In depth - scatter and how it works

I think the scatter area equation is slightly off... Ive got:

((distance + (1 + scatter_offset) * distance_scatter_max)^2 - (distance - (1 - scatter_offset) * distance_scatter_max)^2) * pi * (scatter_angle/360)

You missed a minus I think...

What I DON'T know, does the offset offset ALL the scatter? or just the probabilistic center? If so, higher offset weapons can be used more reliably due to shot clustering.

I think the scatter area equation is slightly off... Ive got:

((distance + (1 + scatter_offset) * distance_scatter_max)^2 - (distance - (1 - scatter_offset) * distance_scatter_max)^2) * pi * (scatter_angle/360)

You missed a minus I think...

What I DON'T know, does the offset offset ALL the scatter? or just the probabilistic center? If so, higher offset weapons can be used more reliably due to shot clustering.

AFAIK from testing the whole scatter cone is offset and the probability density remains the same. for example, an offset of .25 would result in the long scatter cone to be 25% larger than the short scatter cone, but also 25% more shots would scatter long than short. thus, the average number of impacts / area should be constant.

I think the scatter area equation is slightly off... Ive got:

((distance + (1 + scatter_offset) * distance_scatter_max)^2 - (distance - (1 - scatter_offset) * distance_scatter_max)^2) * pi * (scatter_angle/360)

You missed a minus I think...

What I DON'T know, does the offset offset ALL the scatter? or just the probabilistic center? If so, higher offset weapons can be used more reliably due to shot clustering.

Yes you're right about this one. Minus slipped there. Corrected in the OP, thanks for pointing out!

So what advice do u have for us if u claim we've been attack grounding wrong?

So what advice do u have for us if u claim we've been attack grounding wrong?

It was already included in the post. He said that you want to be attack-grounding at max range behind the target.

That said, I was under the impression that around half of the top level players knew to attack ground behind the target. Given air shots and the possible vertical component that was mentioned in this thread, I assume targeting max range many not always be ideal though, especially for vehicles with a smaller hitbox.

That said, I was under the impression that around half of the top level players knew to attack ground behind the target.

True, many people already know this is the optimal way to attack ground in most situations. It is intuitive, since that way you are reducing the chance to come short but not really increasing the chance to overshoot because the projectile can't normally pass over the target.

Nevertheless, interesting read, well done.

Easy to read and very informative, well done. The attack ground advice is also very useful.

The ground attack thingy had to be a bit catchy to keep people interested. Many usually skip these technical posts so you have to offer them something interesting they don't know.

I have seen VonIvan and Luvnest often directly target a vehicle in the smoke instead of a spot behind it, sometimes they also do it correctly. It is possible that they actually know how to ground target properly but just did not remember it under stress, but I would assume that almost nobody knew how far you have to aim behind the vehicle. Still, for some top level players it might be useful information. And even more so for the rest of us 99% that just populate the low ranks.

Thank you for the excellent guide and the suggestion!

Question: Does this mean that attack ground on heavy vehicles is better at wiping infantry especially against infantry that is in cover. Cover will give them a 0.5 accuracy modifier so if they are close, it seems like they will hit more accurately even with scatter.

ABSOLUTELY hug enemies if you can. will massively shrink the scatter. If you are on top of them, each shell will slaughter, no need for attack ground (the area box is so small that the offset is irrelevant).

Tanks already miss infantry with about 100% odds, they ONLY hit on scatter. The offset means their shells can impact the people direct or their cover. Might be worth figuring where to attack ground for most value, but I doubt attack ground has the accuracy of scattered shots...

Thank you for the excellent guide and the suggestion!

Question: Does this mean that attack ground on heavy vehicles is better at wiping infantry especially against infantry that is in cover. Cover will give them a 0.5 accuracy modifier so if they are close, it seems like they will hit more accurately even with scatter.

That should not be the case, but also not make a huge difference.
Infantry has a target size of ~1, which translates to a hit chance of 4-5%. So >95% of your shots will be calculated as scatter shots anyway.

The only situation where ground targeting would make sense is if your tank targets one model that is lagging behind to the rest of the squad.

