RHS Maneuverability Review: Data [ALL Data Done]

[Drongo]

ORIGINAL: Nicholas Bell

Here is another run. This is using RHS aircraft. All devices removed from both aircraft and units.................

I can't comment on what occurs with RHS but I was unable to achieve your results by applying the stated changes in stock.

I grabbed one of the small stock scenarios and cleared all devices from the aircraft and air groups entries for the air units involved in a quick test.

This is the result.
Day Air attack on Dobodura , at 55,91

Japanese aircraft
A6M2 Zero x 60

Allied aircraft
P-40E Warhawk x 60
B-17E Fortress x 58

No Japanese losses

No Allied losses

Aircraft Attacking:
54 x B-17E Fortress bombing at 15000 feet
4 x B-17E Fortress bombing at 15000 feet

Lots of reports of attacks during the combat replay but no actual firing and no hits.

Cheers

[Nicholas Bell]

>Nic is hijacking this thread - and is NOT contributing to developing a better formula or criteria to define maneuverability. I object. He is, in fact, trying to say "don't bother - it cannot be worth the effort." Explicitly so. Well - he is twice wrong: it does matter - and it is something I have been asked to do; and he is wrong to interfere with the process on principle even if it didn't matter.

How about I post some test runs with radically different manueverability ratings to show you how little difference it makes? Would that be okay? All your calibration calculations may result in manuever ratings shifting a couple of points, which has little or no impact in the current game engine. Why not spend your time and energy pushing those programmers to get the code fixed and released ? [:)] Then address calibrating the model once you know what the model is. IMO you're putting the cart before the horse. That is of course your right to do.

I also have the right to express my opinion regarding the data and the program, do I not?

and it is something I have been asked to do

Pray tell, by whom?

[Nicholas Bell]

I was unable to achieve your results by applying the stated changes in stock.

Well, I must be doing something wrong and will have to recheck my test. Thanks.

[Mike Scholl]

"Nic is hijacking this thread - and is NOT contributing to developing a better formula or criteria to define maneuverability. I object. He is, in fact, trying to say "don't bother - it cannot be worth the effort." Explicitly so. Well - he is twice wrong: it does matter - and it is something I have been asked to do; and he is wrong to interfere with the process on principle even if it didn't matter. "


You're being a bit tough on Nic here, CID. If there is any truth to his data (and he seems to have reproduced the results several times), then you may be fighting a battle you can't win without going "outside the box". That would be valuable information which would save you a lot of wasted effort. I think he's trying to be helpfull, and at the same time hoping he's wrong. He's just pretty depressed about the results he's gotten...., and we've all been there a time or two.

[el cid again]

Nic: I am a libertarian - and on principle believe in freedom of speech/etc.
But I don't think it is proper to prevent us from doing a task in a dedicated thread. IF you have something to contribute - contribute.
But don't go on at great length about why it cannot matter - particularly after I show why you are incorrect about that.

As for advocating code reform, I began by doing that. We got the critical change - a decision to support and develop WITP. We got real money committed to do that - and recruits from our own ranks as programmers - people dedicated to that reform. This has well and truly begun and already produced meaningful results. I see no point in not giving them the time it takes to do more, nor for being negative and assuming they won't do it. We have a posting by David Heath himself about the plan - so this is not tea leaf reading. More than that, the problem of getting this routine right is two sided: without good data it isn't practical to calibrate anything. My instructions are "you fix the data and I will fix the code."

I believe you are brilliant. I believe you could be constructive. Suppose for a moment that we will get the critical code reform we want: can we also have good data for it to use ready to go? We can read speed, weight, ROC, etc in books. But we cannot read maneuverability. We have only one field. How SHOULD it be used?

[el cid again]

ORIGINAL: Mike Scholl

"Nic is hijacking this thread - and is NOT contributing to developing a better formula or criteria to define maneuverability. I object. He is, in fact, trying to say "don't bother - it cannot be worth the effort." Explicitly so. Well - he is twice wrong: it does matter - and it is something I have been asked to do; and he is wrong to interfere with the process on principle even if it didn't matter. "


You're being a bit tough on Nic here, CID. If there is any truth to his data (and he seems to have reproduced the results several times), then you may be fighting a battle you can't win without going "outside the box". That would be valuable information which would save you a lot of wasted effort. I think he's trying to be helpfull, and at the same time hoping he's wrong. He's just pretty depressed about the results he's gotten...., and we've all been there a time or two.

