Hornblower wrote:Japan's pilots in '41 did have one thing working for them that the USAAF and USN did not, and that was battle experience. Even if it was against the Chinese. That's my point.
I understand, Hornblower. My feedback to you is your point is mistaken, no matter how intuitively correct you feel it is.
Ask any military man and he will agree that a combat bloodied unit is better then one- no matter how well trained- that hasn't seen combat.
Well, perhaps so, but only with this qualifier attached ". . . other things being equal." That's not perfectly correct, either, as I'll try to illustrate below in more detail.
Our own Civil War will give countless examples of that. Seeing ones comrades dying from a mistake is a harsher lesson then getting yelled at by your drill sergeant. I'm not saying that these Zero pilots were supermen. What I am saying is that they had experience under there belts, whereas we did not. Once we did then we had the upperhand on them.
In a way, in a way not.
What Japanese pilots had under their belts was flying experience in combat versus extremely subpar opposition often flying laughably outmoded planes; as I tried to note earlier this combat experience didn't necessarily make the Japanese better, probably is the primary reason they did not do better versus the Americans when they finally met.
The game has to account for the success of these pilots in early '42 in some way.
Another false belief. The Japanese had virtually no success versus top-drawer competition (1942) save for the statistical aberration which was Pearl Harbor--and even there the "success" was debatable in several ways.
One is to credit the pilots over the plane- which i think is correct.
Considering the quality of Japanese aircraft then yes, if some superiority did in fact exist it must have been with their pilots and not the planes these people drove. But you are still incorrect regarding superiority of Japanese pilots over their American counterparts.
(again this is only Fighters I am referring to) Or to increase the prowess of the Zero over the Allied planes. Which is wrong or else the P-38's, F4U's and F6F's wont have the impact that they did.
Of course they would, or at least should if the model works properly.
But the fact remains that one must account for the effect that the A6M's had early on.
Against whom? Americans began shooting these planes out of the skies as soon as we met them in combat.
In my mind its the Pilots- all be it a small elite group. As they die off, and we developed tactics (thatch weave comes to mind) then the tables turned. This isn't to say that ALL Zero pilots are this way, but that small group who did what they did are, and need to be accounted for.
It was after the introduction of better tactics (e.g., the "Thach weave") that Allied pilots began to immediately improve on that 1:1 kill ratio. When it no longer became necessary to use unsuitable equipment for certain tasks (say, P-39s in the interceptor role) the kill ratio tilted more steeply yet in favor of the good guys. When our fliers were no longer faced with nightly bombardments from sea and air and were no longer sleeping fitfully, if at all, in mosquito-infested mud that kill ratio improved even more. When better aircraft arrived it again got better for us. And so on.
Your basic premise of Japanese superiority in pilot training and experience is mistaken, Hornblower, that's the glitch. It's the same story with Gary's model.
There never was any superiority of Japanese pilots over their American counterparts. There is not a shred of evidence to suggest this no matter how it might appear.
You make the assumption that combat experience necessarily makes green soldiers better. I say this isn't necessarily so. In fact I say that combat experience can actually make green soldiers
worse. And this is probably just what happened to the Japanese pilots.
Of course I can't prove that. For all I know they stayed the same or in fact
did get better from their experience in China, but then if that were the case we must have started off with a bad premise of our own and in fact Japanese pilot training before the war was less effective than we normally give credit for and it was Japanese experience in combat which helped to ameliorate this newly-theorized deficiency in training. Yes?
One thing is for sure. When the Japanese met the Americans the fight became even immediately.
We know this from the kill ratios after just a few battles (enough to form a workable sample). We also know that Americans in early 1942 were not flying
better equipment than the Japanese. Our aircraft were better than Japanese aircraft in some respects, worse in others. So let's call the hardware even-steven.
Then how might we account for the kill ratios?
Well, that must be due to the pilots. Yes?
And what was that kill ratio?
Basically 1:1.
Ergo, some way somehow the pilots must have been pretty much equal (
Ability * Experience) characteristically speaking. Given that formula, if you maintain a Japanese superiority in experience then it must be that Americans were innately better suited to fly planes (higher ability score) than the Japanese.
Are you prepared to make that assumption? If so, why?
Let's get back to actual kill ratios.
These were a hair better than 1:1 for the Japanese through the first two or three months of Watchtower figuring just fighter-vs-fighter action, then started to get better, slowly but surely, for the Allies. This might be accounted for by the fact that early on the Allies out of Henderson Field were often flying aircraft in roles ill-suited to their characteristics: for instance, use of the P-39 as an interceptor as mentioned above. Then, as more suitable equipment became available for Aliied use, the Japanese "advantage" began to fade.
