Issues with tank targeting accuracy

[RobZ]

So from some long experience with this game its become clear to me that the accuracy of tanks and AT guns are way too accurate once they are zeroed in. The AI will aim pixel perfect on the same spot every shot, only the gun accuracy itself will deviate the hits. Here is some tests i did with and without cover infront of the tank (hull down). The lesson here seems to be that a tank with enough armor SHOULD NOT go hull down cus its a death sentence due to how AI aims and mixed with the unreal zeroed accuracy the main gun will get knocked out very quickly.

Tiger 2, behind a 2m hill (hull down) at 1000m vs 76mm guns. At 1000m i do not expect the hit area to be this tiny. The side and top turret is nearly untouched and the muzzle break is completely perforated from existance.

Tiger 2, at 1000m not hull down vs 76mm guns. Here we can see that the AI targeting has changed to the hull instead and the turret is nearly untouched (only 3 shells hit the very lowest part of the turret). In this scenario the shermans ran out of AP so i deacivated the target arc for the tiger and it knocked out all 5 of them, while in the hull down scenario the main gun was knocked out almost instantly and would render the tank useless.

Here we have a jagdpanther at 600m behind a 1m hill vs 76mm guns. Only the lower front is hull down. Again we see the insane accuracy once the tanks have been fully zeroed that gives a unreal hit area. The only deviation is the gun accuracy, not the "humans" aiming it. The mantlet for tank destroyers also seem unrealisticly weak to get penetrated at those insane angles and thus knocking out the main gun. Another thing with this one is that odd penetration on the barrel. How on earth can a shell penetrate the barrel at that angle, this should not be possible.

Jagdpanzer IV L/70 at 600m behind 1m hill vs 76mm guns. Only the lower front is hull down. Here again the insane accuracy and main gun knocked out instantly.

Jagdpanzer at 600m on flat ground vs 75mm guns. Here we see the targeting area has changed cus it has no terrain infront of it. In this scenario the main gun remains operational cus the AI cannot abuse its accuracy on the mantlet area so this tank would be better off than if it was hull down.

 

The thing im saying is not that the overall accuracy is too good, cus that works just fine. What i am saying is that once the AI gets fully zeroed, they have no deviation what so ever in their aiming. Only the gun accuracy itself shows on the hit area of the target and it gives a unrealistic scenario of hits. All rounds land within tiny areas and if you use terrain to get hull down (which should be a good tactic) you will risk loosing the main gun very quickly. I expect to see hits all over the tanks in these scenarios and not within a tiny circle at +600m, remember there is supposed to be humans actually aiming the cannons, but the AI clearly aims at a single dot on the target with no deviation once the gun is fully zeroed. The few shells you see away from the main hit area is made before the gun is fully zeroed inwhich deviation is fine.

I have only terrible experiences with StuGs for example cus the only thing that gets hit on those is the mantlet. And once the mantlet is hit (even by a stuarts 37mm) the main gun will be knocked out. In my games with stugs i get a unreal amount of main gun damages for shells hitting the gun directly or the mantlet (which should be 80mm like the rest of the front, but still get pierced for some reason)

 

EDIT:

Here is the deviation at 2000m. Notice how all rounds hit in a nice circle at center mass, the few shells that hit the sides and lower plate was before the gun was fully zeroed in and still had some aiming deviation.

For refrence this is how the target would look from the gunners perspective, 5x gunner optics zoom. The target is tiny so managing to hit within that circle every time would be nearly impossible.

Edited May 6, 2020 by RobZ
More refrences

[RobZ]

More insanity.

Tiger at an angle from 650m vs 76mm. All shots hit the same tiny area which is "center mass" from the shermans perspective.

[MikeyD]

There's an anecdote out of Hunnicutt's book 'Pershing' where 11th Armored was treated to a demonstration of the new tank. Three new Pershing crews were given ten rounds per tank to test the gun's accuracy. They were plinking German helmets at 625 yards range with the big gun on the far side of a small lake. Opinions of how much dispersion we should see on targets at various ranges is just that, opinions. No more or less valid than anyone else's.

[RobZ]

14 minutes ago, MikeyD said:

There's an anecdote out of Hunnicutt's book 'Pershing' where 11th Armored was treated to a demonstration of the new tank. Three new Pershing crews were given ten rounds per tank to test the gun's accuracy. They were plinking German helmets at 625 yards range with the big gun on the far side of a small lake. Opinions of how much dispersion we should see on targets at various ranges is just that, opinions. No more or less valid than anyone else's.

