Jump to content

CMBN LOS mechanics testing my patience


Recommended Posts

What ASL said.. micro terrain etc. People need to remember that this is not a precise WYSIWYG. This also brings us back to what womble said..

Part of the deal is that they (some of the instances) look "odd" but actually aren't.
Link to comment
Share on other sites

  • Replies 75
  • Created
  • Last Reply

Top Posters In This Topic

Yeah, I wouldn't post these if they weren't.

You wouldn't if you didn't think they weren't. Sure.

I can in theory do this check by going back to the previous PBEM turn but I'm not sure what you mean by "every point of observation" and not sure how it would help.

You can see bits of vehicles (and spot and LOS them) other than the bits of that vehicle that can see and spot you... Eye to eye, it's pure reciprocity. Eye of A to engine deck (for example) of B doesn't mean B can see A.

It's not like this game lets you use the representations of tree trunks, branches/foliage to accurately predict LOS.

I think that's ever going to be the case, and is going to lead to situations where what you thought was probably going to happen doesn't.

I would hope the -2 leadership on the TC is not responsible for 40sec of potential LOS not ending in established LOS despite main guns and MGs being fired 80m away.

It certainly might be a contributory factor. The IV spots the M4 fairly soon after it turns its hull towards the infantry contact it fired its main gun at, so the M4 isn't very well hidden, at least some parts of it aren't. If the IV can see the lower hull and running gear below the trees, and the angles mean that the crew of the Sherman have their view completely blocked in that direction by foliage, then the Amis won't see their executioner. Just because none of the crew positions in the tank can see any valid part of the IV doesn't mean the reverse, and I'm speaking in a geometric sense, rather than considering any sort of game mechanism.

Link to comment
Share on other sites

While mounted in their vehicles tank crews are effectively deaf unless the engine is turned off and the commander has removed his internal communications equipment so I think you need to recalibrate your expectations in that regard. 1 meter is approximately 1.09 yards. 80 meters is approximately 80 yards so very close to the 100 yard length of an American football field. The ability to see the length of a football field through wooded terrain is not as much a certainty as you might think either.

Bull never stated that his tank was 80m inside the forrest and the enemy was stationed just outside the woods. As far as i understand the situation was that his tank is positioned just within the woods and that the enmy tank is 80 m away sitting in unconcealed open terrain. Thats a big difference.

Link to comment
Share on other sites

Bull never stated that his tank was 80m inside the forrest and the enemy was stationed just outside the woods. As far as i understand the situation was that his tank is positioned just within the woods and that the enmy tank is 80 m away sitting in unconcealed open terrain. Thats a big difference.

It is, but still, it's 80m away with trees in the observer's face.

Link to comment
Share on other sites

That realy much depends on the "thickness" of trees and underbrush. Being 10-20 meters inside a normal forrest doesnt realy block your ability to look outside while someone looking from outside into the forrest will have a much harder time spotting due to lighting and general background.

Would be nice Bull if you could provide a screenshot of the situation.

Link to comment
Share on other sites

That realy much depends on the "thickness" of trees and underbrush. Being 10-20 meters inside a normal forrest doesnt realy block your ability to look outside while someone looking from outside into the forrest will have a much harder time spotting due to lighting and general background.

If your head is at a height that is poking up into the canopy of the trees, one tree may be enough that you can't see out.

Would be nice Bull if you could provide a screenshot of the situation.

Pretty sure he's provided two videos in his OP of this particular geometry. I hope we've not moved onto another thing I've not looked at yet...

Link to comment
Share on other sites

I have made quite a comprehensive video of the Sherman/PzIV incident.

PS: The Sherman TC is a +1 (not sure why -2 was mentioned earlier on). The Sherman is kind of more on the edge of an orchard than a forest

I have to removed the transparent trees mod so that the "default" foliage graphics are used in these videos.

The video starts with an overhead look at where the Sherman and PzIV are located so you can get an idea for where the PzIV is located. I then add video I took when I loaded up the previous turns order phase (prior to the PzIV showing up) and did a few LOS checks from the Sherman to and from the spot the PzIV ends up occupying. I also get down at tank level to show what is graphically represented in the game.

PS: Don't watch these with the embedded viewer in Dropbox. The playback quality is poor. Download file and watch.

Sherman LOS checks.mp4

I also have added a video of a LOS check from the movement action spot located where the PzIV ends up back top to the Sherman as a reciprocal check.

