|
pompack -> RE: Paying Player's Poll (no testers please) (12/13/2010 2:38:49 PM)
|
quote:
ORIGINAL: PyleDriver
Ok, I'm breaking this cone of silence. What in the heck are you talking about pompack?
Just disagreeing with redmarkus4 on principle dealing with the philosophical difficulties of creating a simulation that will always produce an exact replica of history. Actually another poster in another similar thread said it better: "if you want a simulation that exactly duplicates history read a book" [:D]
As to the philosophy, Brother Herwin in the WitP forum could explain it better (and at greater length) since he used to teach this stuff and I only studied it.
When you set out to create a mathematical model of conflict you face an initial challenge in your choice of temporal and spatial granularity: at very large granularity you can create a model where you push "start" and it will tell you who won the conflict; at the other extreme you can attempt to construct a model at such fine granularity that it will inform you that on the 14th day of the conflict Private John Smith was killed by artillery fire on the outskirts of Rodevenko. Once you create your model, you must run sufficient trials to achieve some level of statistical confidence in the results. You are then inevitably faced with the fact that your predicted results deviate from the results expected by the designer and/or the customers paying for the model. When that occurs the designer has four basic choices: 1) tweak the initial conditions, 2) change the math portion of the model, 3) add external constraints to the operation of the model, and 4) change the "expected results" in the mind of the designer/customer.
#1 is used most frequently but is constrained by reality since the initial conditions of a conflict of interest are usually known, at least at a high level. Note that one of the primary drawbacks (other than cost) of a fine-grained simulation is the fact that at a low enough level the initial conditions are NOT known and given enough flexiblity to change them the designer can produce pretty much whatever result a customer desires (which in the real world means that the customer has wasted his money unless that was the purpose of creating the simulation in the first place.\
#2 is rarely done to any great extent to the the exteme cost in time and money of any but the must superfical changes in the math model
#4 is unfortunately the rarest outcome in the real world of combat simulations; customers especially don't like to be told by computers that they had it all wrong
#3 is usually the choice after changing the initital conditions (#1) fail to produce the "expected result": for example, for an Armored Cav advocate, if the Russians fail to advance far enough on the first day (thus indicating that your Cav screen is sufficient) you add a constraints that eliminate traffic friction at the border due to "prior planning", increase the effectiveness of first day bombardments due to "prior registration". The point is that you can add as many special rules as you like and pretty much create any result that you like but that is doing nothing to increase the accuracy or the "historical nature" of the simulation.
So that is a lot of words to say that I don't believe that you should change a model that produces good results over a multi-year period by constricting it with a bunch of "special rules" in order to produce an historical result in the tenth week of the campaign (for example)
|
|
|
|
