Hull-down spotting disadvantage

[Childress]

Seems pretty serious. What's the verdict? Stick to infantry battles? Get back into Free Cell? Go ahead and mock but I thinking of buying Nancy Drew: Curse of Blackmoor Manor. Simplistic, yeah, but you gotta love they way they handle the terrain mesh.

[A Canadian Cat]

Good and interesting tests. My thoughts on this... even if there is a bug that doesn't give hull down the full spotting advantage it deserves, hull down is still a viable tactic with CM and you would be foolish to ignore it.

Seems pretty serious. What's the verdict? Stick to infantry battles?

Naw - I'm with Bill. Room for improvement - sure, use of vehicles broken - no.

Play the game and have a good time. Report back here if you see something odd (or awsome) and then go back and play the game some more.

[Sgt Joch]

Interesting results. At first glance it would appear that tanks in the open get an advantage because they have all 5 crewmen spotting while "Hull Down" tanks only have 3 since the driver/co-driver have a blocked LOS, although there might be another reason. Definitely worth looking into.

I presume all tests were done with buttoned up tanks? I did a very quick and dirty test with unbuttoned tanks which showed little noticeable difference in spotting times between tanks hulldown or in the open.

[Childress]

Play the game and have a good time.

Which game? Getting mixed results with the Nancy Drew demo. In one play through I was able to place Nancy and her slingshot hull down behind the abbey and take out the evil Nigel Mookerjee on the second shot. (the first bounced off the glacis) But entering the Manor then she got eaten by Mrs. Drake's carnivorous plant.:mad: The highest level I got to was Talkus Multissimus: For the relentless interrogation of suspects. Overall verdict: Immersive! But intimidating learning curve.

[Battlefront.com]

OK, so looking at this thread I am seeing two things coming out of it:

1. A significant hulldown position does dramatically reduce the chances of being hit. Significant defined as covering most of the lower hull.

2. Tanks with 5 crew able to spot are at an advantage vs. a vehicle with only 3 crew able to spot.

As for #1, this obviously makes sense. A small amount of hull down won't make much of a difference because the shooter is going for center mass and something like 1m berm is irrelevant. A 2m berm is significant because now the shooting unit has roughly 50% less to shoot at.

As for #2, in theory it should work that more eyeballs produces better results. Just like being buttoned or unbuttoned should have an affect, as should moving or stationary. The question should be if the driver/radio operator are contributing too much to the overall spotting picture not if they should be factored out completely. I'm wondering if the internal weights favor information from the driver/radio operator more than they perhaps should.

Steve

[wolf66]

Don't mind the man behind the curtain

[c3k]

FWIW, I've spent the day working on a test for this. Yep, I've thrown out the first several because I did 'em wrong. (That's my arrogance, I guess. )

I finally had it...until I realized I'd borked the map. So: working since 7 this morning and no results yet. But getting closer to a proper test.

It is being looked at.

My perspective: I don't care a whit about hit or kill chances...yet. This is purely a test of spotting hull down vs. non-hull down. That takes care of number 1. Number 2, I've gotten rid of extra eyeballs, so it'll be apples to apples. After I test apples, it'll be bananas, then on to oranges. Then, of course, the range will have to be changed. Then the button up status. Etc. It'll be a few days to get something which can withstand peer review. (Fair warning: I will NOT post the results here. Sorry. I'm not using v2.12, which is probably what most of you are up to. )

Was I allowed to say that?

For HistoryLover (whose thread was closed), Kulik, VaB, and the rest: all your positive contributions have been heard, weighed, and taken into the balance. If there is something off with hulldown, let's start with the spotting. Then we can work to the hitting. Finally, we can finish with the killing.

I'm no testing guru, but I'm posting this so you know that a beta tester has eschewed any more testing of the incredible REDACTED BY MODERATOR, and am instead working diligently on this. If you have input to SPOTTING testing with hulldown status, please let me know.

Regards.

Ken

[Vanir Ausf B]

As for #2, in theory it should work that more eyeballs produces better results. Just like being buttoned or unbuttoned should have an affect, as should moving or stationary. The question should be if the driver/radio operator are contributing too much to the overall spotting picture not if they should be factored out completely. I'm wondering if the internal weights favor information from the driver/radio operator more than they perhaps should.