PzIV LOS.mp4

A few things:

1. Based on the default graphical representation of trees and foliage in the game, you are hard pressed to think that ANY LOS between the PzIV and Sherman could ever exist. It is a shame that the disconnect between the graphical representation and the actual in game LOS mechanics (of which I believe the LOS tool is an extension of) is so great. It does make playing this game very difficult.

PS: The original videos I posted up featured the transparent tress mod which may have given the impression that based on graphics alone, that it was more believable that LOS could exist.

2. For all the graphical representation of the foliage we see in the video, which we know is over-re-representative because the in-game LOS mechanics DOES tell us LOS does/can exist, it is still incredible that the one vehicle with a dead TC that was preoccupied engaging multiple targets and firing it's main gun and MGs for 40sec, was the the one vehicle that remained unspotted at the end of the 40sec, whereas the vehicle that was just sat there unbuttoned with all crew spotting saw and heard nothing.

Link to comment
Share on other sites

The Pz IV/Sherman encounter could be an outlier event. The result you saw seems unlikely but not impossible. The only way to test if something is amiss would be to re-run the same turn many times and see how often each tank see the other first. There is a fair amount of randomness in spotting so a single result doesn't tell you much.

Link to comment
Share on other sites

PS: The Sherman TC is a +1 (not sure why -2 was mentioned earlier on).

Sorry, my mistake. It looked like a -2 to me on the first video.

PS: Don't watch these with the embedded viewer in Dropbox. The playback quality is poor. Download file and watch.

And that would have been the mistake I made... :)

Thanks for the detailed movie. I think it bears out my hypothesis that the PzIV can see the lowest portion of the Sherman below the foliage, whereas the Sherman's vision is blocked by the foliage that's at the same level as its "eyes". Though the Sherman can see through the nearest canopy, the "last tree on the sight line" seems to me to cut off the LOS to the IV's location (and you were pretty unlucky; one AS to either side and you'd've seen it). If you go to the PzIV location and use the maginfication zoom to look back, I think you'll be able to see the running gear and front hull "below the curve" of the Sherman, without any trunks blocking the LOF.

Link to comment
Share on other sites

Wow I was away for a few days - you guys had tonnes of fun on this thread :D

I have to say I don't see an issue with @sburke's rant earlier - other than it might have been a bit much from @Lt Bull's post it certainly was not out of line based on the few other people chiming in and getting really upset and worked up at mine and @womble's explanations.

Having said that@Lt Bull I have a question have you given saves to @sburke or another bets tester? It sure looks like you are making new videos. It might actually help them figure out of there is actually a problem here. If you already have done so off list then please just disregard me.

I will relay a situation that just happened to me: My Cromwell was parked in a bocage lane when a Tiger came into view. I though oh now not another tank loss (already lost three tanks in recent turns) but nothing happened. The two tanks just sat there staring at each other. Lucky for me. During the orders phase I tried to target the Tiger but no luck. Closer inspection showed that the TC can just barely see the roof and TC of the Tiger. Of course in game they either draw the whole tank or none of the tank. So it looks like the tanks can see each other but I think the two TCs are the only things each other can see. So, it looks a little odd, in one way, but is working fine.

Time to get that Cromwell to get away fast...

Link to comment
Share on other sites

The Pz IV/Sherman encounter could be an outlier event. The result you saw seems unlikely but not impossible. The only way to test if something is amiss would be to re-run the same turn many times and see how often each tank see the other first. There is a fair amount of randomness in spotting so a single result doesn't tell you much.

Yes, I would like to do this. Unfortunately my opponent was unwilling to co-operate. I don't know if BFC can do anything with a PBEM without the password either.

I think it bears out my hypothesis that the PzIV can see the lowest portion of the Sherman below the foliage, whereas the Sherman's vision is blocked by the foliage that's at the same level as its "eyes".

How can you say the Sherman's vision is blocked by foliage when you can see in the video I can draw a LOS line with the LOS tool to the zone where the PzIV was? Claiming the Shermans LOS was blocked (ie. it never had "potenial LOS") whereas the PzIV LOS wasn't, contradicts all that has been said about the supposed reciprocal nature of LOS in CM.

I understand that the LOS tool itself is a quirky tool anyway like in this situation. Without a target to snap to, the LOS tool seems to pick a point at what seems to be close enough to ground level, rather than at some distance above the ground where a target would actually be. This can be seen when you might hover the LOS tool over a wheatfield/long grass. LOS may appear to be blocked but if you check the LOS to the same spot which now has a physical target to snap to, LOS will be shown to exist.

