Something Very Wrong with LOS Through Trees

[Cpl Steiner]

Hi all,

I decided to play CM:RT again yesterday after a long hiatus and chose to play a quick battle as the Germans (Axis attack, medium sized town map, human selection of forces). The game launched, and I was greeted by the inspiring sight of my Panther platoon and attached mechanised infantry in a field opposite a large forest, within or behind which I knew some Soviet SU-85 assault guns were lurking. Not wishing to advance into the obligatory turn 1 AI artillery barrage (one of my pet hates about QBs) I did not issue any orders and just clicked the start button. Within seconds, one of my Panther's was in flames - hit by an SU-85 deep within the forest. We are talking extremely deep here - hundreds of metres of dense woodland! Naturally I rage-quit immediately, having wasted a good half hour carefully purchasing the forces for each side to give the best possible battle only to see a ridiculous impossible shot take out one of my best units in the first few seconds of play.

This is a bug, pure and simple. I will not be persuaded that it can be anything but a bug. I checked all the angles and the Line of Sight was clearly passing through multiple trees - not just a few; dozens and dozens.

Now for some "constructive" criticism. To my mind the issue is that tree foliage is being ignored by the game engine and LOS is being calculated purely on the trunks of the trees. It might include the branches also but this is debatable given how unbelievably far this shot went through half a map's worth of dense forest. My proposed solution to this is to somehow treat foliage like crops. Fields of tall crops easily block LOS in the game at present. I can only imagine that this is because the game engine treats fields of crops "en masse" for LOS blockage, as I cannot imagine that every single husk of corn is being checked for LOS. Maybe if the LOS is a certain height over a crop tile, the game just calls it blocked. Well, if this is indeed how it works, I think a similar sort of rule should be applied to tree foliage.

Thoughts and comments welcomed.

Edited January 25, 2016 by Cpl Steiner
Multiple grammar/spelling corrections!

[AttorneyAtWar]

Without a screenshot/Save this is pretty much impossible to be debated or troubleshooted.

Edited January 25, 2016 by Raptorx7

[Cpl Steiner]

Without a screenshot/Save this is pretty much impossible to be debated or troubleshooted.

Yes, I understand, but this sort of game event is not that difficult to replicate. I have experienced it on several occasions. LOS through trees is extremely generous. You could probably replicate it easily using the editor if you wanted to check. I did try to take a save but had already clicked "Done" after viewing the replay, so it was not possible.

[womble]

How hard is it to save the game when stuff like this happens. You might be able to replicate it reliably, but I, personally, have never seen such an extreme "trick shot" happen myself, in spite of similar setups. There's something specific going on here, and to identify it, the saved game will be necessary along, probably, with a full DXDIAG of your machine, in case it's some sort of hardware interaction.

I  have seen shots travel through multiple layers of foliage, on the very, very rare odd occasion, but they were "flukes" and unrepeatable when I tried.

I'm not saying you didn't see this happen. But to get anywhere with fixing it, BFC are going to need more than a "oh it's easy to replicate" handwave.

[weapon2010]

this sort of game event is not that difficult to replicate

then replicate it, and provide some screenshots,better yet a video, so we can all see whats going on

[Cpl Steiner]

OK, I have replicated it as best I can and uploaded a scenario file for you to try yourself as Axis. This was the map and the forces involved in the approximate same positions going from memory. The tanks do engage on the first turn for me when I start turn one without issuing any orders.

I confess there are not as many trees in the LOS as I originally stated but I still think LOS through trees is erring on the too generous side. I doubt you will agree with me but see what you think anyway.

Tree LOS Issue.btt

[slysniper]

Without even seeing your file I am on your side as to the fact the game does not get it right when it comes to viewing and trees.

I have had the same issues at times when tanks can spot tanks through major tree coverage that it should never be able to.

 

Actually one of the best examples is right on this site. a AAR for CMFI with Bil and someone else.

When it was broken down and viewed closely. The player found the baseball size hole that allowed the attacking unit to see through the trees and fire on the enemy unit.

