70 in a HE-T round, but shows a regular duplex detonator on the illustration, so either the collector over at warrelics.eu who told me they weren't in those is mistaken or there's a mistake on that chart. Is it an original WW2 sheet?

Either way the lack of VC70 fuzes can't be a factor in explaining any difference in probability estimates between the US & UK + GER analyses as at least half of the Ausf.A shells were with the VC70, so any difference would have to be a combination of the US one missing the effect from incendiary shells, not accounting for active targeting, cumulative effects of several hits and/or a difference in sample size.

Edited December 27, 2018 by Panthera

December 27, 2018

47 minutes ago, Panthera said:

Is it an original WW2 sheet?

Yes that sheet is original from the german manual. I quote both rounds mit und ohne Zerleger, so you can see the difference:

Mit Zerleger:

And ohne Zerleger:

Edited December 27, 2018 by sevenless

December 27, 2018

@sevenless You misunderstood me, the Zerleger is the self destruct fuze which detonates the round incase it doesn't hit anything. I think the maximum flight time for a round with the Zerleger fuze was ~10 sec, then it would seld destruct to avoid the shell falling down and detonating somewhere unintended.

The duplex detonator is what actually ignites the explosive filler incase of impact, whilst the delayed duplex detonator known as VC70 allows for a further 10cm of flight after initial impact so that the shell detonates only once inside the target.

Duplex detonator (regular impact detonator):

Duplex vc70 detonator (delayed post penetration detonator):

Edited December 27, 2018 by Panthera

December 27, 2018

34 minutes ago, Panthera said:

@sevenless You misunderstood me, the Zerleger is the self destruct fuze which detonates the round incase it doesn't hit anything. I think the average flight time maximum flight time for a round with the Zerleger fuze was ~10 sec, then it would seld destruct to avoid the shell falling down and detonating somewhere unintended.

I think I have the Shell A with delay detonator here also. Wait a minute just need to get the correct pages.

OK here we go:

Ohne Zerleger:

Mit Zerleger:

And some interesting Info from the Fliegerfibel as to the change in ammos due to production streamlining:

In fact they had real problems producing these rounds in the necessary amounts. IIRC I recently came across an order of a Jagddivision about that problem. If I can dig it up I post it here:

http://forum.12oclockhigh.net/showthread.php?t=5459&highlight=108+cannon

The file is National Archives HW5/464 and the document is CX/MSS/T 147/39.

On 31 March 1944, 7. Jagddivision ordered that "In view of the present shortage of 3 cm ammunition MK 108, the units must help each other out. ZG 76 is to surrender to (Roman) I./JG 5 1000 rounds. The units will arrange for handing over between themselves."

British Intelligence added this note: "An order for setting up a Sturmstaffel on 20/11 stated that the MK 108 in the Me 109 had not been very successful (T 9/53). Other mentions of this weapon in e.g. T 19/82 (Test unit 25 Achmer fitting it obliquely in Me 110) and R 66 (B) 26 (to be adapted to engine DB 605 A 1)."

So there were problems with the MK108/Bf 109 combination.

Edited December 27, 2018 by sevenless

December 27, 2018

11 minutes ago, sevenless said:

I think I have the Shell A with delay detonator here also. Wait a minute just need to get the correct pages.

Yes, Ausf.A shells came with the vc70. It is the more streamlined HE-T shells which I was told didn't come with the vc70, and indeed I can't find any cutaway illustrations or pictures of the HE-T shells with the vc70, only the regular duplex detonator:

1) INC 2) HE(vc70) w/ PETN filler 3) HE(vc70) w/ H41 4) HE-T (impact) w/ PETN + H41 5) practice round, normally filled with tar

Edited December 27, 2018 by Panthera

December 27, 2018

3 hours ago, Panthera said:

The US report doesn't really try to arrive at an average, it combines the 1 hit probabilities, in which case it takes 9 of them to reach the average of 50%.

Well no, actually it does. See table 4, p.41, overall probability that one 30 mm hit will result in a "A" kill on a B25. 6 or 10% depending on how much fuel is carried.