Yes we have. And it may be you are right. Nic has at all times been civil - a big deal to me. He has obviously wasted his time reading this thread and running special tests for us - asked or not. So that implies he is committed to setting us strait for some reason. So I am inclined to agree with you. I do sympathize with his depression. But there is more to the story. I have tried to say in broad terms that there are things happening - exciting and constructive things - and since it is not practical to let outsiders inside the propriatory code world - we need to trust our friends who are inside will do well what we all want to see done. We need to do what we can do.

Don't worry - this is going to happen. It is happening. And we might be able to make it better.

[herwin]

ORIGINAL: Nicholas Bell

Mike, you're random thoughts are spot on.  El Cid & Herwin, your desire to mathematically calculate manueverability values is commendable.

Hate to be pessimistic, but it's all meaningless in game terms because the air combat formula is broken, nerfed, borked or what ever you want to call.  No matter how good the data is, it's going to spit out junk.  It's just a numbers game - who's got more planes - with the differentials between manueverability and speed, etc modifying this. 

Please try setting the all the aircraft values to 0 and see what happens.  Then set them all to 10.  Then try speed at 0 and manueverbility at 30.  Ensure all pilots are set at the same experience level (99).

Understand better how the engine works (or doesn't) first, then consider how to standardize the values, is what I am saying.  I've wasted too many hours playing with the values attempting to find some basis to work from for at least somewhat historical results.  What you do with your time is of course your business, but I hate to see you spend so much effort on something which will have so little results.

As long as pilot experience is modeled reasonably, the rest of the numbers can probably be tweaked to get realistic results.

[herwin]

Land combat: operational model. Note that my model is generally consistent with the OCS model, which is the only modern/WWII board game system that actually is realistic IMHO as a retired tactical analyst.

Morse and Kimball indicate that the exchange rate (US losses/Japanese losses) between US and Japanese fighters in 1943-44 was approximately independent of the size of the forces involved in each engagement. The percentage of Japanese fighters lost per engagement was independent of the numbers involved--i.e., Japanese losses were proportional to the size of the Japanese force. The percentage of US fighters lost increased with more Japanese fighters and decreased with more US fighters, so US losses were proportional to the size of the Japanese force. The exchange rate during 1943-44 was about 10 Japanese fighters lost per US fighter lost. A difference of 40 knots in airspeed resulted in an exchange rate of 2-1, all else being equal. This analysis was the starting point for my analysis.

[el cid again]

Note I also use OCS -

[herwin]

ORIGINAL: el cid again

Note I also use OCS -

The model you referred to was not statistically valid--it had more free variables than data points. I did a similar model that *was* valid--a much larger underlying data set and a much smaller set of free variables--and used it develop that CRT I posted.

[Nicholas Bell]

More than that, the problem of getting this routine right is two sided: without good data it isn't practical to calibrate anything. My instructions are "you fix the data and I will fix the code."

Then I stand corrected. That is quite a breath of fresh air if they are going to take your data and work the code to get realistic results. I apologize for ranting, and will get off my high horse. I admit to being very frustrated with my inability to alter the data to get what I consider historically acceptable loss rates in the current engine (I would like to be able to play this game now, as opposed to some unknown distance future date). That frustration was inappropriately directed at your work.

You might consider the information provided in Alfred Price's books "The Hardest Day" and "Target: Berlin". In summary, in both major air battles the approximate loss rate was 1/6 of the enemy sortie rate. Both books describe in detail high sortie/high loss missions, and 1/6 of enemy sorties between fairly equal forces would equate to "average best possible result." My studies of other missions with actual sortie & loss rates (as opposed to claims) indicate the most combats result in less than 1/6 E.A./friendly destroyed.

(Of course, in the Pacific the forces engaged where hardly ever "equal" in even rough terms. We however have the advantage of creating equal opposing forces by ensuring pilot training and fatigue levels are identical. We can even have the opposing forces use identical aircraft in tests.)

Regarding the game value manueverability, it would be better if the game engine allowed for offensive and defensive manueverability ratings. While the code change is not likely to happen, might I suggest that in your creation of a formula for manueverabilty, that you consider breaking it down this way and then weighing each to reach a single value.

As long as pilot experience is modeled reasonably, the rest of the numbers can probably be tweaked to get realistic results.

