ConsimWorld

The official social platform for ConsimWorld

Using Dupuys 'Numbers and predictions in war'

Many years ago a friend of me, Niklas Zetterling, developed a general division scale ww2 battle system inspired by Dupuys 'Numbers and predictions in war'. Up to this day this battle system is the most realistic that I have encountered (though I regrettably don't play that much serious consim games).
In short each unit has a number of steps (usually six for a standard inf div) two attack and two defense strengths, maneuver and prepared, an operation point value, an AT-rating and a type. The turn was I go/You go with combat as part of movement (both movement and combat cost op, and displaced enemy units leaves an op marker behind that will have zoc up to that point.
Turns were 1/2 weeks and hexes 16 km.

It was a very clean system with quite few special effects and such, while combat resolution was rather complex - we usually used computer support!.

Anyway, the bottom line was that it worked surprisingly well on any tested situation, without a lot of tweaking. I was especially impressed over how it worked on the eastern front, where the scale of operations, rate of advance and even type of mistakes made, corresponded well with the historical data. It also scaled nice as long as we didn't venture to far into the tactical realm.
It was however slightly cumbersome to play Barbarossa on division/brig level with half week turns and 16km hexes....

What was used from Dupuy was mainly the method of getting a base strength of a unit by adding a calculated kill-value (well, two - mobile and static) for each single weapon in the unit, and adjusting with his national efficiency factors, adjusted for different branches (like SS and Luftwaffe field div....)

Replies to This Discussion

Permalink Reply by Robert Mårtensson on March 13, 2008 at 10:36am

Well, mainly two ways. When a unit was upgraded it reflected in the combat factors - sometimes not so much though, as bigger tanks and guns usually also meant fewer of them. But then there was an armor-class or AT-level or whatever (I don't remember what we called it), that was an abstraction of the gun vs. armor race. We only considered three cases - you are superior, same level or inferior - and the combat strength of participating armor units (and to a lesser degree semi armor units) were modified by a percentage accordingly (one reason why we needed a computer).

To simplify things, the usual modifications were already considered in the combat strength, so a German panzer division had their printed strength from the assumption that they were superior, and thus got quite a lot weaker when there were a T34 battalion among the defender. A standard soviet AT-unit would on the other hand not make any difference other than by its combat strength.

Efficiency was calculated into the armor class, so German armor units did have a class higher than the gun/armor strength would indicate, while esp. early soviet armor, and Italian of course, had a very low armor class.

Another factor was that modifications wasn't that dramatic. More around +/-25% than halving or doubling factors.

But as I hinted, the system didn't have all those special rules for different types of units (which was a bit boring from a gamers perspective). It was mainly the effect from combat strength variation in mobile or planned combat, movement cost (leg/mot/mech) and the OP (operation points) of the unit, that gave the profile of a unit. Thus, with its high combat strength in mobile combat (which was less costly in OP and losses), quick movement with mechanized movement rate, and more OP than an ordinary unit, a full strength panzer division had not only the punch to defeat an enemy unit, but the swiftness to do it more than once during a turn, which was very valuable!

The units in War in the East is very irritating indeed. The german panzers (and other countries to perhaps) is far to powerless. If I could make a quick fix to that game I would try making the German panzer div 15-8, PG 10-8 and SS 20-8, and then I would make soviet inf 2-4 and corps 6-4

There were of course quite a lot of simplifications in the armor class system. Highest class counts and it did not look at numbers, so one high class tank or AT battalion in a multi division battle was enough to get superiority.

Permalink Reply by Robert Mårtensson on March 14, 2008 at 6:00am

I would also do away with those pesky AT-brigades. I can't really see any justifications for them, other than to balance the game. The point would be to establish the panzers as the one offensive arm in the German army. So I would like them stronger and the German infantry relatively weaker on the eastern front. In France 1940 I just need the Panzers stronger, so you don't need such a lot of special rules to allow for a German break through. Might be that I would need to add a few more strength points...

Permalink Reply by Eric Walters on March 13, 2008 at 7:12pm

QJM is fine when looking over one's shoulder at past battles--it's easy enough to tweak the variables and algorithm to get the results you need. Problem is that QJM is far less useful as a forecasting device...oh dear....so for you designers looking to use it to help you with future combat, don't bother.

Anybody tried to replicate Stephen Biddle's analysis as outlined in his book, MILITARY POWER, for resolving tactical actions. That would interest me.....

Permalink Reply by Robert Mårtensson on March 14, 2008 at 5:16am

QJM = Quarterly Journal of Medicine? No, must be Quantitative Judgment Method...

