Corsair vs FW190

[delcyros]

Thanks, Bill.

I have walked through the 4thFG and will continue with 357th FG now.
It appears that not all informations are aviable (Mikes site does give about 1/6 of all 4th FG claims) but I find them very informative. After a bit fore backprocessing of the datas in spss, I got a pretty neat effectiveness curve. Most datas belong to the P-51B in the combination 4 x 0.50cal BMG with fixed gunsights. A few datapoints refer to later P-51D with K-14 gunsights (white points). I believe this graph shows quite well how significant the change to K-14 was, altough the sample is very small and requires verification on a larger base of datas.

Note that I sanity checked all narratives and therefore could not use up all information aviable.
Hope it helps. Range is either mean range (composite of opening and closing range) or range if only one figure was given in the narrative.

 

[drgondog]

Thanks, Bill.

I have walked through the 4thFG and will continue with 357th FG now.
It appears that not all informations are aviable (Mikes site does give about 1/6 of all 4th FG claims) but I find them very informative. After a bit fore backprocessing of the datas in spss, I got a pretty neat effectiveness curve. Most datas belong to the P-51B in the combination 4 x 0.50cal BMG with fixed gunsights. A few datapoints refer to later P-51D with K-14 gunsights (white points). I believe this graph shows quite well how significant the change to K-14 was, altough the sample is very small and requires verification on a larger base of datas.

Note that I sanity checked all narratives and therefore could not use up all information aviable.
Hope it helps. Range is either mean range (composite of opening and closing range) or range if only one figure was given in the narrative.

Great stuff! For what it was worth most K-14s were 51D installations.

If you want to focus on pre K-14 stick to pre September 1944 Encounter Reports. Anything after November, 1944 likely K-14.. Mixture in between those dates.

 

[HoHun]

Hi Delcyros,

>Range is either mean range (composite of opening and closing range) or range if only one figure was given in the narrative.

Using mean range (if that means arithmetic mean) would only work if effectiveness at both ends would be identical. However, your curve even as it is shows that the effectiveness drops off significantly with range, so the arithmetic mean introduces a systematic error by over-estimating effective range.

Even if the number of rounds fired would be distributed evenly over the range interval between opening and closing range, the number of hits would be weighted towards to short-range end.

Regards,

Henning (HoHun)

 

[delcyros]

Using mean range (if that means arithmetic mean) would only work if effectiveness at both ends would be identical. However, your curve even as it is shows that the effectiveness drops off significantly with range, so the arithmetic mean introduces a systematic error by over-estimating effective range.

Certainly it does introduce an error.
Unfortunately that´s the general problem in methodology. At best You have a good method and a large sample. Here I have to balance between the size of the sample and clear acceptance limits. Sometimes opening and closing range is reported, sometimes only one of both or approximate range of firing and quite often nothing at all is mentioned. I use to neglect the datas for no ranges but had to combine the other in order to keep some samplesize. I hope that widening the statistical base will help here.

 

[HoHun]

Hi Delcyros,

>At best You have a good method and a large sample.

Choosing an arithmetic mean was equivalent to adopting a hypothesis on the analyzed event. The problem is, your result is in contradiction to this hypothesis.

By choosing a different averaging method, you could avoid that contradiction.

This is not a problem of sample size.

Regards,

Henning (HoHun)

 

[drgondog]

Precison methodologies in statistics for this effort, when dealing with variable and imprecise range estimation, coupled with differential skills of the pilots and having no idea regarding the evasive manuevers of the 'target' is always 'interesting'

Delcyros - certainly illuminating even if not 'precise enough' or to the rigorous standards of others.

 

[delcyros]

Thanks Bill but Hennig is right with his critique. I just wonder what will Hennig suggest? Geometric and harmonic mean both come to mind but wouldn´t that also introduce a hypothesis not necessarely reflective for the datapool?

I rather widen out the database in order to perform a factor analysis. This should give us an idea of how developed the main relationships in this multivariable dataset are.

