equivalent man power

There is a possibility that that particular selection criteria is biasing in and of itself. I.e. "better" formations kept "better" records. Not much can be done about it at this point though. Indeed if you are setting up an experiment you can take that sort of thing into account. I don't think it's worthwhile "setting up" a war though just to improve the statistics.

From what I understand the model is doing a pretty good job of predicting results in recent conflicts which is a pretty critical criteria for judging a model. Not to distract the conversation but look at how bad all the current climatological models are doing and all the weight given them.

 

Not really, because it wasn't a case of "better" units keeping "better" records. The problem is the records needed are simply non-existent for all units. Basically, between 10. Armee and the corps and divisional records, it is possible to track daily losses for most units through the end of the year in 1943 with a number of holes. However, afterwards only dekade reports are available and in April even those are missing, apparently destroyed in the retreat during May and June when most of the corps and divisional daily records were destroyed as well. The problem with the second half of 1944 then is the same problem that bedevils all German records for the period; the SOP for all German units was to retire records to Potsdam on a 6-month basis, in January and July. Which means just as the records were getting there in January 1945 the whole thing was falling apart. In theory the Potsdam archives were evacuated to the last capital of the Reich - Flensburg Denmark - in March and April, but the reality is only most of those already in Potsdam ever made it out, except for fragments.

 

In that case it may not be that the quality of the unit is biasing the available data but the phase of the battle /condition of the front is. It sounds like the period that is not well covered is the one that would produce the most losses (general retreat). Now that may make the study even more useful for it's intended purpose i.e. planning. You want to make sure that you have enough to win the battle and taking into account the performance of units after the battle is decided shouldn't be the primary concern. However for those who try to misuse the study by siting it as a general quality reference for German vs Soviet vs Western Allied troops it's a bit more problematic.

 

The system works better with static or near static attrition battles than ones of maneuver and free wheeling action. Dupuy originally noted this problem under the heading "Quick wins."

 

Actually, it works best in single day actions between divisions...maneuver is best defined by the scenario setup and the outcome. "Static or near static attrition battles" is pretty much meaningless as is "maneuver and free wheeling action". One more or less always includes some element of the other. The idea they are two different things is pretty much poppycock. Desert Storm was a free wheeling maneuver action, but it involved gobs load of attrition; it was just one-sided. The German crossing and breakout from the Meuse bridgeheads was essentially a static or near static attrition battle until it broke loose, as did Alamein and Normandy.

 

I'd be interested in what the model might predict for 1) A D Day assaulting brigade group/RCT and 2) a WW1 divisional assault.- unless date fro m these has been input.

 

Again, since the database the model was created from were for the most part all "divisional-size" in scale, the attrition scales at the regimental level or smaller have never been "fine-tuned". The other problem is some of the amphibious assaults were extremely high intesity for very short periods of time. which skews things in a model where the attrition scales are built to an "average" intensity level (it is possible in the TNDM to adjust the intensity, but even then it is for the results in the bell curve rather than the outliers. However, for a demonstration I did a TNDM run of the 36th Division landings at Salerno, which got pretty good results.

 

Arguably the day's actions on each of the seaborne beaches were divisional size, either with one or two Brigade Groups/ RCT up.

But the latter is a tantalising statement - do go on.

 

Er, yes, of course, but you are mixing my replies to two different questions, sorry if i was unclear.

You asked: "what the model might predict for...[a] D Day assaulting brigade group/RCT", which is a difficult question since the model is validated to what were effectively a set of "divisional" engagements and even the "regimental" engagements were rather large in terms of personnel - c. 5,000-7,500 IIRC. Given attrition rates vary markedly for divisions, brigades/regiments/battalions/companies it is difficult to get it work well.

OTOH, your question also asked about "D Day assaulting", implying you were asking about modeling amphibious assaults. Which was why I remarked on the experiment with the 36th ID landings at Salerno. Which was just an experiment - a one off. Unfortunately it was done about 15 years ago and I no longer have access to the TDI files, so can only go by memory. I set it up as a phased engagement, employing the ability of the TNDM to run fractional day engagements (which has its own problems given the database is by day rather than by hour) to try to replicate the Salerno landings. It worked, generating reasonably close matches on casualties for the US forces, but IIRC the advance rate was too high (endemic to the TNDM even when run as an amphibious/large river crossing).