equivalent man power

You really want me to revisit this Rich? You know I can...

Dupuy's system is badly flawed and his conclusions are not supportable by the QJM as found in Numbers, Predictions, and War.

 

No Terry, please don't. I think we can agree to disagree. At least until we can figure out why James Taylor, Professor Emeritus of Military Operations Research at the Naval Postgraduate School, the recognized mathematical authority on Lanchester equations, never seemed to have a problem with it. And wrote the differential equations for the TNDM, which was the successor to the QJM. Nor did Paul Davis of RAND. Of course, his doctorate was in chemical physics, so he must not have had a math or statistics background, right? Yes, many mathematicians and statisticians abhorred Trevor's messy maths, but at the same time many didn't. I don't have a math background myself, so will leave it to the authorities to argue. :circlejerk:

Anyway, it's odd despite your core circularity argument being nearly as old as the QJM it never seemed to engender much excitement in the OR world, which instead focused on the so-called methodological errors commented on by Sheldrake. Which also end up being mostly specious misunderstandings or simple typos.

Cheers!

 

You're welcome. Nice to be back. You can thank Poppy among others for e-mailing me and asking I return.

 

I didn't know you had gone here Terry, but I have only been posting here regularly for a few months. Welcome back.

 

No salutes left to hand out, so I'll throw you one like this;

SALUTE!!!!

Welcome back.

https://www.youtube.com/watch?v=GDm3qr3Zlu0

 

Yeah, Poppy is de man! Go Poppy, go Poppy, go Poppy go!

I'm glad you're back, I greatly admire your posts.

 

Hmm, I am not sure 10 AOK was mainly mechanised formations.

Perhaps I have missed something, but, I wasn't sure the analysis included the 34 Div v 44th Infantry Div or the FEC v 5th Mountain Div, the 65 Inf Div being overrun by the V British Corps or the 71st and 94th infantry divisions v the II US Corps and FEC in May 1944.

 

It was, at least until mid October 1943 when 65. ID and 305. ID began taking over parts of the line. Before that the only non-"mechanised" formation was 1. FJD.

Yes, I'm afraid you have missed something. This sort of analysis requires reasonably accurate daily casualty and inventory data for both sides, which is simply not there for many of the German formations after November 1943 in Italy. Which is also why with few exceptions so many of the Italian engagements in NPW are from that period.

 

Sorry but this does not make sense. The campaign started in September 1943 and continued until May 1945, but we are going to ignore all results after October 1943!

 

That you don't understand doesn't mean it doesn't make sense. Are you saying that instead of "ignoring results" you would rather wish to make up nonexistent data to draw "conclusions from? The Italian engagements cover the period when data is available for 1943, but also for the period at Anzio February. We added additional engagements from the ACSDB and found a shift in fall-winter 1944 in those cases tending to favor the American forces in the Ardennes slightly more than the Italian data, but it wasn't a significant shift. We also added additional cases after Trevor died, including some minor revisions to the Italian data, and nothing changed significantly. For some dozens of engagements at Kursk, in Normandy, and at Aachen. The last I checked the total data set for World War II engagements was approaching double that of the original database, without significant changes.

 

Not quite sure what we are arguing about here.

You have just agreed that only some engagements were included in the model - i.e. those for which data existed. If the data is more available for engagements fought be German mechanised units this still makes the study biased in the direction argued by John Buckley Terry Copp et al.

 

Sorry, but part of the problem seems to be that everyone assumes the QJMA as related in NPW was the be all and end all of it. It wasn't. The QJMA was pencil and paper and calculator with most of the initial factors based upon "analytical assumptions". The next iteration was the QJM, which added engagements from Trevor's studies of the Arab Israeli wars and additional World War II engagements. It was also a self-contained computer program and incorporated actual findings (as opposed to "gut instinct") on advance rates, effects of terrain and fortifications, and the like, based on some dozens of intervening studies done by HERO/DMSi for DoD and DA. The next was the TNDM, which was completely reprogrammed in Pascal, with the original simple "look up table" personnel and equipment attrition calculations recast as differential equations by Dr. James Taylor of the NPS and the other factors further refined by additional studies. It also included additional engagements culled from the Land Warfare Database. It was essentially a 25-year iterative effort from circa 1969 to Trevor's death in 1995. Accumulate data. Test in the model. Refine, Accumulate more data. And et cetera. Since 1995, the then roughly 120 engagements forming the basis of the database (already expanded from the original "80"), was expanded further, including additional engagements from Italy, Normandy, the Westwall, the Ardennes, and the Ostfront. After I left TDI in 2008 I understand that Chris Lawrence expanded the dtabase even further with many additional engagements from Kursk. The last count I recall was over 200. That is still small compared to the tens of thousands of engagements of World War II, but it remains the most robust engagements sample ever collected.