What I found to be interesting in the 4th FG reports is that the Fw-190 and Bf-109 don´t differ much between. As a matter of fact, there are significantly more cases known where pilots expanded very many rounds / kill for the Bf-109, which is something I didn´t expected after reading so much about the Fw-190 structural rigidity and the bf-109´s fragility, respectively.

part of the discussion was related to how effective the .50 cal BMG was. Judging from the encounter reports, it appears that the .50 cal BMG was very effective against Luftwaffe single seat fighters. It certainly doesn´t say anything about how many hits were required to down a Fw-190 / Bf-109 as we don´t know the hit probability. Pilots with excellent marksmanship skills / favourable shooting conditions were able to down a LW single seater with as few as 60-70 rounds API from close range but typically this was around 150 to 400, depending on range and condition (the average for qualified* Mikes 4th FG encounter reports give 229 rounds/kill) and could go as high as 1000+/kill.

*)qualified means that there must been a number of rounds expanded figure known. The e/a had to be engaged in firing ranges (no crash dives counts), shared kills are accounted for. Significant differences in narratives and ammo expandeture figures are excluded (f.e. 334th A.F.Buntes report dating to 29. of march 1944, claiming one Fw-190 around Magdeburg with a three sec. burst from ~150 yards. This report is inplausible wrt rounds/kill figures because his ammo return figures showed that he has completely emptied the ammo-boxes of the P-51B! 1260 rounds in 3 sec. is impossible with four 0.50cal BMG, so am convinced either his narrative is wrong (several burst not accounted for in the narrative) or incomplete (subsequent strafing of targets of opportunity not accounted for?).

 

[drgondog]

Thanks Bill but Hennig is right with his critique. I just wonder what will Hennig suggest? Geometric and harmonic mean both come to mind but wouldn´t that also introduce a hypothesis not necessarely reflective for the datapool?

I know he is right about the critique - but the datapool by the very nature of it is largely subjective except for final ammo count.

I rather widen out the database in order to perform a factor analysis. This should give us an idea of how developed the main relationships in this multivariable dataset are.

Agreed and necessary for anything except further somewhat subjective opinions - (which are still valuable)

What I found to be interesting in the 4th FG reports is that the Fw-190 and Bf-109 don´t differ much between. As a matter of fact, there are significantly more cases known where pilots expanded very many rounds / kill for the Bf-109, which is something I didn´t expected after reading so much about the Fw-190 structural rigidity and the bf-109´s fragility, respectively.

I believe this would have at least four important factors - a.) evasive action/manuevering prior to shooting, b.) skill of the pursued pilot, c.) skill and shooting ability of the pursuing pilot, and d.) the variable nature of the range as the damaged a/c slows down

part of the discussion was related to how effective the .50 cal BMG was. Judging from the encounter reports, it appears that the .50 cal BMG was very effective against Luftwaffe single seat fighters. It certainly doesn´t say anything about how many hits were required to down a Fw-190 / Bf-109 as we don´t know the hit probability. Pilots with excellent marksmanship skills / favourable shooting conditions were able to down a LW single seater with as few as 60-70 rounds API from close range but typically this was around 150 to 400, depending on range and condition (the average for qualified* Mikes 4th FG encounter reports give 229 rounds/kill) and could go as high as 1000+/kill.

*)qualified means that there must been a number of rounds expanded figure known. The e/a had to be engaged in firing ranges (no crash dives counts), shared kills are accounted for. Significant differences in narratives and ammo expandeture figures are excluded (f.e. 334th A.F.Buntes report dating to 29. of march 1944, claiming one Fw-190 around Magdeburg with a three sec. burst from ~150 yards. This report is inplausible wrt rounds/kill figures because his ammo return figures showed that he has completely emptied the ammo-boxes of the P-51B! 1260 rounds in 3 sec. is impossible with four 0.50cal BMG, so am convinced either his narrative is wrong (several burst not accounted for in the narrative) or incomplete (subsequent strafing of targets of opportunity not accounted for?).

The March 29 battle ranged from Celle to Brunswick to Uelzen to Magdeburg - it was a big day for both the 4th and 355th with perhaps 70+ Fw 190s and 30-40 Me 109s. The 4th had a mix, the 355th tangled with 190s.

Bar was in this fight and probably shot down the only 355th pilot lost that day to air combat.

I need to look but IIRC Bunte shot down his bird in the Magdeburg area and strafed targets of opportunity on the way home - which would account for the rest of his ammo.

 

[timshatz]

One variable that is unfathomable, and probably could explain the variation on the 109 kills, is the gunnery skills of any given pilot. Could be the pilots expending large numbers of rounds on their targets were just crappy shots.

