Chance to hit (accuracy) - does not sense in 3.1 (?)

[EvilGuru]

No. No. No more graphs! Please make it stop!

I have to agree with Per on this one: the short/long range accuracy distinction is not really helpful. I also doubt whether any real battle has been won/lost on account of the short/optimum/long range option. That is not to say that accuracy could not do with reworking -- it could. However, the ability to change the firing range of units should not be part of any future system.

Regards, Freddie.

[aubergine]

I find that setting the long range option for firing range makes VTOLs and certain other units far more effective when I am sending them on suicide missions.

[NoQ]

However, the ability to change the firing range of units should not be part of any future system.

+9000.

On the other hand, there was a curious idea somewhere to make this button rely on a unit's being assigned to commander. That is, so that only commanders could make units fire at long distance or retreat on yellow etc. This could even bring commanders back to multiplayer, probably, though will certainly be some pain (:

[aubergine]

I think all units should have the retreat modes and fire modes on them. But agree that range mode might be movable to commanders (except for VTOLs - I almost always have my VTOLs set to fire at max range so at least they get missiles fired before being shot down).

Related ponderance: viewtopic.php?f=42&t=8975

[Jorzi]

I guess having a realistic accuracy calculation would screw up the balance a bit by favouring light bodies (although that would be an interesting change) but maybe we should have different accuracy calculations for artillery and direct fire weapons.

i.e.
For direct fire, the accuracy would simply be defined as the hit chance at optimum range, defining an inverse quadratic function (y=ax^(-2)) with accuracy capped at 100 % at range 1 and less (this is the same function as ray density from a spherical emitter so it should be realistic). This is pretty much what Searge-Major suggested and it won't be affected by the size of the body.

For artillery, we should control the spread instead of just the hit chance. For this I suggest a gaussian distribution where standard deviation radius is calculated using (1 - accuracy)*distance to target

[Searge-Major]

I guess having a realistic accuracy calculation would screw up the balance a bit by favouring light bodies (although that would be an interesting change) but maybe we should have different accuracy calculations for artillery and direct fire weapons.

i.e.
For direct fire, the accuracy would simply be defined as the hit chance at optimum range, defining an inverse quadratic function (y=ax^(-2)) with accuracy capped at 100 % at range 1 and less (this is the same function as ray density from a spherical emitter so it should be realistic). This is pretty much what Searge-Major suggested and it won't be affected by the size of the body.

For artillery, we should control the spread instead of just the hit chance. For this I suggest a gaussian distribution where standard deviation radius is calculated using (1 - accuracy)*distance to target

That's almost exactly what I was trying to say; accuracy is not dependent on the size of the body. Thanks Jorzi.

I think your artillery function should do the trick, but your function for direct fire accuracy gives a really high rate of decay at close range, and a low rate of decay at optimum and maximum range.

The function I gave, 1-(x^2)*(1-accuracy)*2, gives a very low rate of decay at min range, with the accuracy reaching the specified value at 0.7071 of max range, and decaying faster as range approaches maximum. It's not the best, so perhaps someone can improve upon it?

[NoQ]

Shouldn't direct-fire accuracy be just a gaussian random deviation from the shooting angle (left/right), with its parameter being the variance of this deviation?

[Searge-Major]

Shouldn't direct-fire accuracy be just a gaussian random deviation from the shooting angle (left/right), with its parameter being the variance of this deviation?

Like this?

Guassian Probability Density function.png (1.24 KiB) Viewed 6082 times

Errrm.... ahem.... *coughs*... well, I err...

From what little I know about Gaussian Distribution, I would say that it would provide a fairly accurate model. The function would probably have sigma^2 = weapon accuracy, but as I say, I don't know enough about Gaussian functions to say. And, I wouldn't have a clue as to which would be easier to implement in the game, though I assume a random number generator would have to be used, with the probability derived from either function. Umm... can't think of anything else to add... and I'm starting to forget what the function actually does... and anything math related...

*exceeds brain capacity*

*waits for someone who knows more about maths to come along...*

[NoQ]

Yeah, sort of.

x here is the direction at which we actually shoot, μ is the correct direction we should have shoot at, and σ² is the accuracy (which is no longer a percentage).

Emulating gaussian random numbers is something standard, afaik.

This would be a realistic behaviour as long as we believe (lack of) accuracy of a single shot to be a sum of many little independent errors.