The gun accuracy is not the point im making. Im talking about the AI literally aiming at the exact same pixel every shot, once the gun is fully zeroed the only deviation is the actual gun accuracy and not the "human" aiming it. My last post shows what im talking about, the tiger is only hit in one spot cus the AI aims at the exact same location every time which isn't realistic. 

Here is a tiger hit over 200 times by various calibers, and it sure looks like it was hit all over the place and doesnt have a small circle on its center mass.

[RobZ]

This red cross represents 2.5m x 2m used for accuracy tables. Those hits are from sherman 76mm at 2000m in a combat scenario. It seems 2 hits are outside that area. Meanwhile the 8.8cm KwK43 gun in training has 85% accuracy for that same target. So in this scenario the 76mm sherman in combat conditions have better accuracy than 8.8cm kwk43 has in training. This is fully zeroed ofcourse, which would be about equal to the kwk43 training scenario where exact range is known to test accuracy.

[Erwin]

RobZ's tests are very interesting as it reminds me of a discussion we had when CMBN first was was released regarding whether a PzIV should ever be used in a "hull down" position.  Tests seemed to show that because the mantle of the PzIV was more easily penetrated than the front hull glacis the PzIV had higher % of being KIA when hull down.  Obviously, the "to hit" % was lower when a tank was hull down.  But, any hit was far more likely to result in a KIA PzIV tank.  Of course, if the hull down tank kills the enemy tank first, that resolves the issue.

However, the overall conclusion was (in the game at least) it was statistically better to have a PzIV tank in a non-hulldown position when in combat - especially when sitting there trading shots.  Folks have also commented at how frequently any tank can suffer gun hits.  

So, it seems that what your tests are confirming is that (in the game) prolonged combat when hull-down should be avoided for tanks that have a weaker mantle than the front hull and also to avoid suffering gun hits that turn the tank into an expensive MG pillbox.

 

[wadepm]

First, don't trade shots.  Keep moving.

Second, once you have the range it is relatively easy to hit the same spot over and over.  Assuming the commander can see the hit and provide confirmation to the gunner that he has the range it is just the matter of aligning the site and setting the target at the correct range mark. Of course, this is hard to do when in combat and the other guy is shooting back, the muzzle blast is kicking up dust, the tank commander is screaming at you, you're tired, thirsty and hungry, etc.  Maybe the game doesn't include enough off that...  

[RobZ]

1 hour ago, wadepm said:

First, don't trade shots.  Keep moving.

Second, once you have the range it is relatively easy to hit the same spot over and over.  Assuming the commander can see the hit and provide confirmation to the gunner that he has the range it is just the matter of aligning the site and setting the target at the correct range mark. Of course, this is hard to do when in combat and the other guy is shooting back, the muzzle blast is kicking up dust, the tank commander is screaming at you, you're tired, thirsty and hungry, etc.  Maybe the game doesn't include enough off that...  

Yes, dont trade shots is the best tactic, but this isnt about how you should play the game.

Depends what you mean the "same spot", as for the AI in this game they can aim on the same pixel every time which is not humanly possible. The combat stress is something that adds ontop of all this indeed, and as i showed in a previous post the sherman gets higher accuracy in combat than tiger 2 would in training.

[Hapless]

23 hours ago, RobZ said:

Those hits are from sherman 76mm at 2000m in a combat scenario. It seems 2 hits are outside that area.

There could be hundreds of shots outside that area: if they miss the Tiger completely there's no record!

I assume that everythings aims at the centre of mass of the target because it's less likely to give the game engine an aneurysm. This makes sense on a lot of levels and not just to avoid stressing the engine, it's a ridiculous rabbit hole that vomits forth far, far more problems than it would ever solve.*

 

*For example: "Shermans are targeting my Panther's lower mantlet shot trap in 1943 but they don't know about it yet because they're never seen Panthers before" vs "My Shermans are still aiming for the lower mantlet shot trap on late version Panther Gs even though I found this intelligence report from a month earlier that shows they knew all about the mantlet chin".

[holoween]

Since this is something that only really matters against well armored tanks giving a deviation from the center mass aimpoint if the round bounces to simulate the gunner aiming for different areas could help spread the shots around a bit more realistically.

[RobZ]

25 minutes ago, Hapless said:

There could be hundreds of shots outside that area: if they miss the Tiger completely there's no record!

