quote:

---

Originally posted by Chris Lawrence:
Academic marketability......an interesting concept.

---

quote:

---

Originally posted by yadernye:
Posted to the H-War list, April 13, 2001:

From: "Markel, Matthew W MAJ ODCSOPS" <Matthew.Markel@Hqda.army.mil>
Date: Wed, 11 Apr 2001 14:44:14 -0400

Dr. Woodford seems to argue that if Dupuy did not select his engagements
with deliberate bias, than the QJM data base is not skewed. The fact
remains that most of the engagements cited by Dupuy occurred prior to the
late summer breakout of Allied forces in France.

As to meticulous research, after trying to track the data myself, I would
counter by saying that it is almost entirely secondary research. That in
itself does not impugn the database; the U.S. Army Official Histories are
quite good. The fact remains, however, that the authors of this series
made no attempt to systematically record a consistent set of data on all
engagements described.

Wade Markel
MAJ, GS
Strategy and Policy Analyst
Army Transformation Office
(703) 695 8263

---

Major Markel unfortunately also appears to have little knowledge of the QJM Data Base and how it was created. In fact, I suspect he is confusing the CHASE/Land Warfare Data Base with the QJM Data Base. In actual fact, one of the reasons that the Italian Campaign battles were selected for analysis was because the documentation for all three combatants involved -- German, British, and American -- were available and extensive. Thus, the Italian Campaign fit the first and prime criteria for validation, data for both sides involved in the engagement was available.

And if Major Markle would be interested I could direct him to the relevent NARA and PRO document holdings that were utilized as sources for the data. We still maintain the original paper research files with copies and notes from the primary source documents used in the enegagements here in our office. If he is unable to discover those documents himself I suggest it may be because he did not know where to look. I recently had cause to review some of the Italian engagements in the course of a recent study done on enemy PW capture rates. As Chris alluded to before, I found that when estimations were done, it was inevitably with the result that although the Allied strength was occasionally somewhat overestimated, German strengths in the same engagements (especially in heavy artillery) were grossly overestimated.

Sorry, this was actually posted by me.

Rich Anderson
The Dupuy Institute
http://dupuyinstitute.org/
_____

Of the historical ilk.

quote:

---

Originally posted by yadernye:
Posted to the H-War list, April 13, 2001:

From: "Markel, Matthew W MAJ ODCSOPS"

The fact remains that most of the engagements cited by Dupuy occurred prior to the late summer breakout of Allied forces in France.

---

Well, there are a couple of reasons for this. First German record keeping after June 1944 degrades considerably. They tended to submit their war diaries every six months, and the last 6 months of 1944 are notorious for missing log books. Therefore, if one wants to get good two-sided data from WWII, you really need to do research before July 1944. This kind of rules out NW Europe.

Second, what is wrong with using the Italian Campaign? Why is it less relevent than northwest Europe? If the major's arguement is that Trevor Dupuy should have used data from North Africa, Italy (1943-45) and North West Europe (1944-45), than I agree with him. This is a much larger research task and would like to know if DSCOPS would like to fund it?

Third, we have done a little "non-QJM" research into Ardennes engagements versus Italian Campaign engagements, and later Italian Campaign engagements (post-June 1944) to earlier Italian Campaign engagements. First, the data is not radically different (NW Europe is not armor heavy while Italy is primarily infantry battles) and second, there does appear to be some change over time as the German army degrades and the US Army continues to improve. The fact that things may have changed does not invalidate Trevor Dupuy's work.

quote:

---

I would counter by saying that it is almost entirely secondary research....

---

Actually the reseach is from a mixture of primary and secondary sources. Obviously, you cannot get German strength, casualties and inventory data from the "US Army Official Histories". This data only comes from the archives. I wonder how the Major could have reached the conclusion that our German data came from "US Army Official Histories"?

quote:

---

The fact remains, however, that the authors of this series made no attempt to systematically record a consistent set of data on all engagements described.

---

Do not understand what Major Markel means by this "fact".

The WWII Italian data is from a mixture of the US and German unit records and secondary sources. Obviously, the Arab-Isreali data can only be from secondary sources as neither side has opened up thier archives at that time (nor had the USSR). Any post-June 1944 German data is also going to be a problem. So, the data is less than perfect. Still, this does not make, confirm or create the arguement that it is biased and skewed. I think John Sloan, who was involved in assembling much of this data can speak more directly to it. If Major Markel has a problem with the data, he is certianly invited to come by our offices (which are in the area) and see for himself.