December 27, 2018

42 minutes ago, Panthera said:

Yes, Ausf.A shells came with the vc70. It is the more streamlined HE-T shells which I was told didn't come with the vc70,

Please see my edited post above including the Ausf A shells.

Obvious difference between both is absence of the tracer element in the Ausf.A shell. IIRC generally every 5th round was a tracer round on the belts which were prepared at unit level, but that was decided at every unit differently. Some flew without any tracer rounds at all. Usually the Fw190s opened fire with the MK 108 at 400 metres or less, where you don´t need any tracers anymore. You have the tracers anyways from the 20mm 151s.

Edited December 27, 2018 by sevenless

December 27, 2018

11 hours ago, Sgt_Joch said:

Well no, actually it does. See table 4, p.41, overall probability that one 30 mm hit will result in a "A" kill on a B25. 6 or 10% depending on how much fuel is carried.

No, that's not an average for how many number of shots it takes to down the aircraft, it's the single hit kill probability within two criterias, "A" & "B". For the "B" criteria the 1 hit kill probability is 11.7 to 16.1%, i.e. there's a 11.7 to 16.1% chance that a single hit prevents the aircraft from returning home.

Thing is the chance of fire was 1) only considered for the HE shell 2) estimated based on too few actual firings to provide a reliable average as they had run out of B-25's to test against at this point. This is written on page 40 just prior to table 4.

Crucially though the US estimate is without considering the chance of an INC. round, which in reality would be either every 3rd or 2nd round fired, and each one would skyrocket the chances of a fire or damage to structures & installations hidden deeper inside the target. As already mentioned just 3 of these incendiary shells to the wing(s) was considered the average needed in order to bring down a Lancaster or Halifax bomber.

Edited December 28, 2018 by Panthera
11.7% not 10.1%

December 27, 2018

3 hours ago, JtD said:

We can chalk it up to margin of error

Well no, you can't say margin of error is the explanation when gap in the results is so wide. Someone has to produce the actual German report.

So does the German report still exist? If it does not, it calls into question this whole 4x30 mm hits to kill a B17 notion.

December 27, 2018

18 minutes ago, Sgt_Joch said:

So does the German report still exist? If it does not, it calls into question this whole 4x30 mm hits to kill a B17 notion.

It does. Buehler report 182/2 from 9/44, posted somewhere here already. And no, not existence of this report does not call into question something. The US report 59 of 30th September 1945 by Military Analysis Division corroborates the german report.

Edited December 27, 2018 by sevenless

December 27, 2018

3 minutes ago, Panthera said:

No, that's not an average for how many number of shots it takes to down the aircraft, it's the single hit kill probability within two criterias, "A" & "B". For the "B" criteria the 1 hit kill probability is 10.1 to 16.1%, i.e. there's a 10.1 to 16.1% chance that a single hit prevents the aircraft from returning home.

No, wrong again. As I have said from my first post. That table is the one shot probability that a 30 mm hit will result in a "A" kill on a B25 , 6 or 10 %.

Both of your responses to my posts were wrong. Why do you bother responding to my posts if you do not take the time to read and understand them?

December 27, 2018

16 minutes ago, Sgt_Joch said:

No, wrong again. As I have said from my first post. That table is the one shot probability that a 30 mm hit will result in a "A" kill on a B25 , 6 or 10 %.

Both of your responses to my posts were wrong. Why do you bother responding to my posts if you do not take the time to read and understand them?

Now you're just being silly, because you responded to me saying this:

Panthera

"The US report doesn't really try to arrive at an average, it combines the 1 hit probabilities, in which case it takes 9 of them to reach the average of 50%. "

with this

Sgt_Joch

"Well no, actually it does. See table 4, p.41, overall probability that one 30 mm hit will result in a "A" kill on a B25. 6 or 10% depending on how much fuel is carried. "

In other words you've repeated me twice now without at any point actually reading what I said in the first place.