Is there a variable that could be calculated that would show a confidence interval for any given engagement in terms of accuracy of the report? Such as "the average pilot, shooting at aircraft "X", from a distance of "Y", would expend "Z" number of rounds with a standard deviation of "U" rounds?"

For instance, an average Mustang Pilot (not using the K14 gunsite) firing at a 109 from 200 yards will average 247 rounds per kill with a standard deviation of 44 rounds.

Just a thought.

 

[DerAdlerIstGelandet]

First of all great stuff.

Sorry if I missed it, I have been in an out the last week or so, but does anyone know the average rounds required against P-51s by a Bf 109 or Fw 190 pilot?

 

[drgondog]

One variable that is unfathomable, and probably could explain the variation on the 109 kills, is the gunnery skills of any given pilot. Could be the pilots expending large numbers of rounds on their targets were just crappy shots.

Is there a variable that could be calculated that would show a confidence interval for any given engagement in terms of accuracy of the report? Such as "the average pilot, shooting at aircraft "X", from a distance of "Y", would expend "Z" number of rounds with a standard deviation of "U" rounds?"

For instance, an average Mustang Pilot (not using the K14 gunsite) firing at a 109 from 200 yards will average 247 rounds per kill with a standard deviation of 44 rounds.

Just a thought.

No. (Short answer)

Who and what process decides 'accuracy', 'objectivity', 'metrics' and how does one train pilots and Intelligence Officers who are writing down the pilot's narrative how to ask the right questions and then minimally assign - high confidence, medium confidence, low confidence - to the narrative?

There is no recording of radar ranging because there were no radar ranging sights or media for recording, the pilot in question 'was kinda busy'.. the pilot in question may have questionable range estimating skills (reminds of many guys I have hunted Geese with), etc, etc, etc.

If one were to get every Encounter report and had all the film available, and every witness provided a separate statement with specifics regrading range and deflection (assuming he wasn't swiveling his head for guys sneaking up on him - which was his prime responsibility..

Then you might have separate piles marked 'Probably fairly accurate', 'Speculative but middle probability of accurate' to 'Who really knows how much factual data is in this Report'

Having said all this, I am 'Statistical Analysis Aware' by required coursework but would NOT place my skills on a chargable basis to a client!

 

[KrazyKraut]

Thanks Bill but Hennig is right with his critique. I just wonder what will Hennig suggest? Geometric and harmonic mean both come to mind but wouldn´t that also introduce a hypothesis not necessarely reflective for the datapool?

Geometric mean doesn't make sense to me, as it's usually used for growth rates and such, not the metric values we are dealing with here.

Harmonic mean... no idea. It will lead to a lower kill distance for each encounter. However, it is a rather arbitrary decision.

What I found to be interesting in the 4th FG reports is that the Fw-190 and Bf-109 don´t differ much between. As a matter of fact, there are significantly more cases known where pilots expanded very many rounds / kill for the Bf-109, which is something I didn´t expected after reading so much about the Fw-190 structural rigidity and the bf-109´s fragility, respectively.

Two problems: 1. these kills seem to be mostly from 6 o clock. The real advantage of the 190 was the increased survivability against defensive fire coming in from 12 o clock. 2. You are comparing kills vs kills. The analysis shows nothing about those that got away.

part of the discussion was related to how effective the .50 cal BMG was. Judging from the encounter reports, it appears that the .50 cal BMG was very effective against Luftwaffe single seat fighters. It certainly doesn´t say anything about how many hits were required to down a Fw-190 / Bf-109 as we don´t know the hit probability.

Again I don't think your analysis says that as you have no information on how often the enemy got away. Enough rounds expended, even a 30cal will bring down a fighter. Only a comparative analysis can shed some real light into that subject. That 50 cal was effective against single engined fighters is not really in question for me anyways though.

 

[HoHun]

Hi Delcyros,

>I just wonder what will Hennig suggest? Geometric and harmonic mean both come to mind but wouldn´t that also introduce a hypothesis not necessarely reflective for the datapool?

My approach would be to choose a mean that follows the obviously parabolic function you have drawn through the data points. So the question would be, how did you arrive at that function, and can we use it for establishing a mean range that is consistent with the data?

Somewhat simpler, one might hypothesize that effectiveness of the individual rounds is little affected by range so that the predominant factor is the decrease of apparent target size with range, which is inversely proportional to the square of range. Accordingly, the mean distance would be Dm = sqrt(2)/sqrt ((1/D1^2)+(1/D2^2)).