I assume that everythings aims at the centre of mass of the target because it's less likely to give the game engine an aneurysm. This makes sense on a lot of levels and not just to avoid stressing the engine, it's a ridiculous rabbit hole that vomits forth far, far more problems than it would ever solve.*

 

*For example: "Shermans are targeting my Panther's lower mantlet shot trap in 1943 but they don't know about it yet because they're never seen Panthers before" vs "My Shermans are still aiming for the lower mantlet shot trap on late version Panther Gs even though I found this intelligence report from a month earlier that shows they knew all about the mantlet chin".

Yes there is a lot of misses first, but that's when the range is not exactly known, so several ranging shots will miss. But when max zeroing is reached, every single shot hits within that small area. The game engine can already handle deviation as seen when units first fire at long ranges, they can miss by huge margins so the game already handles this. The issue is that when they are fully zeroed, they no longer have deviation in aiming and it shows on the hits cus everything hits the same spot like the AI was spesificly aiming exact center of the target. And yes they should aim at the center of the target, but not with this computer level precision. As I stated earlier, tiger 2 has 85% accuracy within the red Cross, the Sherman in that scenario has way better than that, might even be 100% cus those 2 shots outside that area was early hits before zeroing was complete.

Edited May 7, 2020 by RobZ
typo

[wadepm]

Have you tried the same test but with an MG-34 peppering the firing tank?  I wonder if that will effect how soon he gets the range and subsequent shots.

[RobZ]

25 minutes ago, wadepm said:

Have you tried the same test but with an MG-34 peppering the firing tank?  I wonder if that will effect how soon he gets the range and subsequent shots.

Sherman 76 at 2000m, beeing shot by a HMG 34 team constantly. This is 95% accuracy for the 2.5m x 2m target while under direct MG fire.

[wadepm]

Can't see any difference.

What about crew experience?

[RobZ]

1 minute ago, wadepm said:

Can't see any difference.

What about crew experience?

Right, forgotten to mention this. Crew experience during ALL tests has been regular, 0, normal.

[Hapless]

10 hours ago, RobZ said:

The issue is that when they are fully zeroed, they no longer have deviation in aiming

This is the point of zeroing.

10 hours ago, RobZ said:

like the AI was spesificly aiming exact center of the target

Because that is exactly what they are doing.

10 hours ago, RobZ said:

As I stated earlier, tiger 2 has 85% accuracy within the red Cross, the Sherman in that scenario has way better than that

It's pretty pointless to compare real world data for Tiger 2 and ingame data for the Sherman 76 (Have you got a link or reference for the Tiger 2 accuracy stuff?) It would more useful to compare the ingame Sherman accuracy once zeroed to the Tiger 2 ingame accuracy once zeroed.

On 5/6/2020 at 7:29 PM, RobZ said:

For refrence this is how the target would look from the gunners perspective, 5x gunner optics zoom. The target is tiny so managing to hit within that circle every time would be nearly impossible.

The Sherman gunner's main optic is 4x IIRC, so it looks even smaller! Im not 100% up on my Sherman fire control and gunnery mechanics, but I don't think firing the gun is going to change the gunner's point of aim... so why would the gunner voluntarily aim somewhere else once he's on target?

Finally... in theory you could increase the deviation to make the guns less accurate to simulate the gunner "shifting his aim" or "targeting different points of the tank", but

a) how do you know that BF hasn't already done this?

and b) It's such as an edge case. This setup- one tank plinking another 2000m which isn't allowed to shoot back on a flat map with no cover- is an accuracy test (which I understand is what you're testing) but it's not an accuracy test that takes into account likely battlefield conditions. What you're effectively testing is the maximum accuracy of the Sherman vs the Tiger 2, but there's no indication that this is relevant to actual gameplay. If the Sherman doesn't survive long enough or isn't exposed for long enough to get that maximum accuracy- or the Tiger 2 is smart enough to avoid getting plinked like this- then does it matter?

[RobZ]

3 hours ago, Hapless said:

This is the point of zeroing.

Yes i know, but the AI does not have any "human" deviation in aiming, it manages to aim pixel perfect center line on the vehicle.

3 hours ago, Hapless said:

The Sherman gunner's main optic is 4x IIRC, so it looks even smaller! Im not 100% up on my Sherman fire control and gunnery mechanics, but I don't think firing the gun is going to change the gunner's point of aim... so why would the gunner voluntarily aim somewhere else once he's on target?