(One ace the Allied pilots did have up their sleeves and which is not modeled correctly--actually this isn't modeled whatsoever and suggests yet more problems with the simulation--is that their fighters didn't have to worry much about time-over-target restrictions whereas the Japanese "Zeroes" coming down from Rabaul did. I'm not convinced this directly impacted the dogfight kill ratio, if it affected it at all, but it certainly had an adverse effect on Japanese efforts to control the air space over Guadacanal.)
No matter how you slice it, there are no statistics to suggest Japanese fighter pilots had any significant edge whatsoever, based on their training and/or prior war experience, over their American counterparts in the early going
that I'm aware of. It might just be possible to make the opposite case, but that's not why I'm here.
Models
In the most basic terms, a model of history must use historically representative values or it's bound to give ahistoric results. My contention is Gary's model does just that. The reason must be his model is full of ahistoric, or historically unrepresentative values.
What is reality?
Well, a model's reality is what its builder says reality is.
How is this reality expressed? Gary's model uses numerical expression.
Let's build an ideal model of our own.
As the builder of our hypothetical model I say "model reality" should be expressed as a value of
1.0.
When we mix our "model realities" together (let these values interact with one another to express other more complex "realities"--in order to model real-world events this is necessary) we'll need to multiply and divide and square these values at various junctures, always with our goal to stay at a resultant model "reality score" of
1.0.
A result less than that will represent something short of reality, should we come up with a score greater than
1.0 we'll have achieved a sort of superreality (in theory perhaps a heady thing to some, but I don't know).
If all our model functions work with values of
1.0 the project's a snap. No matter how many times we multiply and divide and square the value of
1.0 the end result remains unchanged:
1.0. However, if we substitute "reality values" to our functions of less than that or greater than that (if we feed our model mistaken premise) then our resultant model "reality score" for whatever event we run will not be
1.0 and thus off to whatever extent indicated.
So, let's now build a model of WWII in the Pacific.
You say Japanese pilots were superior to American pilots.
The data before us suggests they were not, rather pretty much equal; then American pilots became better (for whatever reason), or equipment upgrades took effect, similar reverse conditions overtook the Japanese, or some combination thereof.
So, based on that, if we at first rate Japanese/American pilots/planes the same we're confident the model is working with "reality values" very close to
1.0. (If Japanese pilots and planes are rated equally with their enemies' then the result of our "multiplication dogfights" must surely come out near the historic early kill ratios we can account for: 1:1.)
If we rate Japanese pilots higher than their enemies then our model must be fed a Japanese-pilot "reality value" of more than
1.0. Let's call this new "reality value"
1.1.
Now our model has a problem, though. When we try to multiply this value with other good values the rules of mathematics will no longer allow our model to give us a "reality result" of
1.0 but rather a value greater than that. If the rest of the model is still faithful to history, and the only value at variance is the one for Japanese pilots, and assuming we multiply as our function, our model activity would look like
1 * 1.1 with the result
1.1. This would throw our intended "reality score" off by some 10% with regards to kill ratio. Plug another false supervalue into that same function and the error will be greater still in favor of the Japanese. The same holds true should we underrate American pilots and planes, only in the other direction.
This is always what happens when you feed a model false premise. As long as false premise remains the model is compromised.
We could offset this false premise of Japanese pilot superiority (or American pilot inferiority) with another equally false premise and still arrive at a "reality score" of close to
1.0 (keep those kill ratios in line with history) by plugging in a functional value of, say,
.9 to represent some imagined Japanese deficiency in an effort to diminish effectiveness of Japanese pilots/planes, but the result wouldn't be any more "accurate" (in the sense we're after, to model real-world events), it would merely provide us with a "reality score" that
looked better on its face. Our model would still be in error, indeed, more confused still.
Two wrongs do not make a right, unless it's in math, where two wrongs could make a "right" yet be wrong nevertheless.
Assigning correct values to models is not easy, I don't claim it is. My math skills certainly are not accelerated, might be worse than yours. But I know in theory if reality equals
1.0 and you multiply that by
1.1 or
.9 the result will not be representative of reality.
False values plague all models whenever they're used so every effort must be made to ensure the function values we do assign our models are as close to the "reality value" of
1.0 as possible. More false values lead to more error.
History tells us what the "reality values" for Japanese and American pilots ought to be in early 1942 with regard to overall ability--very close to
1.0 for each. We know this because the kill ratios are straightforward and they come out around 1:1. It is not our model's purpose to generate corrupt results by assigning unrealistically high scores to Japanese pilots or unrealistically low scores to American pilots. Should we, then we'd need to feed our model yet more false values somewhere within its workings in order to compensate for the historic kill ratios which are kinds of absolute references for our model's targeted "reality result." Per the example above, with many false values assigned soon our model would become confused to the point it would neither be able to give results convincing on face nor in a believable enough manner to suspend disbelief for the knowledgeable user.
We would have a bad model.
As for wives: I'm forever wrong; my wife keeps a typed list.