So any engine that works on a concept like that, thinking that something can be spotted through open tree spaces that are that small, has a design flaw.

 

The problem is, its not fixable in the present engine I fear.

So you better just accept it, realize it is not changing in the near future.

I just tell myself, the trees are much more open than they appear, and that large viewing gaps are there, even though they are not to your eye.

 

Amazing how much better I play when I look at trees in that light.

 

 

[AttorneyAtWar]

Without even seeing your file I am on your side as to the fact the game does not get it right when it comes to viewing and trees.

I have had the same issues at times when tanks can spot tanks through major tree coverage that it should never be able to.

 

Actually one of the best examples is right on this site. a AAR for CMFI with Bil and someone else.

When it was broken down and viewed closely. The player found the baseball size hole that allowed the attacking unit to see through the trees and fire on the enemy unit.

So any engine that works on a concept like that, thinking that something can be spotted through open tree spaces that are that small, has a design flaw.

 

The problem is, its not fixable in the present engine I fear.

So you better just accept it, realize it is not changing in the near future.

I just tell myself, the trees are much more open than they appear, and that large viewing gaps are there, even though they are not to your eye.

 

Amazing how much better I play when I look at trees in that light.

 

 

Yeah that's pretty much the same thought I have on the matter. The trees/bushes in game are not in 1:1 scale, the leafs and all that are merely there for looks, most of the time you have pretty clear sight through them.

Edited January 25, 2016 by Raptorx7

[kevinkin]

Su - 85 view before shot (note the end of the gun ready to rumble in the jungle):

 

[kevinkin]

View as you push the view through the tree line (note the white smoke wisp from dead Panther):

[kevinkin]

Overhead with LOS tool:

[kevinkin]

Not sure where the miracle lies. The fact the Panther is visible to the SU-85 or the shell makes it through the tree trucks ... I guess both. Maybe some tree types at a certain coverage should be treated like the side of an invisible brick wall ballistically for direct fire beyond a defined range. 

Kevin

[Cpl Steiner]

Not sure where the miracle lies. The fact the Panther is visible to the SU-85 or the shell makes it through the tree trucks ... I guess both. Maybe some tree types at a certain coverage should be treated like the side of an invisible brick wall ballistically for direct fire beyond a defined range. 

Kevin

As stated in my original post, I think the game only checks for where the LOS intersects a tree trunk or branch (maybe). It's as if the season is perpetually winter and all the leaves have fallen, from the game's point of view. That's one issue with trees anyway. Not sure if this particular scenario file demonstrates it that well as I think the tree foliage closest to the Germans is high enough to not intersect the LOS but I have done other tests in which it seems pretty clear that the leaves are not considered at all in LOS checks.

[Michael Emrys]

I have experienced it on several occasions. LOS through trees is extremely generous.

Although I would not use the word 'extremely' in this context, I tend to agree with you. I have been burned several times by the AI getting an LOS/LOF where I would have expected it to be highly unlikely or even impossible. As to how credible this is, well that depends not only on how many trees there are along the line of sight, but what species they are, how closely they grow together, what the season is, presence or absence of undergrowth and lighting. But I suspect that at present, CM uses a rather more simplified algorithm. What I earnestly wish for is a more WYSIWYG representation. For men on the ground, it is not too hard to see to what extent LOS/LOF might be blocked or at least questionable. I would like to have the same facility.

For the particular case that Cpl Steiner brings to us, I notice that one row of the trees between the two units of concern are tall fir trees having no foliage for the first two or three meters above the ground. They would hardly block LOS at all. But I have seen other cases where the argument could be made more strongly.

Michael

[Cpl Steiner]

I agree that the type of tree here is not the best for the "foliage" issue that several people have seen (me included). I believe this is because the foliage is ignored and only the 3D model of the tree is considered.

Here is a proposed solution, perhaps for another iteration of the game engine a long way into the future.

LOS Test Sequence (assumption: only trees considered)

1. Does LOS cross any part of a tree tile?

No = End (LOS is clear, as far as trees are concerned)

Yes = go to 2.