True in small engagements using the current game engine. There is of course the well known major problem with engagements over 50 a/c ( which El Cid has confirmed with the programmers). I have also observed instances of extreme statistical aberration, probably caused multiple bad die rolls. This may have to do with the randomizer, but in any case it can skew even small combats out of the realm of plausible historical results.

[el cid again]

ORIGINAL: herwin

ORIGINAL: el cid again

Note I also use OCS -

The model you referred to was not statistically valid--it had more free variables than data points. I did a similar model that *was* valid--a much larger underlying data set and a much smaller set of free variables--and used it develop that CRT I posted.

You refer here to one field only - the maneuverability field - which this thread is meant to work on improving - if we ever get back to doing that.
RHS does not strive for perfection - it strives to do something quick better than we had. Then to do something better still. Integrated with changes in the code and changes in general understanding of how it works. We want to get something better - but within the limits of what we can do at any given time. It seemed better to start with the design than with something purely theoretical - and I was guided by a programmer in my methodology - one who knows more than either of us about how WITP works. If you want a better algorithm - help me define it.

[Nikademus]

Then I stand corrected. That is quite a breath of fresh air if they are going to take your data and work the code to get realistic results.

Cid is not part of the developer team.

[el cid again]

What is the point of this remark?

In a technical sense, it is certainly true. No one said otherwise either. Yet in a broader sense it may actually be false. I certainly have been surprised at the degree of interaction and cooperation which has occurred in the last year - and it may indeed be that CHS and RHS are part of the Matrix developer process in some semiofficial sense. The coordinators of both have been given unsolicited technical help - and official responses to even suspected techincal problems seem to be virtually instantaneous.
Other forms of communication have also occurred - details of which we are told not to disclose. At a very minimum we feel valued by, and supported by, Matrix, specifically in the quest for a better WITP product line.

I do not see how this remark helps us define a better maneuverability rating? Nor do I see how it helps anyone do anything better? It is beyond dispute formally true - and it may be less than accurate in a broader sense. Why say such a thing at all?

And can we please get back on topic - or is that for some reason not an option?

[Nikademus]

I do not see how this remark helps us define a better maneuverability rating? Nor do I see how it helps anyone do anything better? It is beyond dispute formally true - and it may be less than accurate in a broader sense. Why say such a thing at all?

Are you even capable of speaking simply and concisely?

The 'remark' was not meant to address the topic of maneuverability. It was meant to clarify to Mr. Bell the question you avoided answering. You stated and it is something I have been asked to do and he asked Pray tell, by whom?

You are free, as is every other owner of WitP to mod to your heart's content. What you do with your mod is your own business. To continually imply however that your goals somehow/maybe/otherwise/hint-hint are connected with Matrix are erronious and misleading. As such, i provided clarification. You are not under contract to "fix" any data.

Other forms of communication have also occurred - details of which we are told not to disclose

Yet, if this were true, you nevertheless continue to feel compelled to tell people that such and such has occured with a wink and a nudge whenever someone questions a part of your reasoning or something involving your mod. When someone is told not to disclose a communication, it means don't mention it.....at all

[el cid again]

Are you even capable of speaking simply and concisely?

Most people familiar with me would say "no." Or even "NO!!!"

The technical answer is "yes" - but only with great effort - and
careful planning. The web is not the best way to see me do that -
I write quickly whatever pops into my mind.

[el cid again]

Quote
The 'remark' was not meant to address the topic of maneuverability. It was meant to clarify to Mr. Bell the question you avoided answering. You stated and it is something I have been asked to do and he asked Pray tell, by whom?

REPLY: Since I have agreed not to disclose the answer - and since I keep my promises - I studiously avoid answering questions outside the bounds of what I agree to.

I am, however, impressed with your diligent attention to detail. And also with your picture!

[Nikademus]

REPLY: Since I have agreed not to disclose the answer - and since I keep my promises - I studiously avoid answering questions outside the bounds of what I agree to.

How very convenient.

[ChezDaJez]

El Cid,

This flight data may help you with your modeling. It is based on the same type data (wing loading and others) that you intend to use. It goes into a fair amount of detail into how it was determined and what parameters were used.

The link is: WWII Fighter Data

Take a good look through the entire site. There is a lot of good data there. The formulas used are listed also.

A few others:

Speed Graph

Climb Graph

Roll Rate Graph

Acceleration Graph

Chez

[Iron Duke]

Hi cid

Have you looked at 'America's Hundred Thousand' it has quite a bit of info on the tech side of the a/c's covered in the book including graphs showing roll rates at various speeds.