With indirect weapons ... well, we could do the same, or we could do something two-dimensional (they occasionally fire too high or too low) (the same could be applied to direct weapons too, some bullets flying above tanks, while others hit the ground too early)

[Searge-Major]

No. No. No more graphs! Please make it stop!

What's wrong with graphs? I mean, sure, if they're not popular, then I won't post any more, but I gave the function, so you can test to make sure I didn't skew anything.

*waits for someone who knows more about maths to come along...*

Now that I think more about it, I think that the process would be as follows:

Grab point in 3D space. (WZ has height, and there are Vtols to be considered as well)
This is the target point.

Provided the distance from the weapon to target point is equal or less than maximum range, continue with targeting function.

Use the target function (whichever is best or actually works ) to give two angles, one from the direction north (whether that is x, y, or z in wz coordinates), and one from a line between the target and weapon to the vertical axis.

Vary the angle according to the accuracy function (again, whichever is best) to provide a miss angle based on the range and the weapon accuracy, so that a target at minimum range has virtually no chance of being missed, and even then only by a minute amount, whereas a target at maximum range... you get the picture?

who will multiply the HP with the size of the pie ? To fit the stats with your new reality ?

Iluvalar, not sure if this part of your post was in answer to my post or not, but with the functions being discussed, the size of the target has absolutely no bearing on the end numbers. Provided collision testing is accurate, and not based on the exact centre of buildings and droids (I'm fairly certain it isn't ), a larger droid, and especially buildings, have a much greater chance of being hit, just because of their size, it is not function related. Hope that's cleared any misunderstanding, and sorry if I've misunderstood.

[Searge-Major]

Yeah, sort of.

x here is the direction at which we actually shoot, μ is the correct direction we should have shoot at, and σ² is the accuracy (which is no longer a percentage).

Thanks for clarifying. Gaussian distributions was something I never really covered.

Emulating gaussian random numbers is something standard, afaik.

That's good.

This would be a realistic behaviour as long as we believe (lack of) accuracy of a single shot to be a sum of many little independent errors.

With indirect weapons ... well, we could do the same, or we could do something two-dimensional (they occasionally fire too high or too low) (the same could be applied to direct weapons too, some bullets flying above tanks, while others hit the ground too early)

Yeah, I was assuming inaccuracy to be the result of (simulated) human error, any discrepancies in the weapon itself and the projectile, etc.

Also, just to add, my previous graphs were graphing the actual accuracy of a given weapon over it's range given as a percentage, and were not really capable of calculating accuracy in themselves.

[Jorzi]

Well, the reason I suggested a different accuracy system for direct fire is specifically so that size/shape of the model would not affect hit probability, which I suppose is what iluvalar meant.

Hit probablity is essentially a function of accuracy (i.e. deviation in the starting speed/direction), range and target size. The challenge is to make the percentage work in an intuitive way. (i.e. 80% accuracy means 80% hit chance on any target at standard range)

[Iluvalar]

I'll try to explain again, please focus with me :

The actual stats of the games are enjoyable, not perfect but kind of balanced. You are talking about making the borgs 12x harder to hit compared to a python. It will not be fun anymore . It will be realistic but boring. Someone will obviously have to drastically change data for HP of all bodies. So borgs have 40hp and python now have 1000hp to make the game work close to as it is right now. And no players will be able to understand what is resistant or not anymore.

[Jorzi]

Yeah, I agree that if we want realistic hit mechanics, they should have been there from the beginning so that the game would have been balanced accordingly.

[Reg312]

iluvalar, again you suggest to sacrifice game mechanics and fun of game to "balance"
where you got "12 times" ? all formulas are adjustable
no one tries to find "ideal" or "universal" formula of accuracy

Again i'm repeating: "balance" can be easily redone, player should understand what game becomes better (realism is thing that like by all players, except Iluvalar )
i see bunch of ways how problem with "balance" can be avoided
btw, i did not suggested to change so much, only formula of missed shots

if accuracy formula would taken into account size of body, even in that case all can be adjusted
i think its OK if cyborg have higher chances to avoid projectile

i'm agreed with Jorzi

so we have many people who want change WZ by various ways and developers who afraid to change something
i think all should be agreed with that: in WZ should be changed mechanism of missed shots

My list of suggestions:
1) Change mechanism of missed shots (simple)
- at least range of missed shot should be larger than area of weapon splash damage
- can be used any formula (that does no sense for me, i think that all formulas are ok)
2) Remade subsystem of accuracy (its is complex and can be suspended for a time)