Combat prediction simulation

[Fumbles88 10]

Hi guys,

I am currently heading up a new campaign for the unit I am involved in. Part of this is going to simulate being on 'the front line' and there will be other forces working alongside us engaging the enemy at different parts of the map (simulating the front line etc). I want to make this campaign as dynamic as possible and I don't want offensives to go down to a 'dice roll' as it will not account for weaponry, ambushes, morale etc. What we're looking at is British Forces taking on Lingorian Rebels so there would be, as I mentioned, balancing issues with a 'dice roll'.

I'm wondering if anyone has ever made a 'simulator' or a 'predictor' for firefights, ie you input how many of each side, weaponry, armour, support etc which will then simulate 100 times or so and give the most likely outcome. I've searched online and obviously my search has not come out good so I doubt that there is one in existence. Failing no one having one of these can anyone give me any kind of suggestions as to how I could go about modelling this? I have basic C, C++ and a higher end knowledge of .SQF for scripting in game.

Cheers for any help you guys can give me, it's much appreciated.

[maturin 12]

War, much less ArmA, is too complex for some things. The U.S. military does have a certain threshold by which they like to outnumber a defending force, however.

[Rydygier 1302]

Indeed. Such thing could be very useful to me. If I could, I would create such predictor. Perhaps in dreams...

When I'm thinking about... What must be done to achieve such thing? Let's try rig up some basics hastily:

1. Must be decided, with how numerical data we should represent result of firefight - numerical definition of victory/defeat;

2. All factors affecting such defined battle result must be specified;

3. All of them must be translated somehow into measurable, numerical form; for some, like terrian shape, we rather can determine only statistical approximations;

4. Must be known, how exactly each of them is affecting resulting numbers and other factors, so we could create a main formula properly (most difficult? Empirically perhaps? Oterwise perfect understanding of game mechanisms required, so you can literally calculate course of firefight step by step on numbers level - insane, so IMO will end up on emprically obtained (very) statistical approximations anyway - say "bye" to the reliability);

5. Writing the code, that will pass input data through all calculations and spit out the result.

A few points (if list is complete and proper, do not know), but expected complexicity is intimidating and results far from accurate. This is of course about predicting of Arma firefights, much, much easier than real. IMHO a much, much bigger task than dynamic campaign itself...

You could also treat Arma environment itself as prediction tool in some kind of trail and error tests. Create a test mission simulating "clinical" certain combat siutation, with cyclically spawned forces, that on chosen ground, on one of parts of planned front line terrain, in chosen configuration and other conditions will perform same firefight as many times, you like. Each time one side fulfill victory condition, or loss, "board" is cleared, and new wave of such series of tests is spawned. Results are summed and processed statistically. Manipulate with initial configuration as long, as you achieve in many tries result close enough to the desired optimum.

Do it for each particular part of front line, save for each tested spot forces configuration (initial state), that in yours tests leads to the desired result of fight, then you can for each spot dynamically spawn firefights of expected proper course with only slight randomization perhaps. Weakness of this method is spots are hardcoded (determined), unless you decide to disregard/simplify terrain shape factor influence by risky assumption, that firefight of given forces will proceed similar on similar kind of terrain or even on any kind of terrain. Another limitation - is assumed known, certain composition of both sides. It also not taking into account possible interferences between particular firefights nor unexpected participants, like players.

Edited August 25, 2013 by Rydygier

[gammadust 12]

You could also treat Arma environment itself as prediction tool in some kind of trail and error tests. Create a test mission simulating "clinical" certain combat siutation, with cyclically spawned forces, that on chosen ground, on one of parts of planned front line terrain, in chosen configuration and other conditions will perform same firefight as many times, you like. Each time one side fulfill victory condition, or loss, "board" is cleared, and new wave of such series of tests is spawned. Results are summed and processed statistically. Manipulate with initial configuration as long, as you achieve in many tries result close enough to the desired optimum.

I've been working exacly on this, since I believe we indeed require a statistical approach... i develop my ideas more in the op of this thread.

In regards to a predictor of outcomes per se, in advance of any tests, based only on some specific yet sufficiently generic initial conditions, is something that could come about only after we settle on good criteria imo. The trial and error tests could help us settle that criteria (i would call them "arma's language").