RAF Fighter Gunnery Analysis

Spending too much time in old documents and spreadsheets - I made a graph some might find interesting.

Based on ORS reports on the examination of hundreds Spitfire gun camera results.

red = Spring 1942
blue = Summer 1944

Solid line is the percentage of combats
Broken line is the percentage of combats that resulted in a destroyed enemy (trend line)
Points connected by the thin line are the specific data points for the above trend line



So - for example:

250 yards mean closing range was 10% of combats in spring '42 and 20% of combats in summer '44.
At this range 12% of combats resulted in a kill in spring '42 - while in summer '44 this had risen to 48%.

I have taught a lot of statistics classes. Are your data points aggregates of data? The individuals cases (each gun combat record) would give a lot more points. Using averages or medians as data points is better than not having any data but a data set of several hundred individual cases would be interesting. However, I no longer have access to JMP or SAS and analyzing a lot of individual data by hand would take some time. You can do it with EXCEL but I always avoid doing stats with EXCEL when possible. The data indicate that the closer the more probably a kill but the difference between the results from 42 and 44 could result in a number of hypotheses.

They are aggregates. The ORS (Operational Research Section) reports grouped the closing ranges of combats into the points above. More granularity would be nice but - as you say, this is a lot better than nothing.

The increase in lethality from mid-42 to mid-44 was quite surprising to me. Though it certainly makes sense when looking at the firing ranges.

One explanation would be that by 1944, the German pilots just entering service were so inexperienced that they froze instead of breaking/jinking etc. I have seen an analysis recently of the probability of being killed for German fighter pilots in WW2. Using number of missions as the X variable (horizontal axis) and the probability of being killed as the Y (vertical axis), the plot was basically an inverse exponential function. That is, it was like a reverse J for the first 50 or so missions and then it just rode along the ordinate. In other words, the probability of being shot down or killed was very high on the first missions, decreased rapidly to around 50 missions, and then stayed low. Someone with 500 plus combat missions was a hard target. Look at how many of the highest scoring German Aces were victims of operational accidents on the 262. Perhaps, there were some Japanese who flew as many missions but they were so overwhelmed by the quality of U.S. (and British) planes, they were hopeless. I know there are some accounts of experienced Japanese pilots who did will in the Frank and Rex but they were very few and far between. Also, the best Japanese planes had big problems with engine reliability.

My thoughts:

Better aircraft

• Spitfire IX is certainly a lot better than the Spitfire V in terms of speed, climb and acceleration vs. the 109/190 - allowing for much closer closing ranges and favourable angles

Better armament

• better 20-mm fuzes
• spread harmonisation pattern
• 20-mm ammunition magazines doubled
• SAP/I ammunition
• possibly less cannon stoppages

Better pilots

• RAF training hours went up as Luftwaffe hours went down
• More accurate range estimation, or should I say, less tendency to underestimate range (this was pointed out by the ORS analysis)
• More determination/emphasis in training/etc. on closing in as close as possible when shooting

Also better gun sights. This is related to better range estimation but gun sight improvements are probably worth highlighting separately as a mechanism for enhancing fighter pilot performance.

The GM2 sight used in the Spitfire would have changed very little from spring 42 to summer 44. And I excluded gyro gun sight data from the '44 figures in order to compare like with like.

GM2 vs. GGS is a whoolle different can o' worms.

OldParts said:

I have taught a lot of statistics classes. Are your data points aggregates of data? The individuals cases (each gun combat record) would give a lot more points. Using averages or medians as data points is better than not having any data but a data set of several hundred individual cases would be interesting. However, I no longer have access to JMP or SAS and analyzing a lot of individual data by hand would take some time. You can do it with EXCEL but I always avoid doing stats with EXCEL when possible. The data indicate that the closer the more probably a kill but the difference between the results from 42 and 44 could result in a number of hypotheses.

Click to expand...

Check out the SAS University Edition which is available free for non-commercial use. It is server based so it looks a little different from Display Manager but has almost all the functionality.