[Consim-l] Analysis of Coral Sea

[Mike NotSpecified]

Mircea, thanks for bringing this to our attention!  Very interesting.  I'll be 
eager to hear Markus's take on it.

For me, one section really stood out:
"The aircraft level model assumes that each aircraft's attack or interception 
is statistically independent of all other aircraft, whereas the squadron level 
model implicitly assumes that the aircraft within each squadron are perfectly 
correlated; e.g. either all of the dive bombers in a squadron are intercepted, 
or none of them are"

I have been playing a lot of games on Midway, including Smithsonian Midway, 
Victory at Midway and CV.  The first two use large groups of planes, which may 
or may not be squadrons I don't remember exactly.  CV uses a counter for every 
three planes which makes it much more interesting in my opinion.  A very real 
limitation in the first two game is you just can't run very many different 
missions, you are pretty much committed to all or nothing missions because 
each carrier may only have 4 or 5 air counters.  CV gives you a lot more 
flexibility and thus does a better job of holding my attention.  I am working 
on a comparison of all three games, hopefully in the next month or so...

A second thing that struck me was that after stripping away all the acadmeic 
language, at heart what these guys are reporting is not all that 
sophisticated.  Or maybe I should say not a lot different from some of the 
games I've been playing, which might suggest that our games are fairly 
reasonable models in their own right.  Not that I'd want to play most games 
enough times to generate a range of outcomes....

Last thing.  I built a program to emulate the combat model in Smithsonian 
Midway.  So far I've only built it to capture the interaction of the CAP with 
different ratios of escorts to bombers. At this point I can run it and say how 
many bombers make it through to attack the ships and what happened to the 
fighters on both CAP and Escort.  My eventual goal is to run it all the way 
through AA fire, bombing and results, and then make it all with enough 
variables that I can determine the likely results from different numbers of 
aircraft, ratios of fighters to bombers, composition of the task force etc.

But I built it mainly as an exercise in learning how to generate random events 
in my programming (well, that happens on its own, I mean designed random 
events within the code) and, as noted above, I've become somewhat 
dissillusioned with the squadron approach.  So it may be awhile before I have 
anything to report, coding the combat model in CV (or Flat Top, same thing) is 
a pretty big job.

Still, very nteresting article.  Thanks again for the tip!  Mike


On Thu, 16 Nov 2006 23:34:04 +0200
  "Mircea Pauca" <mircea.pauca at gmail.com> wrote:
>    Even 'official' analysts can be almost
> like gamers ;-) An article especially of interest
> for Markus Stumptner - calibrating a Salvo
> combat model to the Battle of Coral Sea:
> http://www.carleton.ca/csds/working_papers/ArmstrongWP03.pdf
>    By that theory, most of our games use
> stochastic Salvo models, right ?
> especially 'fistful of dice' systems, War at Sea,
> Victory in the Pacific etc.
> 
>    Thank you for thinking about this,
>    Mircea Pauca, Bucuresti, Romania
> _______________________________________________
> Consim-l mailing list
> Consim-l at mailman.halisp.net
> http://mailman.halisp.net/mailman/listinfo/consim-l