Short of trying to recreate the situation manually by editing the scenario and adding the tanks to the map as close as possible to where they were in my PBEM (I think I will actually try to do that), I can not check to see if the Sherman would ever have spotted the PzIV or if this was even an outlier.

Link to comment
Share on other sites

How can you say the Sherman's vision is blocked by foliage when you can see in the video I can draw a LOS line with the LOS tool to the zone where the PzIV was?

It looked to me that while the cursor was over where the IV was, the Sherman didn't have LOS. Perhaps I was misjudging it. One way of helping resolve LOS questions to units that can't be seen by the unit you're interested in, but which are spotted by other units on your side is to have a unit that can spot the enemy unit plot a move order to that location. With "Show all Paths" on, you can be fairly clear as to what ground spot you're interested in from your currently-unsighted unit.

Claiming the Shermans LOS was blocked (ie. it never had "potenial LOS") whereas the PzIV LOS wasn't, contradicts all that has been said about the supposed reciprocal nature of LOS in CM.

Not so. LOS is drawn from eyeball to "point of interest". If eyeball A is blocked, but another set of eyeballs can see a part of eyeball A's ride, then that vehicle can be shot at. The IV is starting high and looking at the ground where the Sherman is sat.

That's if the IV is sat where I thought it was. I'll go back and check again. Yeah. To my eye that IV is just out of LOS. You can see the centre of the AS it's mostly in, but because it's a "spotted thing", the "Special relationship" kicks in and LOS is calculated more precisely and to the vehicle itself rather than the LOS matrix reference points in the AS. The reciprocity of LOS is only a reciprocity of calculation, not a guaranteed reciprocity of circumstance.

I understand that the LOS tool itself is a quirky tool anyway like in this situation. Without a target to snap to, the LOS tool seems to pick a point at what seems to be close enough to ground level, rather than at some distance above the ground where a target would actually be.

Yep. I think the last time anyone authoritative said anything, it was "1m" above the ground. It takes the centre point of the AS, too, not anywhere else. But it isn't necessarily a binary decision, the LOS tool recognises some conditions where the presence of a unit might allow LOS. Once a unit spots, or has been spotted, the calculations get more detailed because a special relationship is created. However, in this case, I think the canopy is blocking even the view of the ground, from the Sherman's perspective; the IV being further away from the blockage can see under it.

Short of trying to recreate the situation manually by editing the scenario and adding the tanks to the map as close as possible to where they were in my PBEM (I think I will actually try to do that), I can not check to see if the Sherman would ever have spotted the PzIV or if this was even an outlier.

That would certainly be informative. It looks like a very small patch of "can't see" though. Perhaps your opponent would consent to some turn sharing once the match is over? Or to pass his turn and password to one of the testers for a shufty?

Link to comment
Share on other sites

I am surprised by some things that are being said and claimed here with such boldness/authority/experience/affiliation to BFC. They deserve to be rigorously tested as they certainly sound bogus to me, misleading at best.

eg.

Claiming the Sherman's LOS was blocked (ie. it never had "potential LOS") whereas the PzIV LOS wasn't, contradicts all that has been said about the supposed reciprocal nature of LOS in CM.

Really? So like in my example above, if one tank can spot a part of another tank it should be able to shoot it? Lets set up probably the most basic clear cut situation one could think of to test this.

I have set up a simple map that has a building, a Sherman and a PzIV (eyeballs A). Unlike the representation of say trees/foliage, the representation of the corner edge of a building in the game is at least much more definite and clearly defined (less abstract, even ZERO abstract?) to a player as far as assessing LOS/LOF etc. I placed the PzIV beside the building so that it's rear hull peeps out from the side of the building.

2vwwdjd.jpg

I then placed a Sherman on the other side of the building. From what is graphically represented, and if CMx2 LOS mechanics work as mentioned above, then the Sherman crew eyeballs should have no trouble "seeing a part of eyeball A's ride", establishing LOS "then that vehicle can be shot at".

But what happens (can happen) instead?

The Sherman can not spot the PzIV that graphically should be clearly visible to the Sherman even at virtually point blank range (12m)!

eesrt.jpg

Even more bizarrely, it is the PzIV that can spot the Sherman! (albeit with no LOF and after I rotate it on the spot).

Here is the video of that (spots Sherman at about the 47sec mark):

PzIV sees Sherman (better to download the linked video files locally and watch, the Dropbox viewer is a bit sucky)

What is interesting is that the vision of the Sherman coincides when the driver vision slot seems to swing past the building corner (this could be purely coincidental though given the abstract nature of what CMBN LOS mechanics appears to be).

