If you want the most accurate lists of air combat claims see the books by Frank Olynyk. While John Foreman published the series RAF Fighter Command Victory Claims of World War Two. There are various tabulations of claims by aircraft type in WWII, but little on how they were calculated.
John Foreman's idea of combat losses, year / Hurricane / Spitfire / Fighter Claims (monthly figures)
1939 / 8 / 0 / 29
1940 / 1,002 / 531 / 4,127
1941 / 294 / 595 / 1,384
1942 / 76 / 710 / 723
1943 / 14 / 469 / 1,066
1944 / 3 / 797 / 1,783
1945 / 0 / 386 / 732
Total / 1,397 / 3,488 (total 4,885 out of 7,523 aircraft losses) / 9,844
It's interesting here that 1943 fighter losses were much lower than 1944, despite an apparent completed switch (in 1943) to mostly Spitfire activity judging from the distribution of losses.
The 1945 losses are almost as high for
four months in 1945 as for 1943 for the entire year!!! (or about 7 months if you include the Pacific).
The ratio of kills to victory is also very constant at just over 2:1.
To me this indicates the later models of the Spitfire did not improve survivability despite a declining Luftwaffe. Or that the later models took a lot of ground attack losses.
I do not find it hard to believe the Hurricane out-scored the Spitfire, though I have no opinion either way.
The "F4U vs Zero" Osprey book had an end chapter that analysed encounters were
only F4Us and A6Ms were present, and then tabulated actual failed to return losses from
both sides in the areas of the respective involved squadrons... This is the only reliable way to get anything anything even resembling an accurate picture.
The result for the entire first year of the Corsair introduction, again not for all combats, but for all encounters were it is
certain no other types were present (again, the blood sample analysis: A large enough sample becomes representative for the whole exchange ratio between these two types), covering a fairly large number and a long period.
It turns out that an air to air exchange ratio of 1:1 was the typical average of that period for those two types, in those combat were other types can be excluded... Note USN claims for the F4U are of an 11:1 ratio (all types combined)...
The way they are probaby wrong is that a lot of axis air to air kills are counted as flak kills or unknown, or accidents. Also, a huge amount of Allied air to air kills are likely vaporware. The same applies to some dubious Axis claims, particularly Eric Hartmann. Using failed to return data from the Soviets, his actual kills went as low as 80 in a "Fana de l'Aviation" article (and he went down some 12-15 times, mainly from hitting the debris of his victims)... While I don't think his kills are that low, I would not be surprised if he was far under 200, or even just above 100.
It goes to show the uncertainty of this data is high. I vaguely recall reading that British kill ratio data was of a more stringent sort, which is probably why it is so low...
As to the P-51 vs P-47 debate, one set of data I memorised from another French article was that the P-47 scored 140 of 220 Big Week kills, the P-38 and P-51 sharing 80 in an unknown ratio.
The data presented in this thread was like this:
During Big Week, with 1/5 the sorties, the -51B destroyed 64.5, the P-47 destroyed 78. The P-38 destroyed 10. (This totals 152.5, a long way from 220)
Similar to the Corsair/A6M correlation compilation, from both sides, a failed to return correlation from both sides for the Luftwaffe vs the Allies in the Normandy theater yielded, again, a 1:1 ratio for Luftwaffe fighters, despite overwhelming Allied superiority. I don't know if Allied bomber kills were included in that ratio.
It does seem the Luftwaffe did very, very poorly at high altitudes compared to low altitudes, but beyond that I would not trust kill ratios that have no failed to return data from both sides, along with the specific encounter and the specific types and specific units involved. The way to get an idea of the whole is to look at the most precise encounter possible, as defined by location and units and the failed to return at the relevant time data, these in numbers sufficient so that these individual parts gain, cumulatively, the size of a ballpark representative sample.