I think there is some confusion what accuracy actually is: When a shot is fired, the attacker has to make a accuracy role first. If he succeeds, his shot will always hit (unless it is blocked by world objects). If not, the shot will be calculated as scatter but can still hit if the scatter was low. You won't see any difference between those shots though.

EDIT: Wait, I'm not sure if I completely understood your question. If you were asking whether you should close in or not: Yes, close in if it is safe. Angle scatter will decrease at all times, while the horizontal scatter will only be decreased if you close in further than distance_scatter_max

I was just wondering if attack ground would be better than right-clicking to attack infantry. The way you put it, seems like right-clicking is still the go-to method.

I was just wondering if attack ground would be better than right-clicking to attack infantry. The way you put it, seems like right-clicking is still the go-to method.

Yes, do the right click. You might be able to max it a bit if you attack ground exactly the middle of the squad, but I highly doubt it is worth the additional micro.

Yes, do the right click. You might be able to max it a bit if you attack ground exactly the middle of the squad, but I highly doubt it is worth the additional micro.

Unless we are talking about units like the Brummbar or Bulldozer.

Those are basically howitzers. Treat their auto attack like a 1-shot Barrage, not as an actual auto attack. (I know you know elchino7, just in-case anyone new reads this)

Yes, do the right click. You might be able to max it a bit if you attack ground exactly the middle of the squad, but I highly doubt it is worth the additional micro.

Hi here, I'm going to do a bit of necro posting to reply to this post. But since the topic of interesting for everyone I guess its fine.

So I have been testing it a lot, I mean every time I built a tank in my last 50-100 games and obviously I can't sort out stats but share some feeling about it.

- Yes to do the right click works with tanks. I always do the right click now, not that I put the tank on hold fire but I'll always supervise myself the shot when I'm on it with better result than let it shot itself.

- Since i've got more experience with the sherman what I'll say may be not true for all tanks. For some reason I don't understand the sherman almost always overshot and almost never undershot (undershot are usually the result of ground colision). I don't know if it is particular to the sherman but I tend to think, from observation, that it is the same for all tanks.

- So in term of microing the sherman, I do supervise the shot with a right-click but also always undershot, I never click on the targeted squad center, always closer. And even more with a vet0 sherman, once the tank reach vet2 I'll target the closest model's feet.

And as I say I don't have stats, but from observation I tend to get twice kills as if I let the sherman doing the shot alone. This is evidently more true while the tank is unvetted, once vet2 usually I can start microing it less.

Now maybe I'm wrong, my sherman having more kills could also be the result of overmicroing it thus better avoiding losing it in battle. And that's true that I manage to keep my shermans much longer but I'm also getting it veted much faster.

So here it is, my recommendation at least when using a sherman is to always undershot by some meters to compensate its habit to overshot. Since I do that I usually get around 30/40 kills per game with it.

Unless you have the source code and are looking at the defined functions for the accuracy or whatever it would be called, you can't really know how things behave.

I have noticed that MGs suppress much worse if slightly behind some large cover like a car. You can notice it by looking at all the bullets that hit the large object in front. Even though some say that small arms fire does not have any collision data and does not behave like tank shells. I've also noticed that KT pintle MG in spearhead, and MGs in buildings suppress almost instantly, even against green cover. Confirmed it on Across the Rheine by building sandbags, putting 4 man IS behind them perfectly, and using that large top-left building (below the top fuel) to shoot at the Infantry section. They got suppressed in one volley and I noticed that the MG was positioned on the 2nd floor (so the tracers could easily be seen going directly onto the IS, rather than the sandbags).

So nobody except Relic really knows. Hence why you have bugs like JP4 shooting across the map, killing an ambulance.
You can open the attribute editor and check all the variables that are defined somewhere inside the source code, but unless you see the functions themselves, you will never completely know. You can only assume and deduce.