Finally... in theory you could increase the deviation to make the guns less accurate to simulate the gunner "shifting his aim" or "targeting different points of the tank", but

a) how do you know that BF hasn't already done this?

and b) It's such as an edge case. This setup- one tank plinking another 2000m which isn't allowed to shoot back on a flat map with no cover- is an accuracy test (which I understand is what you're testing) but it's not an accuracy test that takes into account likely battlefield conditions. What you're effectively testing is the maximum accuracy of the Sherman vs the Tiger 2, but there's no indication that this is relevant to actual gameplay. If the Sherman doesn't survive long enough or isn't exposed for long enough to get that maximum accuracy- or the Tiger 2 is smart enough to avoid getting plinked like this- then does it matter?

The first shermans had no telescope at all, just a periscope. Then came the 3x12 telescope for sherman 75s and also existed for 76s. Sherman 76 tanks mostly had a 5x13 telescope. You must be refering to the M83 which i honestly don't know where it was used as i find no sources on it, but it was a variable 4x-8x telescope.

 

a)

This shows how the AI aims for the "center line" of the vehicle in all cases. The obvious aiming point for a ww2 tanker in this scenario is to shoot the rear, but they dont. This is sherman 75s targeting tiger 2s rear-side from ca. 1000m, and not just 1 tank, but 3. So either they all agreed on where to aim or the game only tells the AI to shoot the absolute center with no aiming deviation, and this is the issue im talking about. Im not saying gun accuracy is too good, but the AI is too perfect at aiming and add that with the more or less real gun accuracy already in the game, it gives a unrealisticly small hit zone at the exact same area. The only time the AI adjusts their aim is if terrain is infront of the target where they adjust upwards.

 

b) I believe we see this game from different perspectives when playing. I would not have started any of these tests if i did not notice any weird behaviour from actual playing. I know that this is a game so i notice when "gamey" moments happen. Like all my stugs getting main gun damaged cus the AI aims perfectly dead center every shot and thus the gun is more likely to get hit. You mentioned earlier that the game engine might not handle this type of aiming deviation. And that might be true for the horizontal plane. As shown in the pictures the AI doesnt aim sideways at all, but they can adjust up and down. So it might infact be a engine limit for actually aiming around on the target. I already know that all vehicles in the game is just a single vertical line to the AI. I have had situations where a single tree is blocking the AI from shooting an enemy tank while you could clearly see the whole tank stick out around the sides of the tree, but the center line of the vehicle is blocked and thus the AI couldn't fire.

As for reaching maximum accuracy. This can be done with target refrence points. In these tests i have used those most of the time to quickly get max accuracy. And at 2000m it usually only takes 1 miss before every shot starts hitting (occasional miss here and there). Another way is if your units have already engaged a target at that range and a new target appears, they will already know the range and have near max zeroing already. Mix this with the flawless center line aiming of the AI and you get shots landing exact center and not in a more realisticly spread manner.

Why shouldn't the AI be a perfect center aim? That's because in real life its not always so easy to perfectly line up your sights at long ranges. Targets get pretty small through the telescope sight and the "crosshair" in your sight will start to become larger than the target and that makes it way harder to line up a dead center shot like the AI can, which would lead to more deviation on the target.

When it comes to fire controls and if firing the gun moves the point of aim, i don't know the answear to that. But at 2000m the gun doesnt need to move much for the aim point on the target to move. For the gunner it might look like it didnt move at all, but at the target it might have moved 25cm to the right/left.

 

Sherman 76 vs tiger 2 at 2000m

Tiger 2 vs tiger 2 at 2000m

These results doesn't tell me much. It looks rather similar, except the shermans got more rounds on target cus the tiger 2s actually killed their target. Both shermans and tigers missed their first shots and then started hitting. Shermans missed 2-3 times after that aswell while tiger 2s did not, but that might be RNG and the fact that shermans fired more.

[Erwin]

The main question is whether the "effect/end result" reflects/simulates reality. 

Are the % of gun hits we experience in CM2 WW2 games similar to what was experienced in RL? 

Was it known that PzIV crews need to avoid trading shots from a hulldown (presumably ambush) position as eventually, the mantle would be hit and that was a sure kill as it was weaker than the glacis?  

If so, then no problem...

[RobZ]

3 minutes ago, Erwin said:

The main question is whether the "effect/end result" reflects/simulates reality. 

Are the % of gun hits we experience in CM2 WW2 games similar to what was experienced in RL? 

Was it known that PzIV crews need to avoid trading shots from a hulldown (presumably ambush) position as eventually, the mantle would be hit and that was a sure kill as it was weaker than the glacis?  