2. Does LOS intersect a tree model in any of the tree tiles crossed?

Yes = End (LOS is blocked)

No = go to 3.

3. Does LOS pass through an imaginary cube above the tree tile that represents the approximate location and dimensions of the tree canopy?

No = End (LOS is clear)

Yes = Might be blocked  as LOS is going through the tree's leaves. Other factors could then be considered to determine if the LOS is actually blocked (e.g. season, distance of sighting unit from tile, distance of target from tile, etc. etc.

I am not saying this is perfect but I think it would give an effect similar to what we already have for crop doodads. I am also not trying to suggest that BFC need any help from me as regards coding the game engine! I have enormous respect for what they have achieved given their limited resources but there is always room for improvement and unbelievable spotting through tree canopy would be high on my list.

Edited January 25, 2016 by Cpl Steiner
Missing equals sign

[MikeyD]

The tall pines that are in the game we might as well be in a forest of telephone poles for all the concealment they give.  You'll notice those lighter colored trees in the screenshot foreground have foliage closer to the ground. They'd have been a safer bet. I've seen a clump of bushes block LOS while a row of tall trees didn't. I recall no less a personage than George Patton felt compelled to instruct his armor in the ETO to stay away from the friggin' tree line. You've got no cover from fire and less concealment than you think you have. Get yourself behind a proper terrain feature instead! Per General George S.

[Wicky]

[DLaurier]

I've had a similar experience, only in the opposite direction. In a QB against the AI, my T-70 spotted a panzer of some sort through a deep zone of trees, and pummeled it for over a minute.
I save/surrendered, and took a peek. I had immobilized a panther G, and disabled it's gun.
Happiest missmatch result EVER.

IRL I have seen just how variable sight lines can be in summer forest. Gaps can be a narrow as your arm, but can run hundreds of feet. If you are in the right spot you can see out into the open, and beyond. Step a mere half a pace to either side though, and you cant see past spitting range.

Your SU 85 got lucky.

[A Canadian Cat]

LOL nice example.

I just tell myself, the trees are much more open than they appear, and that large viewing gaps are there, even though they are not to your eye.

Amazing how much better I play when I look at trees in that light.

Good advice there. I can see a reasonable argument for foliage thickness not matching its visual representation. I really do not think it is as a extreme as @Cpl Steiner is making it out...

 

To my mind the issue is that tree foliage is being ignored by the game engine and LOS is being calculated purely on the trunks of the trees.

There is no way that is right. I realize you are deep in observation bias territory here but you must have seen times when your tank, just a short distance into the woods cannot see squat in the next field. I know I have. There is no way that your conclusion above is even remotely correct. If you want to argue that foliage does not do enough to block LOS sure I can see that perhaps even agree but it is being accounted for and has the expected effect a lot of the time.

 

[womble]

The problem lies in the answer to:"How much of a target does a shooter have to be able to see to be able to spot it?"

There have to be hard parameters, it's a computer game. LOS is traced as 1-dimensional lines with no width, so it can be threaded through the unlikeliest of "baseball-sized" lacunae in foliage. So the alternative is to recheck every "potential" LOS hit (by the narrow ray) for others near it until a "big enough" area of target polygons has been spotted. Lots of extra work for the ol' CPU, there.

Or we can live with it and not trust too blindly to the trees keeping us hid.

[kevinkin]

It's also how likely a gunner would fire through a multi layered keyhole with the prospect of wasting ammo, revealing their position and spraying friendlies with debris caused by a bad shot. But programming that sort of AI thinking is beyond the scope of the overall game. Players think with confidence their units are in safety and then this LOS/LOF issue pops up. Their sound battle plan is ruined by some quirk in the rules. There seems to be enough evidence to modify the code when the opportunity arises. How? We can think it through. But the simplest solution would be the one most likely implemented.   

Kevin

PS: Anyone have a reference for Patton's advice on tree lines?