Summary: The 1 hit kill probability estimate is not an attempt at estimating the average number of hits it takes to down a B-25, something the US study doesn't really seriously attempt to predict as the cumulative effect of several hits wasn't tested. Hence all they do later is combine the 1 hit kill probabilities, ending up at 9 to reach 50% AKA the average. A far cry from the 3-4 hit average concluded by the British & Germans based on both testing AND operational results, the reasons for which can be down to any of the below or all them combined:

- The US study missing the effect from incendiary shells

- The US study not accounting for the deliberate targeting of wings & engines

- The US study not accounting for the cumulative effect of several hits

- The US study not properly accounting for airstream effects

- A difference in sample size

Edited December 28, 2018 by Panthera

December 28, 2018

6 hours ago, sevenless said:

It does. Buehler report 182/2 from 9/44, posted somewhere here already. And no, not existence of this report does not call into question something. The US report 59 of 30th September 1945 by Military Analysis Division corroborates the german report.

It would be interesting to see that report, or at least what part of it corroborates the German report, and how. The US report BRL 727 analysing German gun camera results very much calls into question the German estimates. Hits are questionable, and so are "victories" awarded. The only purely MK108 attacks analysed are 5 attacks, leading to 8 assessed hits, producing 1 fire. From these gun camera records "4" is little more than a guess.

Edited December 28, 2018 by unreasonable

December 28, 2018

11 hours ago, JtD said:

We can chalk it up to margin of error even though it is a large gap, because the margin of error in this case is also very large, for instance structural damage on a B-25:

B kill probability: They are 95% confident that a hit will lead with a 1 to 17% probability to a B kill, estimated value: 6.4%.

It's obvious that with a that large confidence range basically any outcome can be covered by uncertainty of prediction. Just go with the 17% and you'll instead of an about 80% find a less than 50% survival chance for 4 hits just based on structural damage, not accounting for fires, fuel or pilot kills.

That is all true, but going with the 17% means that you expect to overestimate vs true value 97.5% of the time, which is hardly encouraging. Surely better to only expect to overestimate (or underestimate) 50% of the time? The observed rate is still the best estimate, provided that you have done your best to eliminate or adjust for known bias.

Of course these considerations also affect the German gun camera estimates, also done with tiny samples for the MK108.

Edited December 28, 2018 by unreasonable

December 28, 2018

Panthera, it is useless arguing with the guy that thinks adding single shot probabilities to 100% yields the average number of shots needed to bring down a plane...

The American study does attempt to make survival rate calculations. These calculations are based on single shot probabilities. Why do they do that? Maybe they had spare time left and wanted to produce another statistic.

If you take into account that there is not only an infinite amount of possible ways to parametrisize a single dependence (cumulative effect) between the kill-p (since the the survival never actually goes down to 0 so you have to set a convergence alas your implicit dependence changes) but also an infinite amount of possibly different covariances between single shots, you see that you end up with an infinite amount of degrees of freedom that can also be parameteisized by an infinite amount of parameters to arrive the same average of 4 hits.

Any reasonable person would thus conclude that comparing an implicit p for a kill derived from an average to a kill percentage derived for single shot data is absolute nonsense but I guess the name already says it all... This has been said so many times that it is kind of sad that this idea is still being clinched onto like Hitler believing in the Endsieg in his Führerbunker in Spring 45.

As strange as it may sound, actually recorded data on averages is the best data available if that is what you want to measure. A model that does not have any parameters vital for measuring an average amount of hits to bring down a plane is not suitable to measure this, what a surprise. This is crystal balling as hard as it gets.

I have no doubt the Führerbunker will keep up the fight as long as the Endsieg is so near.

For me pictures and actual data is the equivalent of the Red Army knocking on the bunker door.

December 28, 2018

35 minutes ago, unreasonable said:

going with the 17% means that you expect to overestimate vs true value 97.5% of the time

I understood the statistics in that the estimated value came with uncertainty. Meaning if we were to continue testing, there's still a 2.5% chance the estimate settles at 17% or above. I'm not sure if you're saying the same thing with a different wording, or if you're saying something else.

But, of course, going with the upper limit of the confidence range is quite optimistic, and the reasonable thing to do is to go with the estimated value. Additionally, it's always good to make an error effect analysis, for all studies.

Which DM do you like more - before 3.008 or after?