What I usually am looking for is a way to handle what ifs in a historical context, in a way that could give some indication about what really would have happened. My guess is that QJM must be the way to go, but the parameters you could tweak must be general enough to be valid in both the situations from where you get your data and to the situation where you will apply it. And that can be frustrating! There are always unknowns that are hard to quantify. Like for instance how to quantify Swedish troop quality in ww2, when they never were tested in combat.

Permalink Reply by Don Chappell on March 14, 2008 at 11:16pm

This has been a good discussion, and I hope it continues. I was sorely tempted to drop what I was doing at work in order to answer.

Dupuy's book is literally within arm's reach when I'm sitting at my computer at work. But to tell the truth, it's mostly in such an "honored" position because it makes great eye wash. My copy has a white book cover with bold red letters, so it jumps out.

As most of you know, "selling" the analysis is the final critical step to performing an analysis. So when somebody comes in to consult with me, there are a bunch of books RIGHT IN THE MIDDLE of their line of sight. I intended for them to say: "I know a lot about this shit and you don't." That's not necessarily very collaborative, nor PC of me. But then, at work, the analysis isn't for entertainment purposes. I literally deal with rocket scientists and like all very smart people, they are prone to erroneously think if they're smart in one topic, then they're smart in all.

Gee, this sounds like a rant that might better go on my personal blog. (blush)

I'll come back later to comment on Dupuy, QJM, and WEI/WUV. I very much loath the method for real analysis. I'm a lot less judgmental about using it in wargames! In fact, it makes sense in certain designs, IMO.

Permalink Reply by Robert Mårtensson on March 15, 2008 at 6:36am

I think I can see a bit of what you are aiming at, partly at least (hmm, a part of a bit - that's not very much :-).
Dupuys work draws quite far reaching conclusions on rather thin data, so I guess it's kind of worthless in an academical context.
I can also see that you have to have a good knowledge on all the aspects of the involved forces, which should make it less valuable in forecasting the future.
And it would only work on a large scale, as it is a purely statistical method. All individual variation in competence etc, all freak events and other upredictables would have to be drowned in sheer numbers.
I guess there are situations better suited for this kind of analysis and situations where it is less applicable. As I said it worked like a dream on the eastern front battles we tried it on, and worked well on France as well (both 1940 and 1944), while we couldn't make it work on the Norwegian campaig without a lot of special rules (well, we didn't even try).

So, in a situation where the ultimate intentions of both sides were known at the time, and there are no important hidden cards on one side (like ULTRA), and where the scale is big enough to make individual events less important, my experience is that those statistical methods works quite well. On the other hand there were a lot of if's there...

Permalink Reply by Mircea Pauca on March 26, 2008 at 4:25pm

I got Dupuy's book and somewhat-like the results. The "sausage machine" for his QJM numbers seems a bit suspect - too much emphasis on "firepower" and too little on survivability and weapon range. (If one can always stay out of range of the enemy and keep firing with some effectiveness, the relative value should be infinite ;-)).
But the philosophically worst part of his QJM is "Surprise" ! I ran Dupuy's numbers at the back of the book through my multiple regression, and found his factors not bad, but only explanining ~20% of actual variance of outcomes. This jumps to ~60% only when including Surprise, which is purely tautologic in definition. (If one won, one successfully "surprised" the enemy !).
Then, the almost only useful question is shifted to "How do we achieve Dupuy's Surprise", that is "How we win" ? so the analysis gained exactly nothing.

Permalink Reply by Mircea Pauca on July 28, 2010 at 11:36am

Mr. Wittman: Do you mean to debate my statement above, or the previous string of debate ?

I really would like to know if you consider Dupuy's fudge factor of Surprise as methodologically valid - that is, explaining something new before a battle comes, not only for historical 'retrodiction'.

Or else, you remain with the other known hard factors with relatively weak influence (results from Dupuy), and the degree of Surprise achievable gets a probability distribution, obviously depending on preparation and leadership skill etc.
Then one gets overall... just a CRT ? ;-)

The aggregating formula for each weapon's OLI seems too firepower-oriented and too little rewarding the other facets, especially the context-dependent ones.
From history, stronger/almost-invulnerable armor *against one epoch's adversaries* like the Matilda II, KV, Tiger and Abrams when they appeared give a disproportionate jump up in relative value, which also degrades relatively fast once countermeasures are found. And it seems offense can escalate faster/cheaper than defense as technology. So, the "proper" numbers may not even be properly multiplicative and transitive !
e.g. "numbers from my belly"
1 PzIV ~ 2 T34
1 Tiger ~ 8 T34
but it does NOT follow that 1 Tiger ~ 4 PzIV against any other kind of enemy...

Also I think the terrain and defensive postures too small in Dupuy - only 1.55x (if remembering well; don't have the book here) for prepared defense in rough terrain ?

Permalink Reply by David J. Conklin on November 20, 2011 at 11:28am