ORIGINAL: wdolson
Flak only accounted for 54 planes. The 20th AF switched over to low altitude bombing in March 1945. 39 of the loses from flak were after March 1945. 17 of the 19 combined losses from aircraft and AA were after March.
I think it's obvious that large caliber Japanese flak was not a concern to the B-29s. Switching over to low altitude caused more flak losses.
Bill
I lived (until his death) near the retirement home of one of the few Naval members of the USSBS. I suggest you read the USSBS itself - and if you can get them - the supporting analysis materials. Losses to flak were as high as 4% per raid - and often caused revision (or at least consideration of revision) of changes in tactics. It appears many flak losses in your data are classified as "other" - probably because the compiler has too many "unknown" causes. We don't always know the cause - and what the cause is when we do know is sometimes a judgement call. All data can be quibbled. But the USSBS study is the best we have - and it was made in part with data provided by the other side. Data from only one side is notoriously bad - in all wars - in all eras. We long claimed (and many still believe) an 11:1 kill rate in Korea - but it was not. We claim only about 1:1 in Viet Nam - and even that is an exaggeration. That for air : air - but the point is - one sided data is never good. You can only build an integrated picture when you include data from all sources - and from crash sites.
Air defense theory is a bit counterintuitive and it is very easy to make invalid assumptions when looking at data. The date you have provided us is essentially useless and not at all indicative of what we need to build a proper model. Consider, for example, the tactics used by B-29s before the switch to low altitude bombing. The typical mission was flown at 32,000 feet. This is above the effective reach of almost all AAA systems and casualty rates do not indicate anything at all about what would be achieved if the raids were flown below the effective ceilings of the guns. More than that - and this is a surprise for nonspecialists - we AA guys call the choice to fly at that altitude a victory for the defense. Consider that as altitude increases accuracy declines - even sans winds. In this case, the winds made it utterly impossible to hit the target. Targets not hit = victories for the defense. AA is not measured in terms of planes shot down, but in terms of targets not hit. That in turn is a function of how many targets could have been hit had there been no air defense at all - so you could hit any number of targets at optimum altitude for the bombers - whatever that is for any given mission type.
It is more complicated than that as well: we cannot build a model based on gross data without technical details because - for example - if a jammer was used against enemy radars - that may have degraded AAA effectiveness - and we cannot derive the base case without knowing to what extent that is the case? But even more important are the targets not engaged at all. We need to know that they didn't target this factory - or that town - because the AAA was too well placed and we wanted to avoid the consequences. Naval air defense is easier - the ship carries the guns with it. On land, you don't know what target will be aimed at - so you may have to engage targets which are not closing (the optimum case for the defense). Looking at gross numbers tells you nothing about the base effectiveness of the very same weapons if the attackers were closing.
For a good description of the way AAA is evaluated, see Weapons and Hope by Freeman Dyson - although it is only a small part of the material in the book it is well written for laymen by a man who was intimately involved with the WWII case. For a more general history, see On Air Defense. Air defenses are used to minimize the amount of damage any given enemy force can inflict. It is a dynamic and never ending tactical game in which both sides vary what they do. Our game is remarkable insofar as it does model this to a degree: if you fly low you will pay a higher price in losses; if you fly you will inflict less damage. In a very basic sense, that is quite right. We also have some ability to influcence what happens. Messing with ceilings for guns as well as aircraft (effective ceiling for guns and operational ceilings for aircraft), gun lethality, bomb accuracy, and a number of other factors, RHS has been able to create a situation in which early and mid war raids generate about the right results for any given tactical situation. But that is new, and not the norm, and it is not clear if the late war period is properly modeled or not even with these changes?
A more complex system might not work as well. In many ways Gary Grigsby's attitude about dice rolls (lots of em) is right - and as theater or area commanders we would not control the defenses in a tactical sense - so it produces a relatively correct (if not absolutely statistically valid) outcomes. If we had more factors under our control - I for one would love it - but in manual games where complexity isn't limited - this confirs a strong advantage to the technical players over the more general ones. That might not serve the general community as well. And there is a practical limit to what can be done: that limit is ultimately a cost limit - how much programming time can we afford to invest before the product loses marketability? The answers to such questions properly belong to those with the capital - which is to say not with us.