This is certainly the crux of the matter, but there is one other factor: how much more difficult should a hull down tank be to spot. It's not entirely clear if it makes any difference in the game presently. My testing suggests a small difference in spotting times vs. fully hull down tanks, but it is so small (about 16%) it may be statistical noise. Common sense says a hull down tank should be harder to spot, but whether it should be 10% harder or 50% I don't know.

[Vanir Ausf B]

I presume all tests were done with buttoned up tanks? I did a very quick and dirty test with unbuttoned tanks which showed little noticeable difference in spotting times between tanks hulldown or in the open.

Yes, all my tests were with tanks buttoned.

[Vanir Ausf B]

I'm no testing guru, but I'm posting this so you know that a beta tester has eschewed any more testing of the incredible REDACTED BY MODERATOR, and am instead working diligently on this. If you have input to SPOTTING testing with hulldown status, please let me know.

Sorry, Ken. This is probably a lot less fun than what you were doing

One other thing you may want to look at is the "hull down" and "partial hull down" text on the LOS tool. It seems to lie sometimes, as AKD and myself noted earlier in the thread.

[PhilM]

This is certainly the crux of the matter, but there is one other factor: how much more difficult should a hull down tank be to spot. It's not entirely clear if it makes any difference in the game presently. My testing suggests a small difference in spotting times vs. fully hull down tanks, but it is so small (about 16%) it may be statistical noise. Common sense says a hull down tank should be harder to spot, but whether it should be 10% harder or 50% I don't know.

How hard it is in absolute terms to spot the still visible portion of the hull down tank I guess would depend on factors like: is it silhouetted against the skyline (not good!); is it turned at an angle to the spotter, such that the "unnatural" aspect of the gun tube is exposed; etc. But if those sorts of factors are assumed to be the same for both the hull up and the hull down samples, isn't it reasonable to equate the changed spotting probability to the portion of the cross section on show that is that still visible versus that being masked? That is, if the hull down tank is front on, and the turret forms, say, 25% of that cross section, then the probability of being spotted when hull down is reduced by 75% of the original value, equating to the view of the hull that is being lost?

And the point Steve made about the internal weights given to the different sets of eyeballs does seem to be a likely cause of any issue, if there is one. I don't know if they do now start off at equal shares / precedence? Probably not I assume, as the TC and gunner would have the larger share. But if the driver and W/Op - gunner currently have, say, 30% of the spotting weighting (depending on the view direction), then to lose that in a hull down position is a significant reduction. But if that should only ever have been 10% of the total to begin with, then hull down should leave the tank with 90% of its original spotting capability?

Figures of this latter sort of order sound more plausible, on the basis of hull down positions being so favoured. Sure, it gets you cover: but if in the process you also significantly reduce your ability to spot targets in the terrain to which you are going hull down, and allow them to have the same - or more - chance of spotting you than you have of spotting them, then why would you bother?

Perhaps you would even lose some rear view in this position? On the basis that the TC, knowing there is no driver / WOp forward view, will replace that by more forward (hull down direction) spotting of his own, at the expense of less all round/rear spotting, which only he has the capability for?

Complicated, this ...

[c3k]

Sorry, Ken. This is probably a lot less fun than what you were doing

One other thing you may want to look at is the "hull down" and "partial hull down" text on the LOS tool. It seems to lie sometimes, as AKD and myself noted earlier in the thread.

Yeah, that seems to be an oddity. I've read and re-read that part of the thread several times. First, I need to satisfy myself that I can create an adequate baseline.

[John Kettler]

BFC et al.,

I tried posting this before, but the BB gods saw fit to exile me to Login, many times, while I was trying to post. And, yea, did they eateth my post withal! Once more with feeling!

Here is "A Review of Literature on the Theory of Visual Target Detection Probabilities." It's a 1973 master's thesis for an M.S. in Operations Research at the Naval Postgraduate School at Monterey, California. It provides a complete definition of the terms relevant to the topic, looks at the various mechanisms associated with optical detection, surveys the literature, analyzes the similarities and disparities among various field tests, lab tests and the associated models developed in consequence. Of particular interest is Figure 6 (p. 31) which shows that Pd (Probability of detection) is indeed a function of increasing target size. An intuitive result.

https://archive.org/stream/reviewofliteratu00macc#page/n1/mode/2up

Regards,

John Kettler

[Battlefront.com]

FWIW, I've spent the day working on a test for this. Yep, I've thrown out the first several because I did 'em wrong. (That's my arrogance, I guess. )

Well, it's why we "hired" you

Thanks for putting in the time. Even when you guys test the crap out of something (with focused, disciplined tests) and find nothing, it's valuable. Sometimes there is a problem still lurking about, but a test that "proves nothing" at least shows us where the problem isn't.