If so, then no problem...

This is true, but sadly its probably impossble to get real life data on this. When it comes to hull down though, im pretty sure hull down is always positive in real life. As you do have a human aiming so the aiming skill is still the same, but the overall target size is much smaller and that just automaticly makes it less likely to get taken out. The game represents this well with hull down beeing less likely to get hit, but when your tank has enough armor to actually resist those hits then there is a much higher chance of the tank beeing mission killed than if it was in the open. And this is due to the AI aiming, they never aim at the turret if the vehicle is completely exposed. They aim center-low hull which means the turret is very clean like shown in many of my tests. Panther is one of the tanks that has a clear advantage with exposing its full self, as the hull is much stronger than the turret and the AI automaticly aims for the hull.

[Erwin]

Part of the issue is that in the game, we are usually fighting at what would be considered "point blank range" due to most maps either being small and/or not having long LOS opportunities.   So, a tank gets hits a lot quicker (maybe) than in RL?  

Presumably in RL, a PzIV in a hull-down position would have a higher % chance to hit and kill the enemy tank before the enemy tank had enuff time to accurately hit the smaller PzIV turret mantle.  

So, perhaps what needs to be tested is who hits and kills first in a duel between a hull-down PzIV (or Panther) vs an equivalent enemy tank that is not hull-down.

[RobZ]

12 minutes ago, Erwin said:

Part of the issue is that in the game, we are usually fighting at what would be considered "point blank range" due to most maps either being small and/or not having long LOS opportunities.   So, a tank gets hits a lot quicker (maybe) than in RL?  

Presumably in RL, a PzIV in a hull-down position would have a higher % chance to hit and kill the enemy tank before the enemy tank had enuff time to accurately hit the smaller PzIV turret mantle.  

So, perhaps what needs to be tested is who hits and kills first in a duel between a hull-down PzIV (or Panther) vs an equivalent enemy tank that is not hull-down.

I did do one such test, 5 jacksons vs 5 panther/tiger. The panthers and tigers were put in a 2m hull down position so only the turret was showing. The jacksons were placed completely open on flat ground. The units were 2000m apart and had target refrence points on enemy position. Jacksons won 3/4 times vs tigers, and 3/3 times vs panthers. More tests would have to be done to give conclusive results, but so far its looking very good for the completely exposed jacksons.

[Aquila-SmartWargames]

I am curious aswell on how the probability of CM tank gun destructions compare to reality. Never digged into that. I´ve also noticed that deploying specific tanks into an hull down position might drastically raise the chance to end up with a firepower/mission killed (i. e. gun destroyed) tank.

[RobZ]

Did another test

5 stuart vs 3 panzer IV @700m, no target refrence points used

First 5 rounds was with the panzers in hull down position

This is how much is exposed from the stuarts perspective.

When in hull down position the panzers won 2/5 times, and those 2 times they lost 2 tanks the first time and 1 tank the last time.

 

Then i repeated the tests at the same range, but now panzers were placed in the open and not in hull down position

In this scenario they won 5/5 times, and lost 0 tanks in all of them. Only lost 1 tank twice to mission kill (main gun damage)

This happens because the stuarts aim low on the hull and once zeroing is taking good effect they will only strike the hull. Meanwhile in the hull down position they will obiously strike the turret and they do miss more, but in this scenario that doesn't matter as the panzers get quickly knocked out when zeroing reaches good levels.

So this is one of the scenarios you do not want your panzer in a hull down position.

 

[Erwin]

1 hour ago, RobZ said:

When in hull down position the panzers won 2/5 times, and those 2 times they lost 2 tanks the first time and 1 tank the last time.

Then i repeated the tests at the same range, but now panzers were placed in the open and not in hull down position

In this scenario they won 5/5 times, and lost 0 tanks in all of them. Only lost 1 tank twice to mission kill (main gun damage)

Thanks for doing the tests.  That's a bit disturbing.  Hope you have already reported this as something for BF to look at for the next upgrade.

(Did you look at any gun damage?)

Edited May 8, 2020 by Erwin

[MikeyD]

So in theory PzIV should win 2/5 times against Stuart in all circumstances because the 50mm turret front is no match for the 37mm gun, whether its hull down or not. People would be upset if that happened, I suspect. My experience with hull down most often is the first shot will go high, the second and third will impact the ground in front, then the gunner fill find the range and start plinking the turret front consistently. Except in modern war titles where you'll probably get a catastrophic first round turret front hit