Here is a save to play around and see for yourself:

CM LOS Mechanics.bts

It is quite clear that if anything, it should be the Sherman that (both in reality and as per the claims made as to how CMBN LOS mechanics actually work):

a) readily spots the rear part of the PzIV sticking out from behind the building just 12m away

B) is (as per the claims as to how CMBN LOS mechanics works) able to readily establish LOF and shoot at the "part of eyeball A's ride" that is visible.

Lets investigate further.

OK so the Sherman can't get LOS (let alone LOF) on the PzIV rear. Lets see if we can "trick it" in to hitting the "invisible" PzIV to prove to us that the PzIV really is there waiting to be hit by the Sherman if it could establish LOS/LOF naturally by itself. By targeting a point in the distance with a LOF that would pass through the area next to the building where the "invisible" PzIV would (should!) be clearly visible to the Sherman:

Like this:

2hqdvsj.jpg

Here is the video of that (from the perspective of the "invisible" PzIV):

Hitting INVISIBLE PzIV

The fire from the Sherman is clearly able to hit the "invisible" PzIV. LOF technically CAN exist between the Shermans main gun and the portion that would (should!) be visible to it.

The PzIV really is there, just waiting to be spotted and targeted, but the Sherman TacAI can't see it.

What I have presented indicates to me that the LOS mechanics in CMBN has a very obvious fundamental flaw which prevents it from achieving the kind of realistic LOS/LOF fidelity and resolution you would expect from a game that deals with tactical level combat.

Clearly there seems to be a three way disconnect between:

a) what is graphically presented to and interpreted by the player via the graphics

B) the LOS/LOF allowed/not allowed to be established by the TacAI to potential targets

c) the actual LOF that actually/theoretically does/can exist in the game

Given how LOS/LOF can/can't work in this very clear cut case, is it really any wonder players like myself have had reason to doubt the fidelity of the CM LOS mechanics here and in other posts?

I set up another situation and swapped the Sherman with the PzIV. This time, both tanks get LOS to each other after 10sec, but apparently have no LOF to the other. (I am still yet to determine if the LOS to/from the PzIV is fundamentally any different to that to/from the Sherman).

What is graphically represented leaves no doubt that at least the PzIV should have been able to shoot at the "part of eyeball A's ride"

Check out the video:

BIZZARE LOS-LOF mechanics

Shermans view

So lets assess these words (and a host of other things I have seen posted here and elsewhere on theses forums) again:

LOS is drawn from eyeball to "point of interest". If eyeball A is blocked, but another set of eyeballs can see a part of eyeball A's ride, then that vehicle can be shot at.

Really?

With clear cut cases of odd LOS behavior like this evident in the game, it really is disconcerting and a shame that so much effort at these forums is put in to discrediting, detracting and talking down anyone who dares question the accuracy/fidelity/realism of the LOS mechanics in CMx2 (or virtually any other feature in the game). As a player I don't want to be BSed to. I just want a warts and all understanding of what this game is meant to be/not be and how it is supposed to work/not work.

I probably should have carried out these tests the first day I started playing CMBN. They speak more truth about how CMBN LOS mechanics work/don't work than many of the words being thrown around in this thread and in other threads by individuals who frankly should know better.

That would certainly be informative. It looks like a very small patch of "can't see" though. Perhaps your opponent would consent to some turn sharing once the match is over? Or to pass his turn and password to one of the testers for a shufty?

My opponent has abandoned this game and refused to tell me his password or co-operate in other ways so as I could investigate these issues further. Perhaps if the request came from BFC he might assist in that way but I don't think he cares too much about trying to improve this game.

The saved game above however actually has the section of map and the two tanks positioned as close as I could to where they were in the PBEM. I have not yet got them to be exact so as the LOS tool feedback in the PBEM is 100% identical, though it's very close. I might still keep trying to fiddle around. Regardless, what I find can happen is just as extraordinary. I might post about this later.

PS: If the same types and findings of the LOS/LOF tests I conducted are "old news", please, let me know and where/when it was described.

Link to comment
Share on other sites

I believe the building situation that you show above might already be known because there is a situation with LOF and buildings, although I'm not certain if that's related or not. I believe that spotting goes from center of action spot to center of action spot so that's probably why you got the result that you got, although Steve has said on multiple occasions that LOS is always reciprocal so there must be something involving the positioning of the spotters somehow. It does look very odd though.