If the number of data sets you have makes this worthwhile, you could compare the "single range" data points with the "opening/closing range" data points to see if they match each other.

>As a matter of fact, there are significantly more cases known where pilots expanded very many rounds / kill for the Bf-109, which is something I didn´t expected after reading so much about the Fw-190 structural rigidity and the bf-109´s fragility, respectively.

Hm, I guess when you say "significantly", it's actually used in the technical sense? Interesting indeed!

>Pilots with excellent marksmanship skills / favourable shooting conditions were able to down a LW single seater with as few as 60-70 rounds API from close range

This probably tends towards the "best case" because as Krazykraut pointed out, we only have data on the kills. We don't know if 100%, 50% or just 10% of the targets that were hit with the expenditure of 60 - 70 rounds API were killed, so that we'll inevitably end up with an optimistic figure. Still, a highly interesting analysis!

Regards,

Henning (HoHun)

 

[timshatz]

No. (Short answer)

Who and what process decides 'accuracy', 'objectivity', 'metrics' and how does one train pilots and Intelligence Officers who are writing down the pilot's narrative how to ask the right questions and then minimally assign - high confidence, medium confidence, low confidence - to the narrative?

There is no recording of radar ranging because there were no radar ranging sights or media for recording, the pilot in question 'was kinda busy'.. the pilot in question may have questionable range estimating skills (reminds of many guys I have hunted Geese with), etc, etc, etc.

If one were to get every Encounter report and had all the film available, and every witness provided a separate statement with specifics regrading range and deflection (assuming he wasn't swiveling his head for guys sneaking up on him - which was his prime responsibility..

Then you might have separate piles marked 'Probably fairly accurate', 'Speculative but middle probability of accurate' to 'Who really knows how much factual data is in this Report'

Having said all this, I am 'Statistical Analysis Aware' by required coursework but would NOT place my skills on a chargable basis to a client!

Some judgement call on the part of the pilot but most of it would based on the numbers of the expenditures. Problems abound on this one, no doubt. Multiple kills, random short bursts at fleeting targets, ground strafing would all affect the number that would be credit to each kill. But by massing evidence, there should be some type of pattern that reveals itself. The more data, the better your results.

However, the old "garbage in/garbage out" problem could assert itself.

But much like the scientific analysis of crash lead to more effective training, the same should be true with air to air. My thoughts would just be the beginings of the analysis.

 

[drgondog]

Some judgement call on the part of the pilot but most of it would based on the numbers of the expenditures. Problems abound on this one, no doubt. Multiple kills, random short bursts at fleeting targets, ground strafing would all affect the number that would be credit to each kill. But by massing evidence, there should be some type of pattern that reveals itself. The more data, the better your results.

However, the old "garbage in/garbage out" problem could assert itself.

But much like the scientific analysis of crash lead to more effective training, the same should be true with air to air. My thoughts would just be the beginings of the analysis.

Tim - I agree your thoughts, Delcyros positioning and HoHun points.

You did summarize the biggest difficulty from initial expression of data all the way through analysis - namely what is reliable input?

 

[timshatz]

Yeah, it would be a bear to get a decent line on the inputs. I was thinking last night that gun cameras and the combat report might be an excellent way to analyse each kill. It would give both a variable (pilot report) based on pilot observation and a variable (gun film) based on observed data. It would bring in all sorts of interesting details (pilot range to target observed vs percieved, point of aim variables, burst length observed vs percieved, ect).

Cripes, you could get half a dozen Phds just figuring out the details!

Would be pretty interesting to read too. Wonder if anybody's AF ever did it?

 

[bada]

Two problems: 1. these kills seem to be mostly from 6 o clock. The real advantage of the 190 was the increased survivability against defensive fire coming in from 12 o clock. 2. You are comparing kills vs kills. The analysis shows nothing about those that got away.

Again I don't think your analysis says that as you have no information on how often the enemy got away. Enough rounds expended, even a 30cal will bring down a fighter. Only a comparative analysis can shed some real light into that subject. That 50 cal was effective against single engined fighters is not really in question for me anyways though.

Well known RAF Report:

Shooting from 6 o'clock at 190 fuselage

 

[Glider]

Very interesting, there doesn't seem to be a massive difference between the 303 and the 0.5. Is there any clue re the range at which these results were achieved?