I presume that you posted these replies to the H-War list as well? It will be interesting to see if they generate any _substantive_ discussion. Perhaps Major Markel will take you up on your invitation to visit TDI.

quote:

---

Originally posted by yadernye:
I presume that you posted these replies to the H-War list as well?

---

No I have not. My one experience with posting on H-War was "difficult". Maybe I should.

quote:

---

It will be interesting to see if they generate any _substantive_ discussion. Perhaps Major Markel will take you up on your invitation to visit TDI.

---

I kind of doubt it. My experience is that these critics quickly disappear when confronted. The "Trevor invented his data" rumor is one I've heard before, and it appears to come out of a certain locale in the US Army. I suspect we won't hear from Major Markel again, but he will not actually change his opinion, just be less vocal about it.

quote:

---

Did TDI develop a formula for modeling the effects of fire suppression?

---

Suppression and friction are implicite in the model. As the model is developed and calibrated from historical data, it automatically ends up with historical levels of friction and suppression. As such, unless one is in a combat situation with unusually high or low levels of suppression (or friction), one will get close to reasonable results. The question becomes, for those models that are not built from historical data, how do they model suppression and friction?

quote:

---

Originally posted by yadernye:
If you have the time and inclination, please do join the H-War discussion on this topic. You can do a far better job of rebutting and clarifying than I can.

---

Sigh, oh all right, I'll go there.

quote:

---

BTW, I was queried directly on the availability of the HERO/TDI reports. Am I correct in assuming that they are all available to the public, except those listed as restricted that deal with the mechanics of the TNDM?

---

Absolutely. They are listed under publications in our web site, and under there you can check "Books for Sale", "HERO Reports" or even the more recent "TDI Reports". All are available (about 170 reports) except for the TNDM user guides.

From: "Trent Hone" <thone@rubicontechnologies.com>
Date: Fri, 13 Apr 2001 17:59:32 -0400

On Mon, 9 Apr 2001 23:19:51 -0600, Shawn Robert Woodford wrote:
> With regard to the matter of combat effectiveness, I find it
> interesting that none appear to dispute the notion that such a
> factor exists in the first place.

I dispute that such a factor exists, or rather that such a factor can be
reduced to a definitive number. I have remained quiet through this
discussion because I am not overly familiar with Dupuy's method, but Shawn
Robert Woodford has provided the opportunity to not only criticize the
method, but the assumptions behind it.

I see an essential contradiction in Dupuy's efforts. By developing and
assigning values for a multitude of variables he is acknowledging that
combat is a complex combination of factors which cannot be easily
quantified, but then by extracting from these disparate values a single
number he is attempting to argue that combat effectiveness can be reduced
to a single value. This seems terribly inconsistent.

Combat is certainly a very complex environment. To suggest that it would
be possible to precisely identify and quantify the majority of the factors
involved is itself controversial, let alone using those using those factors
to develop equations. While I believe the development of such equations
could be an effective learning experience, I do not think they could serve
any useful purpose other than predicting _possible_ outcomes for a given
combat. There are just too many variables involved to predict combat
effectively, let alone develop a single combat effectiveness value.

Consider the systems we have in place for predicting the weather. Yes, the
variables involved in predicting the weather are arguably more numerous
than those we would use to predict combat, but the complex interaction of
the variables is quite similar; the interaction of several variables may
not have any discernable effect until a certain threshold is reached, and
at this point the effects of each are considerably intensified. This is
particularly true where C2 (command and control) or morale is involved. I
believe Dupuy’s equations are algebraic in nature, and as such it is
impossible for them to model such complex relationships. Even if they were
as complex as the weather system models available today, we are all
familiar with how accurate these are at predicting the weather.

Just as the morning's forecast is no more than a prediction, Dupuy's models
cannot be considered any more than a prediction of a possible outcome.

Elsewhere, John Sloan has argued that the "weaker" side will never win a
battle. If we believe the "weaker" side has won, then we have inadequately
analyzed the situation. I assume then that he would argue that the Bismark
and her consort, the heavy cruiser Prinz Eugen, were in fact superior to
the Prince of Wales and Hood when they encountered them in the North
Atlantic in 1941. I assume he would argue this because it is readily
accepted that the Germans "won" the engagement; they managed to escape and
also succeeded in sinking Hood in the process.

I do not believe such an argument could be sustained; the circumstances
leading to the loss of Hood were a series of events that appear to be
highly unlikely even in modern analyses of the battle. The explosion of
Hood's after magazine was, in effect, the result of a combination of
unlikely circumstances; it was a lucky hit.