I do think that it would be helpful if you continued testing spotting, not only to help BFC to see if spotting can be improved or not, but also to help you get a better handle on how it works within the game. The more you nail down the spotting routines in the game the better positioned you will be to work around those limitations when you are playing. Hopefully that would reduce your frustration since you would then be aware of the limitations and how they affect you in the game.

Link to comment
Share on other sites

^^^

LOL. TLDR. At least, not every post.

LOS is abstracted. There are DEFINITELY situations that can/should be improved. Your PzIV butt-out is a perfect example.

Here's my understanding of how the LOS system works.

- Pre-game, the game creates a table of all POSSIBLE two-way LOS. In a simplistic model, let's call it a 10x10 grid, A->J and 1->10. Checking every possible LOS from one grid square to another (assuming reciprocity) gives 100 locations, each checking LOS to the other. Using a Sum function for the series 1-99 (the location does not check LOS to itself so only 99 pairs in the 100 location grid), gives 4,950 possible LOS's from one location to another. (The last location, J10 does not need to check LOS to any location, since every location has already checked LOS to J10. That's the benefit of assuming total reciprocity for this example.)

If A1 can see J10, then J10 can see A1. At least, they POSSIBLY can. More on that, later.

Let's just assume that every other LOS is blocked in that grid. (Go with it, please.)

The game creates a table of 4,950 possible LOS pairs. In my example, only ONE pair has a (possible) clear LOS. This is the lookup LOS table. (It is more complex than that. Obviously.)

- In-game, the engine now has a pre-made LOS table. That's all it ever accesses, the lookup table.

- During your turn, you want your unit at B7 to area fire at F9. The game sees that B7 to F9 are the locations. It checks the lookup table. "Ahh," it says, "that LOS is blocked." No LOS, no LOF.

- Next turn, you try A1 to fire at J10. The game checks the lookup table. "What ho? There MAY be LOS", it thinks to itself. It looks around at the grid squares which underlay the diagonal from A1 to J10. (That forms part of the table.) Smoke in B2? The POSSIBLE LOS is now blocked.

You get no LOS. (If there were no smoke, etc., along the grids which the table has entered for possible LOS blockage from A1 to J10 (B2, C3, D4, etc, as well as the points of the other squares, such as A2 and B1 since they geometrically touch the corner of the diagonal line), then you would have LOS.)

- The table is pretty complex. (Assume A1 and J10 have a base height of 50m. A 2 floor building is at E5, base height 37m, roof height 51m. Now the LOS is provisional based on the status of the roof at E5. It's in the table...)

- Okay, so you've got POSSIBLE LOS. Now, the game looks at the ELOS locations. Is your tanker at 0m, .5m, 1m, 2m, or 3m elevation above the grid? It also checks more finely within the action spot, to see if anything blocks the POSSIBLE LOS.

That's how it works. Using 8m grids to run the initial table calculation is an APPROXIMATION. Couple that with unit locations being accurate to less than 1m, and you have more than 64 sub-locations within each action spot.

Remember, only the whole 8m action spot has been accounted for in the lookup table, not every sub-location. A 4km x 4km map has 250,000 action spots. (More for tactically important locations like each side of a wall.). My example used 99 pairs. Using a 4km map gives...31 BILLION pairs. Actually, more. That's with 8m resolution. (Using 1m resolution would have over 128 TRILLION pairs.) Every pair needs to be assessed, including every location under and adjacent.

The game assumes a unit is at a point location for LOS purposes. If it has an LOS, then the entire model is used for intersection purposes to determine hits, etc.

In your PzIV example, the PzIV is considered in an action spot behind the building. A "point" location for the vehicle. Since the Sherman, no matter the ELOS or status of the crewmembers, is in a location that cannot see through the building, then no LOS can exist. Hence, no matter how much of the PzIV sticks out, the Sherman cannot see it. The lookup table "blinds" the Sherman.

Now, does that mean this is a GOOD system?

It's a COMPROMISE.

I'm told that computing the lookup table to a more fine resolution goes up geometrically and would bog down most computers for hours...or longer.

I'd LOVE to have the lookup table resolved to a 1m grid. Hell, I'm lying: I'd like .1m. That way I can shoot that guy in the leg when he's not quite in cover. That's me.

I'd also like the lookup table to apply an algorithm which takes into account terrain deformation. If I blow a crater in the top of the ridge, shouldn't that help my guys?