Show Spoiler

What you write makes sense from the mechanistic point of view.
Almost every vehicle - by design - has the tendency to overshoot. That is the 'ghost point' I described. I can't recall the exact numbers, but it should be in the range of 1-1.5 meters for most mediums. On average, that is. Shots might still be shorter than that. So the best option is to aim those 1-1.5 meters short of the squad center. This will maximize your chance of dealing maximum damage.
I just doubt that it is really worth it with the exception of a few vehicles like the Brummbar and maybe heavies. 99% of players (myself included) don't have the micro to max out the damage of a normal medium tank. I am better off investing that elsewhere, capping a point or building cover etc. The tank will find a decent solution on autofire. Those additional 5% damage or what it will be does not compensate me forgetting squads that I could micro instead.

I don't see a reason why the Sherman should be special in its mechanics compared with other mediums. I assume this is probably an observation bias that stems from the HE shell. I think you'd need to sink in tons of hours to >really< get the correct feeling of manual targeting being that much better than autotargeting with medium tanks. The difference is not that large to begin with, and everything becomes even more noisy with non flat terrain and rubble/world objects around, as you already mentioned.

Unless you have the source code and are looking at the defined functions for the accuracy or whatever it would be called, you can't really know how things behave.

I have noticed that MGs suppress much worse if slightly behind some large cover like a car. You can notice it by looking at all the bullets that hit the large object in front. Even though some say that small arms fire does not have any collision data and does not behave like tank shells. I've also noticed that KT pintle MG in spearhead, and MGs in buildings suppress almost instantly, even against green cover. Confirmed it on Across the Rheine by building sandbags, putting 4 man IS behind them perfectly, and using that large top-left building (below the top fuel) to shoot at the Infantry section. They got suppressed in one volley and I noticed that the MG was positioned on the 2nd floor (so the tracers could easily be seen going directly onto the IS, rather than the sandbags).

So nobody except Relic really knows. Hence why you have bugs like JP4 shooting across the map, killing an ambulance.
You can open the attribute editor and check all the variables that are defined somewhere inside the source code, but unless you see the functions themselves, you will never completely know. You can only assume and deduce.

I can't confirm what you say about MGs.
What happens though is that Relic had to find a solution to the height dimension. What I described assumes a 2D map, which is likely the basis also for Relics calculations. Still, they have to spawn the shot somewhere in the barrel of the tank and move it down to the ground while dealing with their glitchy engine. We can regularly see shots disappear in the ground and other stuff.
What you describe for the JP4 can happen to all tanks. I've seen it with the SU85, Panther, P4 and others as well. My best guess is that Relic calculates the (2D) shot similar to my model and based on that calculates where the barrel etc must be pointing to.
I don't know where those long ass overshots come from. Rare glitches or different calculations fighting it out and presenting an odd solution. It is hard to tell. But casemates will be slightly more prone to tjat due to their higher first shot inaccuracy.

I can't confirm what you say about MGs.
What happens though is that Relic had to find a solution to the height dimension. What I described assumes a 2D map, which is likely the basis also for Relics calculations. Still, they have to spawn the shot somewhere in the barrel of the tank and move it down to the ground while dealing with their glitchy engine. We can regularly see shots disappear in the ground and other stuff.
What you describe for the JP4 can happen to all tanks. I've seen it with the SU85, Panther, P4 and others as well. My best guess is that Relic calculates the (2D) shot similar to my model and based on that calculates where the barrel etc must be pointing to.
I don't know where those long ass overshots come from. Rare glitches or different calculations fighting it out and presenting an odd solution. It is hard to tell. But casemates will be slightly more prone to tjat due to their higher first shot inaccuracy.

Which matrices and which calculations it's taking, only God and some ultra senior Relic employer know. 2D platform which only has simple height calculations for the graphics. Or if the 3rd dimension actually plays a role. Or whatever. Nobody can tell. Your assumptions are logical and human-like. Seems like something a human would program. Still, I've also seen shots disappearing into the deep space on a map like Hill 400. Even confirmed it in the replay by rotating the camera into a cinematic angle. Pershing fired a shot, onto the hill, which had a trajectory into the sky, disappeared. Paused, tried to find a crater anywhere... nowhere.
Funny relic engine stuff.