 

[JoMac]

 

 

The tall pines that are in the game we might as well be in a forest of telephone poles for all the concealment they give.  You'll notice those lighter colored trees in the screenshot foreground have foliage closer to the ground. They'd have been a safer bet. I've seen a clump of bushes block LOS while a row of tall trees didn't. I recall no less a personage than George Patton felt compelled to instruct his armor in the ETO to stay away from the friggin' tree line. You've got no cover from fire and less concealment than you think you have. Get yourself behind a proper terrain feature instead! Per General George S.

And here is MikeyD's Post on Patton's quote regarding the Tree-Line. 

Edited January 26, 2016 by JoMc67

[CarlWAW]

I recently finished a scenario with lots of infantry hiding in small wood pockets. I was positively surprised how low the visibility is.

I think what is crucially important to understand: in CM trees do not make a wood but only are one part while the other part is the ground foliage which must be thick wood, too for them to match expectations of thick wood.

Probably some of the complaints about LOS could be avoided, if the game would display the action spot map elements the cursor is hovering over. If it would say i.e. "grass, trees" then players would not expect wood blocking LOS.

[kevinkin]

Carl, you are right about using heavy forest ground cover which tends to work except not with the tree type in the screen shot above. I switched the light to heavy cover thinking along the same lines as you and the LOS keyhole did not go away in the experiment. 

Kevin

[JasonC]

Consider Wicky's picture.  Yes light penetrates some distance into the trees, with wide gaps between the physical obstacles of the trunks, and no higher foliage to speak of.

But look across the field at the distant treeline on the next hill over.  Scan along it.  How deep can you really see into that treeline?

It isn't a matter of physically blocking every photon.  It is a matter of the human eye being hopelessly unable to pick up any shape in that solid, high, blocking line, down in the shadow under the canopy.  The eye focuses involuntarily on the lines of high contrast.  We have actual edge detection cells on our retina.  What you actually see is a high contrast line across the top of the canopy, between canopy and sky.  Plus another high contrast line along the base of the trees, where trees meet ground.  Plus, less distinctly, a shadow line on the snow some distance in front of that ground line, wavering with the tree heights and undulations of the ground.

Between the skyline and the ground line you see a block of trees.  Some individual lines of contrast are detectable in the tree shapes and foliage, but each is so like all the others in the mass that you get a texture signal, not really edge signals.  You can detect smudges of interrupted white below the foliage line, from snow on the ground back inside the tree line.  Together those can give you another faint line of the lower edge of the foliage, blurred across the multiple trees that define it and the angle of the ground relative to teh angle of the sightline.

But any distinct object on those interrupted snow patches below the foliage line?  Nearly hopeless.  If the trees weren't present, you would edge-detect such a shape as a hole in the field of uniform white of the snow on the ground.  But within the trees, it is all interrupted by trees anyway, and in shadow.  There is no high contrast to pick out the shape of an object against that field.

This is fundamentally a camouflage effect, not a physical LOS blockage effect.  But it is very real.  And it causes a sighting asymmetry - it is far far easier to see something out in that snow covered field than it is to see something inside the tree line from the field.  Photons are intercepted or not, either way, so you might expect the LOS effect to be reciprocal and equal.  But it isn't, not with the camouflage effect included.

If a tank were sticking out of that far tree line it would certainly be visible.  But 10 or 20 yards back inside of it, not moving, and it would have great camouflage effect compared to the same out in the middle of the snow white field.  The tank inside the treeline would see the one out in the field ages before the reverse happened.

Imagine trying to spot a tank painted bright hunters orange.  That would depend only on physical LOS blockages.  But actual tanks are not painted orange, they are painted in camouflage colors and patterns to make spotting them harder.  Put one back 20 yards in a treeline across a field like that distant ridge, and they'd be effectively invisible until they moved or fired.

What the game seems to be underrating is this camouflage type effect.  Otherwise put, the spotting routines pick up marginally "visible" targets far too rapidly and reliably, in the absence of the movement or fire that actually attracts the eye and enables us to pick out a particular shape at range in such uniform visual environments.  The routine probabilities seem fine for large objects standing out against a skyline or a uniform field background of high contrast to the vehicle itself.  But that situation is an outlier of best spotting chances on real battlefields, not the rule.

Edited January 26, 2016 by JasonC