RNG results general discussion

[Zumwalt_446]

Hi all,

I thought I'd start a new thread to continue some of the discussions that have been going around lately on RNG and various weapon/target combinations. Some observations below:

1. It would be useful (if possible) to have a more detailed tooltip going through the elements of a result calculation (similarly to how Command MO offers both simple and detailed results in the event log). This would likely help with some of the frustrations (justified or not) with RNG results and their consequences in game.

2. It seems to me (granted, without exposure to the "back end" of the game code) that the current system does not reflect a "realistic" result in all cases. Specifically, there seems to be a lack of weight assigned to an enemy unit being spotted by the unit engaging it in the overall pK that is drawn against. What I mean by this is that, in many situations, a small unit (particularly recon) will be spotted by one of my units in near-ideal conditions (lack of cover in particular), but that unit then repeatedly fails to kill it (often with multiple vehicles in the unit engaging). This makes me think (again, without background in the development process) that something about how small units' spottability and/or vulnerability is calculated may not be yielding the correct "top-level" result.

3. A further (more abstract) discussion on RNG (again, based on some available but limited mathematical knowledge in statistics). In my interpretation, the appropriate way to implement a probability distribution for real-world (physical) cases is not to allow for the entire range (for instance, capturing the -infinity->infinity tails of a Gaussian), but rather to bound the generated result by a certain Z score/# of standard deviations from the nominal/mean value. One example of this would be using a normal distribution to model bullet dispersion from a weapon; if not bounded, a result could be generated that is simply not possible in reality (for example, a 90 degree or more off-boresight angle for the bullet path). I assume that this has been accounted for, but would appreciate any background or thoughts here.

[SgtZdog]

Hi all,

I thought I'd start a new thread to continue some of the discussions that have been going around lately on RNG and various weapon/target combinations. Some observations below:

1. It would be useful (if possible) to have a more detailed tooltip going through the elements of a result calculation (similarly to how Command MO offers both simple and detailed results in the event log). This would likely help with some of the frustrations (justified or not) with RNG results and their consequences in game.

Putting more info on them is on our wishlist of things we'd like to do. But it is a very complex algorithm, so it will probably still not be a complete explanation of how we fully got to the chance of something happening.

2. It seems to me (granted, without exposure to the "back end" of the game code) that the current system does not reflect a "realistic" result in all cases. Specifically, there seems to be a lack of weight assigned to an enemy unit being spotted by the unit engaging it in the overall pK that is drawn against. What I mean by this is that, in many situations, a small unit (particularly recon) will be spotted by one of my units in near-ideal conditions (lack of cover in particular), but that unit then repeatedly fails to kill it (often with multiple vehicles in the unit engaging). This makes me think (again, without background in the development process) that something about how small units' spottability and/or vulnerability is calculated may not be yielding the correct "top-level" result.

It would be more accurate to say that you can see a unit and we show the marker on the OpMap as a boolean True/False (let's ignore spotting level of detected/Classified/ID'd for the moment), but that does not necessarily reflect the quality of the spotting to generate accurate fire on the target. Think about how you can kinda see someone through several trees vs see them just fine standing in the open. Both of them give you enough info to know they are there, but in the woods the quality of shot is going to be terrible. Small units can be particularly difficult with this too as it takes very little cover/concealment for them to stay hidden. At 500m scale even a relatively open patch of land will still have lots of things to hide behind, small dips/rises in the terrain, etc. Next time you're driving around, try looking around and spot just how many places you can see more than 500m. We don't think about it in our normal, every day lives, but it's actually kinda rare.

3. A further (more abstract) discussion on RNG (again, based on some available but limited mathematical knowledge in statistics). In my interpretation, the appropriate way to implement a probability distribution for real-world (physical) cases is not to allow for the entire range (for instance, capturing the -infinity->infinity tails of a Gaussian), but rather to bound the generated result by a certain Z score/# of standard deviations from the nominal/mean value. One example of this would be using a normal distribution to model bullet dispersion from a weapon; if not bounded, a result could be generated that is simply not possible in reality (for example, a 90 degree or more off-boresight angle for the bullet path). I assume that this has been accounted for, but would appreciate any background or thoughts here.

For most of the distributions we do, a normal distribution is the right way to go. This is because a lot of them we're just trying to figure out a boolean Is This Good Enough. For that, we're not really fussed about the distribution of the hits in the range, but rather what percentage of the hits are good. Put another way, we don't care about whether something exceeded the threshold by 3 or by 4, just did it exceed. For that type of calculation an even distribution RNG against the percentage of values that would represent a success is fine.
Note, for things like the placement of artillery shell impacts we don't calculate the value as a roll against the percentage because that's getting into the geometry. So there are cases where we are not using a normal distribution.

[SchDerGrosse]

1. It would be useful (if possible) to have a more detailed tooltip going through the elements of a result calculation (similarly to how Command MO offers both simple and detailed results in the event log). This would likely help with some of the frustrations (justified or not) with RNG results and their consequences in game.

A very much needed feature.

In its current state, it is impossible to determine whether I am using my units effectively or not. All the game gives me is either a hit or a miss.

Am I firing from too far? Who knows? Am I hitting but not penning, who knows etc.

A CMO like detailed breakdown of "weapon endgame calculation" would be awesome.

[SgtZdog]

A very much needed feature.

In its current state, it is impossible to determine whether I am using my units effectively or not. All the game gives me is either a hit or a miss.

Am I firing from too far? Who knows? Am I hitting but not penning, who knows etc.

A CMO like detailed breakdown of "weapon endgame calculation" would be awesome.

Until we can do something like this, know that units do an evaluation of will they be able to inflict casualties by firing and will they take more damage from returning fire if they do. So in general, units are already making the is this effective check for you when deciding if they should shoot.

Whether you are firing from too far is more subjective. If we're talking a company fending off a platoon, some wasted firepower to keep the enemy at a further distance is probably worthwhile. Or a recce unit that needs that extra time to scoot before it is overrun. On the other hand if you're needing to lure an enemy in to inflict maximum damage with the limited ammo you have, you'll want to set a shorter range to use. So the range question is more down to the tactical situation than it is the relative strength of the weapon vs the armor/accuracy. (Though note that engaging beyond effective range can come with additional malus for its effectiveness of course.) And of course because of the above the unit will not start the engagement beyond its practical effect (though of course it can't account for what hasn't yet been seen).