Vegetation, buildings, rubble, terrain, and elevation all play into the complexity.

I hope part of this was useful.

Ken

Link to comment
Share on other sites

Ken, do you know that for sure, or is it just your guess as to how the LOS system works?

I'm not a programmer nor mathematician, but surely it would be easier just to do checks for each soldier/vehicle about what it can see, using a trace line, instead of checking up each square of terrain against every single other?

As for Lt Bull's very thorough testing: It seems the problem is that vehicles are treated as points in space for spotting. If the magic point is hidden behind a building or even a tree, the whole tank counts as hidden.

The point should be replaced by a hitbox, either cubic or (ideally) the exact shape of the tank itself, if it's technically feasible.

Link to comment
Share on other sites

Really? So like in my example above, if one tank can spot a part of another tank it should be able to shoot it? Lets set up probably the most basic clear cut situation one could think of to test this.

Sorry. Lack of clarity on my part there.

I'm pretty sure I've already said that I don't think every plate of a vehicle is modelled as a "potential point of aim". Your old example where the two track fenders "should" have been visible either side of the tree trunk, but because the centreline was invisible, the opposition couldn't get a LOF/LOS is an example of that.

Unlike the representation of say trees/foliage, the representation of the corner edge of a building in the game is at least much more definite and clearly defined.

Yup. You're running into another problem, though, and that is the initial LOS/LOF check to see if it's even possible to draw LOS to that square. That is drastically affected by buildings. Since the Sherman cannot see the centre of that square where the PzIV is (because of the awkwardly non-abstract building), there aren't any checks for LOS made by the engine.

AIUI, this problem is somewhat compounded by the artificial setup: if there was some shared spotting info about the PzIV, or if the sherman had seen it before, there would be a "special relationship" where the spotting is calculated unit-to-unit, not assessed to the AS.

What is interesting is that the vision of the Sherman coincides when the driver vision slot seems to swing past the building corner (this could be purely coincidental though given the abstract nature of what CMBN LOS mechanics appears to be).

The PzIV has got a "tentative contact" of the Sherman by that stage, which, AIUI, is enough to create the "special relationship", meaning that the driver can get a spot on the Sherman. Or maybe the geometry just means that the driver's far enough across to see the centre of the Sherman's AS, so it gets made eligible to spot without needing the "special relationship". It's 100% clear that when the tank rotates back to point straight at the Sherman, the driver's vision block is pushed back behind the corner of the building and the Sherman is no longer visible. Which you can take how you like, but I'd say is a pretty good indicator that some eyeballs saw something pretty much as soon as they could.

Having played with the save game a bit, if you Slow the Sherman forward in short legs with Pauses, the Sherman and PzIV spot each other at exactly the same time. Then the Sherman kills the PzIV.

As a player I don't want to be BSed to.

No one's BSing you.

I just want a warts and all understanding of what this game is meant to be/not be and how it is supposed to work/not work.

Then stop looking for edge cases where the mechanism fails?

The reason I take exception to your characterisation of the mechanics as "unrealistic" is that you're ignoring the 99.99% (or more) of the time when the mechanics produce at least plausible results. And the fact that of those times when you find an edge case where it doesn't represent exactly what it seems to, the actual end result is often pretty close to what you'd hope would be the case (in my fiddling with that save game the Sherman killed the PzIV that had its arse hanging out in 100% of cases, for example).

It's not perfect. The AS LOS matrix is borked when it comes to house-to-house fighting. But it's on the whole pretty realistic. Yes, it's possible to synthesise cases where the LOS mechanics fail. Some people's play style seems to generate them more than others, but even then , outside of test situations, there have been a hundred thousand checks for every one that ends up looking problemmatical and only one in 10 of those actually is. So throwing up your hands and exclaiming that "It's not realistic" is a gross overreaction.

Link to comment
Share on other sites

Ken, do you know that for sure, or is it just your guess as to how the LOS system works?

I'm not a programmer nor mathematician, but surely it would be easier just to do checks for each soldier/vehicle about what it can see, using a trace line, instead of checking up each square of terrain against every single other?

As for Lt Bull's very thorough testing: It seems the problem is that vehicles are treated as points in space for spotting. If the magic point is hidden behind a building or even a tree, the whole tank counts as hidden.

The point should be replaced by a hitbox, either cubic or (ideally) the exact shape of the tank itself, if it's technically feasible.