The second thing that is problematic is the continual belief that German "mechanized units" were somehow inherently superior to other units. Stronger, with more select personnel, and with better weaponry. Unfortunately, the belief is very difficult to validate. It appears to be intuitively correct, but our intuitive sense of correlations is defective in the extreme. 3. Panzergrenadier Division in Italy is probably one of the best examples. When committed to battle as a "mechanized unit" in the fall of 1943 it proved to be an extremely problematic formation. It was filled with Volksdeutsche III personnel who chose to desert at a moments notice. Just a few months in, the division commander was bewailing its lack of ability. 26. Panzer Division is another example. When committed it had virtually no tanks, so essentially was a weak motorized infantry division. 16. SS (brigade and then division) is yet another. "Elite SS?" Hardly. 15. Panzergrenadier? A hodge-podge of units and personnel thrown together, mauled in Sicily, and then hastily thrown together again for Salerno. 16. Panzer? Yet another division recreated from bits and pieces as recreated "Stalingrad divisions" with about six months of existence before it was committed with a questionable Panzeregiment organization. And so on. Then, there is the simple fact that German non-mechanized units are included in the data base and simply do not vary much from the norm. The same is found in the Ostfront engagements.

As an aside, I was tasked to re-evaluate the data for accuracy back around 2005-ish when we were working to expand the overall TDI engagements database. What I found, consistently, was a pattern of error. The error was usually based upon a misreading of original German sources, which led incorrect assumptions about German organizational strengths that in turn led to frequent underestimations of German strength vis a vis Allied strengths in the engagements.

Perhaps more amusingly, I have noticed over the years that historians appear to focus their criticism on what they see as "statistical errors" while mathematicians and statisticians usually concentrate on what they perceive as "historical errors" in the methodology. The second thing is those who winge so much about how "small" the database is, how "unrepresentative" it is, and the lack, have done exactly jack shit to expand the knowledge base on their own. I understand why, it takes time and effort to do the work to assemble a valid engagement, time which is much better used in working the academic cocktail circuit, writing articles based upon whatever book they could grab off the shelf while keeping their bum in their comfy office chair, and the like (and yes, I do not think Buckely, Copp, Daglish, et al really do so, but too many other do). I understand, but also think both says more about the historians, mathematicians, and statisticians and their bias than it says about any actual faults in the methodology.

 

I agree with that last Rich. Most historians and people I've read who commented on NP&W tended to focus on the OLI part of the system. I start by focusing on the results formulae given on pg. 61.

That is, the formulae (R-R)/5 = (P/P)-1 and P/P = (R-R) /5+1

Setting either to 0... P(R-R)/6P = 0 for example will result in a correct answer only by chance. Neither equation is an equality.

The CEV equation is a variant of the above. PR/PR = (R-R)/5+1. Again, set to zero it won't give a correct answer except by chance.

I also have problems with the graph on pg. 60. The plot of data suggests an exponential fit for the data meaning that the outcome equation should be some sort of exponential equation but the results formulae are all linear equations (ignoring that they are really inequalities too).

My other big problem with the QJM is that it is static in nature. That is, it doesn't account for time and distance in its design. I usually use the battle of Crecy as an example of why this is a major problem. If you use the QJM with time and distance factored in the outcome reflects the actual outcome, a major English victory. With Dupuy's static model the predicted result is diametrically opposed to the actual one.

Dupuy does mention time and space on pg. 38 but sluffs it off as "...it is neither possible nor necessary to express directly..." Well, I say it is and the above battle, as but one of many demonstrates it absolutely.

I also have problems with things like the table on pgs. 52 and 53. Three battles are presented in detail showing, purportedly, how the QJM works. Unfortunately, you can't reproduce the results given because the tables are incomplete and lots of data is missing that would be necessary to work all of the system. For example, WTF did the numbers for the entry "Training" come from?

The OLI on its own can be jiggered around considerably. Accurate? Maybe, maybe not. But it isn't the problem unlike most historians seem to focus on.

 

Good to see you back!

The units in Italy had large number of post Stalingrad or Tunisia "rebuilds" of units that had surrendered. So rather "non standard" compared to other German 1943 formations that at least kept a structured original core when rebuilt. I would add 90th PzG, and 14th, 16th and 24th Panzer, to 26Pz and 15 and 3PzG, so it's not the "first team".
BTW we also had the SS panzer corps rebuilding there after Kursk but it never was committed to the front line.

 

Very true, except the last. It was just 1. SS-Panzergrenadier Division rebuilding and it was way up in Northern Italy.

 

Terry, I hate to be rude, but frankly at this point I could care less that you don't like the maths. For one thing you are arguing about the earliest iteration of a methodology that went through a number of major iterative changes over the course of about 20 years. It is a bit like arguing against the tank in 1944 because of the defects of Medium Mark I.

For another, if you want to argue the validity of the maths I have already suggested who you should argue with. James Taylor isn't accessible, but Chris Lawrence is. Have fun.

Cheers!