By suggesting that the "weaker" side can never win, John Sloan appears to
be suggesting that there is no room for luck or chance in combat. I do not
agree; the counter example I have chosen is certainly extreme, but I
believe it effectively illustrates the difficulty involved in attempting to
model combat with mathematical precision.

quote:

---

Originally posted by Alex H:
Trent H doesn't understand what 'weaker side' means.

---

quote:

---

From: "Trent Hone"

I dispute that such a factor......can be
reduced to a definitive number.

---

Well, Combat Effectiveness Value (CEV) does exist and it can be mesaured. We have obtained measurements of it using different methodologies. Not only has Trevor Dupuy done so using his QJMA (Quantified Judgement Method of Analysis) but recently, I have done a series of exercises in our Enemy Prisoner of War reports that did a simple statistical analysis that produced similar figures and levels of performance differences (between Germans vs US, Germans vs UK, and by comparison US vs UK, German vs. Soviet, Egypt vs. Isreali, UK vs. Argentine, US vs Iraqi).

quote:

---

....but then by extracting from these disparate values a single
number....This seems terribly inconsistent.

---

The actual model was structured from 70+ variables. Some of those were easy to measure, most were difficult to measure and a number could not be measured at that time (all analysis suffers from a very basic limitation called "funding"). These factors that could not measured at that time included such things as training, morale, motivation, initiative, leadership, etc. In effect, a whole range of human factors (and some other things).

The model was constructed from the non-human factors elements (which were easier to measure) and then they were tested against a series of engagements. Obviously one never gets a perfect fit. There are four reasons for this:

1) Factors you have not considered or were not important enough to place in the model (and modeling error)
2) The factors that you have not yet measured
3) Random variations ("luck")
4) other random errors

If after you have tested the model to the real world data (meaning historical combat data), if one is getting fairly close fits.....then you probably have a good model. If you find that one side is consistently underperforming compared to the other side, both in the attack and in the defense, and across a large number of engagements (I used about 60 for my tests), then the unexplained "error" (or delta) is probably a combination of factors (morale, cohesion, training, leadership, etc.) That we sum up to be called Combat Effectiveness Value (CEV).

quote:

---

Combat is certainly a very complex environment. To suggest that it would
be possible to precisely identify and quantify the majority of the factors
involved is itself controversial.

---

Well, either we do, or we cease doing any combat modeling in the US. It is not possible to "precisely" indentify and quantify the majority of factors. Almost none of those factors is "precisely" identified, they are approximated. The key is whether the entire model can then been tested (validated) to real-world data. If the model matches real-world data closely, than it is indeed a "model". The single most important step in model building, in my opinion, is validation.

quote:

---

Consider the systems we have in place for predicting the weather.

---

Oddly enough, Trevor always used the same example. In fact, he used to explain the difference between "forecasting" and "predictions".

The rest of your arguement wanders into the "angels dancing on the head of pin area", which is not an area I live in. Quite simply.....the goal is to build a model of combat. As it is a model, the proof that we have acheived that goal is that the model can be sucessfully measured against real combat. If it produces approximately the same result (it will never match exactly)....then you have a "model." The arguements on methodology are all very interesting....but if the model fits....what is the issue? If you think a different or better methodology will work, then please use it....but the final measurement of its usuefulness is ....does it fit the data (validation).

We do not recommend using the QJMA for naval combat (unless one had a large number of participants) any more than we recommend using the QJMA for the Battle of the OK Corral. Obviously, with a small number of combatants, luck can play a major part in the outcome of the engagements. If one is measuring division-level ground combat (which is primiarly what the model is used for), then the luck tends to be far less of a factor (or shall we say, averages out).

Hopefully this answers all your concerns.

I can also see why Col. Dupuy complained about the parochialism of historians and soldiers. You all work with the same historical evidence they do, yet becuase you quantify it, it's somehow no longer valid. I could speculate on the reasons for this, but shall refrain since it is a largely fruitless exercise. Let's just say that one of the reasons I chose not to pursue an academic career was because of my perceptions of the greatly varying levels of professionalism among academic historians, a variation that seems that much more acute among military historians.

quote:

---

Originally posted by yadernye:
How can a rational debate over the QJMA/TNDM be undertaken when so many critics are not familiar with the methodology?

---

Well, becuase it appears to be a simple model, most people think they understand it. I have discovered that, in fact, most people pretty much miss the boat entirely. This is true of supporters as well as detractors.

quote:

---

Col. Dupuy clearly addressed all of these recurring criticisms in his books, which few seem to bother to read.

---

Actually, I don't think he did a very good job of expaining the reasoning and logic behind the methodlogy. This is partially because it was so obvious to him that he simply did not understand how to explain it.

quote:

---

You all work with the same historical evidence they do, yet becuase you quantify it, it's somehow no longer valid.....