I know that the game makes a pre game LOS calculation from action spot to action spot. Steve has said on numerous occasions that individual spotting 'on the fly' by each individual unit in game is so computer intensive that it's simply not feasible so you can just forget about any fancy theories anyone may have as to what they want the game to do. What some want and what is possible are often two different things. The problems come when those who want something get confused about whether it is possible or not through a lack of appropriate knowledge of the issues from a computer standpoint. The forum is littered with hundreds of posts from those who say "I want X and BFC is incompetent because they can't just implement something so simple." Typically those who say these things have no idea what they are talking about. ;)

Link to comment
Share on other sites

Ken, do you know that for sure, or is it just your guess as to how the LOS system works?

As, @ASL Veteran says Steve has talked about this a few times. The issue with LOS checking is how much work it takes the computer to do. So having a list of locations that can possibly see another location means they can not waste time checking if your guy in the valley can see an enemy in the next valley. The idea is to reduce the amount of checking and only make the computer check for the units that could possibly see each other.

Even that takes a lot of computing time. To make the game run on a wide range of machines not every unit gets to check what it might see every second of a turn. That is where the 7s spotting cycle comes from (someone spend a bunch of time experimenting to find that number but I don't think that it has been confirmed by the programmers although the concept of only some units get to check for new things they can see at each opportunity has been confirmed). Even that cycle has been tweaked to help with some of our spotting issues. Steve has said that there is now code that gives some priority to units in close proximity to the enemy so your guys near the enemy will get more opportunities to see if they spot something new.

We are not sure of the precise cause of what @Lt Bull is seeing in this case but it sure fits with what Ken described. So, several of us are not speaking with actually experience of the inner workings but from the explanations of Steve and his staff over the years. He cannot be here and answer every post so some of us try to help out with the explanation part. And Ken and I think @ASL Veteran have tester status (I hope I am right on that and not outting anyone) so they can take saves from us and test things out before submitting defects to the developers.

Link to comment
Share on other sites

^^^

LOL. TLDR. At least, not every post.

LOS is abstracted. There are DEFINITELY situations that can/should be improved. Your PzIV butt-out is a perfect example.

Here's my understanding of how the LOS system works.

- Pre-game, the game creates a table of all POSSIBLE two-way LOS. In a simplistic model, let's call it a 10x10 grid, A->J and 1->10. Checking every possible LOS from one grid square to another (assuming reciprocity) gives 100 locations, each checking LOS to the other. Using a Sum function for the series 1-99 (the location does not check LOS to itself so only 99 pairs in the 100 location grid), gives 4,950 possible LOS's from one location to another. (The last location, J10 does not need to check LOS to any location, since every location has already checked LOS to J10. That's the benefit of assuming total reciprocity for this example.)

If A1 can see J10, then J10 can see A1. At least, they POSSIBLY can. More on that, later.

Let's just assume that every other LOS is blocked in that grid. (Go with it, please.)

The game creates a table of 4,950 possible LOS pairs. In my example, only ONE pair has a (possible) clear LOS. This is the lookup LOS table. (It is more complex than that. Obviously.)

- In-game, the engine now has a pre-made LOS table. That's all it ever accesses, the lookup table.

- During your turn, you want your unit at B7 to area fire at F9. The game sees that B7 to F9 are the locations. It checks the lookup table. "Ahh," it says, "that LOS is blocked." No LOS, no LOF.

- Next turn, you try A1 to fire at J10. The game checks the lookup table. "What ho? There MAY be LOS", it thinks to itself. It looks around at the grid squares which underlay the diagonal from A1 to J10. (That forms part of the table.) Smoke in B2? The POSSIBLE LOS is now blocked.

You get no LOS. (If there were no smoke, etc., along the grids which the table has entered for possible LOS blockage from A1 to J10 (B2, C3, D4, etc, as well as the points of the other squares, such as A2 and B1 since they geometrically touch the corner of the diagonal line), then you would have LOS.)

- The table is pretty complex. (Assume A1 and J10 have a base height of 50m. A 2 floor building is at E5, base height 37m, roof height 51m. Now the LOS is provisional based on the status of the roof at E5. It's in the table...)

- Okay, so you've got POSSIBLE LOS. Now, the game looks at the ELOS locations. Is your tanker at 0m, .5m, 1m, 2m, or 3m elevation above the grid? It also checks more finely within the action spot, to see if anything blocks the POSSIBLE LOS.

That's how it works. Using 8m grids to run the initial table calculation is an APPROXIMATION. Couple that with unit locations being accurate to less than 1m, and you have more than 64 sub-locations within each action spot.