---

Yea, the historians don't trust it becuase its quantified and the operations research people don't trust it because its history.

quote:

---

What sort of reception have Col. Dupuy's methodologies and interpretations received in other quarters since he published NPW?

---

Of course, this had lead to some habits that just annoy the hell out of some mathemeticians (like multiplying bore diameter in mm by the vehicle speed (kph), creating a new mathematical unit call "mm kph")

quote:

---

Originally posted by Trent H:
One thing I'm still not too clear on is this 'weaker side' business.

---

This was John Sloan's description, so I didn't address it.

The basic structure of the QMJ/TNDM comes from a quote from Clasuwitz, which says something like (I'll post the exact quote later) if one strips out all the variables and conditions of an engagement, including human factors (which is a measurable condition) than one ends up with the bare description of the engagement, which is numbers.

As such the formula for combat strength is numbers (weapon strength) times effects of the conditions (variables) times human factors. This is done for both sides. The stronger side in this equation wins. For example:

If I had an attacking force with a strength of 1.0 attacking a defending force with a strength of .8 but on good defensive terrain (say a value of 1.4), then the stronger side is the defender (1.12) and it wins.

Now the real model is much more sophisticated than that (70+ variables). To start with every weapon is scored. Furthermore, there are 20+ charts addressing various conditions (effects of terrain, weather, air superiority, length of day, defensive works, rate of advance affects on casualty rates, etc.), and these conditions affect different weapons systems in different ways. Furthermore, there are some fairly complex equations that attempt to address the interaction of some elements of the model (like the vulnerbility equation and the mobility equation) and finally there are human factors, which can often be a "force multiplier" by as much as a factor of three (or even six, see below). The final result of these calculations in the model is that the mathematically stronger side always wins, but that can be in many cases the physically smaller side (yes, we have modeled a few South African engagements, and those are wonderfully lop-sided exercises where you have a company sized infantry force attacking a brigade size defender with armor and air superiority, and winning).

For those unfamiliar with Col. Dupuy's work, I would recommend reading Understanding War first, because he lays out his general theories in detail there. The QJM is explained and presented in NPW. Col. Dupuy's analyses present a number of quite interesting and often counter-intuitive insights, of which the notion of combat effectiveness is only part.

Cheers,

If we.....strip the engagement of all the variables arising from its purpose and circumstances, and disregard the fighting value of the troops involved (which is a given quantity), we are left with the bare concept of the engagement, a shapeless battle in which the only distinguishing factor is the number of troops on either side.

These numbers, therefore, will determine victory. It is, of course, evident from the mass of abstractions I have made to reach this point that superiority of numbers in a given engagement is only one of the factors that determines victory. Superior numbers, far from contibuting everything, or even a substantial part, to victory, may actually be contributing very little, depending on the circumstances.

But superiority varies in degree. It can be two to one, or three or four to one, and so on, it can obviously reach the point where it is overwhelming.

In this sense superiority of numbers admittedly is the most important factor in the outcome of an engagement so long as it is great enough to counterbalance all other contributing circumstances. It thus follows that as many troops as possible should be brought into the engagement at the decisive point.

Whether these forces prove adequate or not, we will at least have done everything in our power. This is the first principal of strategy. In the general terms in which it is expressed here it would hold true for Greeks and Persian, for Englishmen and Mahrattas, for Frenchmen and Germans.

Cheers,

quote:

---

Originally posted by Trent H:
Which of the institute’s publications would you recommend if I were interested in discovering more about the modeling system and the philosophy behind it?

---

Actually I think Shawn covers this. There really isn't a book that gives one the wherefore and whys of the model. There is a book that tells you what can be shown with the model (Understanding War), although a lot of that book was also based upon the Land Warfare Data Base (LWDB). There is a book that decribes how the model works (NPW) and there are plenty of reports that show how the model and the data was developed (Attrition, et al), but there really is not a book that describes "why" the model was developed the way it was.

quote:

---

Have any software programs been produced which model the system?

---

The QJM was programed originally in BASIC. It was then reprogrammed in PASCAL. The TNDM is programmed in Turbo Pascal, although the rights to this program is owned by the Dupuy family and others. We do sell the rights to use the model here, but it is beyond the reach of a private individual.

Over the years, a number of people have devleoped their own spread sheet versions of the model. The latest one I am aware up was done from the Attrition handbook to come up with a method to estimate casualties for division-level medical requirements. The DOD had not been able to provide such a division planning tool for the medical officers, so one division medical officer created his own.