Remember, only the whole 8m action spot has been accounted for in the lookup table, not every sub-location. A 4km x 4km map has 250,000 action spots. (More for tactically important locations like each side of a wall.). My example used 99 pairs. Using a 4km map gives...31 BILLION pairs. Actually, more. That's with 8m resolution. (Using 1m resolution would have over 128 TRILLION pairs.) Every pair needs to be assessed, including every location under and adjacent.

The game assumes a unit is at a point location for LOS purposes. If it has an LOS, then the entire model is used for intersection purposes to determine hits, etc.

In your PzIV example, the PzIV is considered in an action spot behind the building. A "point" location for the vehicle. Since the Sherman, no matter the ELOS or status of the crewmembers, is in a location that cannot see through the building, then no LOS can exist. Hence, no matter how much of the PzIV sticks out, the Sherman cannot see it. The lookup table "blinds" the Sherman.

Now, does that mean this is a GOOD system?

It's a COMPROMISE.

I'm told that computing the lookup table to a more fine resolution goes up geometrically and would bog down most computers for hours...or longer.

I'd LOVE to have the lookup table resolved to a 1m grid. Hell, I'm lying: I'd like .1m. That way I can shoot that guy in the leg when he's not quite in cover. That's me.

I'd also like the lookup table to apply an algorithm which takes into account terrain deformation. If I blow a crater in the top of the ridge, shouldn't that help my guys?

Vegetation, buildings, rubble, terrain, and elevation all play into the complexity.

I hope part of this was useful.

Ken

Having tried to understand, and taken a brief part in, some of the other threads referred to in which Steve has explained this issue, this is a pretty darn good explanation and summary of what I understand the position to be.

It also gets my vote for the most useful post of 2014; heck, probably of the decade so far!! :D

Link to comment
Share on other sites

Yeah, a bit simplified. In actuality, I believe that if the POSSIBLE LOS exists, then it is checked with the unit in-game, in the manner I mentioned. It uses actual action spot location (not a generic holder for the 8m square), as well as ELOS for both parties in the LOS check. This is where the size of the vehicle matters. Collision detection applies for 3D items, with some fuzzy logic for vegetation, as well as vehicles. This is the "trace the ray" kind of logic you'd expect. It's done -as needed-, not continuously, and not everywhere. The lookup table provides the information for when LOS checks between units need to be checked. Spotting, the ability of the UNIT to use the LOS, is done at intervals. Again, to save cpu cycles.

If the "owning" action spot (the one the game says the vehicle is in) is a non-clear LOS per the lookup table, nothing will allow your unit to see the enemy vehicle. Your location and his location have a blocked LOS.

In the PzIV behind the building example, I assume the game has assigned the PzIV to be in the action spot which is blocked. Therefore, no LOS.

Hopefully there will eventually be a workaround so that the PzIV would get spotted.

Ken

Link to comment
Share on other sites

Putting aside some of the more negative and emotional comments - I think some good points have been raised and I have a better understanding of the spotting system. So I'd just like to say thanks to those guys who have taken the time to provide examples and reply with some constructive comments. My 2c is: I think BF have reached a good compromise for most situations. However, I can understand how it could be more frustrating if you like playing tank battles in non-open terrain e.g bocage or around buildings. For me I'd rather the occasional wtf moment then waiting 10mins for a turn to resolve los calculations :). Also, I think BFs track record over the last ~15 years (has it really been that long? omg!) has been for continuous improvement -so I think things will only improve over time.

cheers

Link to comment
Share on other sites

I have often wondered if some of the edge cases could be removed by enabling units to "deploy" in such a way that they spend x amount of time removing obstacles from their line of sight. For example in RL, it is doubtful that the Sherman commander in the original post would have deployed in such a way that large portions of his line of sight were blocked by overhanging branches. Similarly, the other thread on ridgeline deployments.

Link to comment
Share on other sites

^^^

What if I set up a unit behind a wall, say, and want them to wait out of sight. If they start removing stones from the wall, soon enough they'll be in the enemy's sight and die.

A lot of frustration would/could result from units autonomously changing their cover/concealment status.

Having said all that, yeah, there is room for improvement.

Link to comment
Share on other sites

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.

Guest
Unfortunately, your content contains terms that we do not allow. Please edit your content to remove the highlighted words below.
Reply to this topic...

×   Pasted as rich text.   Paste as plain text instead

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.


×
×
  • Create New...