This is certainly the crux of the matter, but there is one other factor: how much more difficult should a hull down tank be to spot. It's not entirely clear if it makes any difference in the game presently.

AKD's tests seem to indicate that it does matter if the vehicle is behind a discernible piece of terrain, such as a wall or hedge. Where it might not matter is if it is just a lump of ground. Obviously in theory it should, so it will be interesting to see what further tests produce.

Common sense says a hull down tank should be harder to spot, but whether it should be 10% harder or 50% I don't know.

The problem is that real life is not so neat and tidy. Angle of light, amount of intervening cover, air quality, what's behind the hull down vehicle, how obvious the spot might be (think Monty Python's skit about how not to be seen), how good the eyesight is of the spotting crew, etc.

Some of these things can be accounted for to some degree, others are too difficult to effectively simulate. For example, if there is one lump in the middle of a 10km2 field you are definitely going to check that out first and carefully. Might even shoot it just in case. But if there are hundreds of places something could be hiding you're probably SOL.

Definitely difficult to deal with.

Steve

[Vanir Ausf B]

Yeah, that seems to be an oddity. I've read and re-read that part of the thread several times. First, I need to satisfy myself that I can create an adequate baseline.

Not sure what sort of baseline you are looking for, but here is a save file (2.12 of course) that demonstrates the problem in several different ways.

https://www.dropbox.com/s/zfvho3ef2hyxg7b/That%20lying%20LOS%20tool.bts

[altipueri]

Yes, it is actually quite important, statistically at least, that the same experiment be repeatable by other people using the same methodology. That is why the same version e.g. 2.12 should be used, and that the data and the results be shown. Otherwise there are more variables being introduced which muddies rather than clarifies things.

In the meantime Vanir can dust off his stats books and remind himself how to do the Chi Squared test or whichever one it is that can test whether two or more sets of data results come from the same population.

[Vanir Ausf B]

From the locked thread.

I'd be really curious about the results of this testing ...

I'm currently in a large CMBN match with nothing but PzIV's against Shermans. The map is huge and the engagement distances are very long range.

I'm getting the floor wiped by Shermans who spot my PzIV's, even sitting in wood lines and I have yet to see one of them first. In fact, I'm lost 2-3 PzIV's to one Sherman that was actually sitting out in the open picking them off. It wasn't until 2 turns later until a Jagd crew elsewhere on the map spotted it. I realize it's an anecdotal observation, but it's consistent in this particular match.

I thought the German tanks were supposed to have superior optics that shined at long ranges for spotting and accuracy...

Regards,

Doug

I'm not sure I can provide a definitive answer. Here are my test results.

All tests done at 1200 meters. All tanks buttoned and in the open. 10 tanks per side per test run. Each test run 10 times (except Sherman vs Sherman which was run 50 times), so 100 iterations per side per test.

Sherman 76 vs Sherman 76

Average spot time: 106.1 seconds. Standard deviation = 100.5

M4A3w Sherman 76 vs Panzer IV H late

Sherman average spotting time: 123.5 seconds. Standard deviation = 105

Pz IV average spotting time: 119.4 seconds. Standard deviation = 117.6

M4A3w Sherman 76 vs Panther A late

Sherman average spotting time: 111.5 seconds. Std dev = 107.4

Panther average spotting time: 139.1 seconds. Std dev = 138.5

If I tally the number of "wins" by comparing which tank spotted the other first in each test run we get:

Sherman 76 wins: 49

Panzer IV wins: 51

Sherman 76 wins: 58

Panther wins: 42

I would not put too much faith in those numbers given the huge standard deviation. Also, the Sherman 76 has a 6x gunners sight which is better than the 3x sight on the Sherman 75 and most other Allied tanks. Then again, the Panzer IV gunners sight is only 2.5x :confused: The sight on the Panther can be switched between 2.5x and 5x.

But I do think there is enough here to conclude that whatever German optics advantage there may be in the game is too small to be tactically significant.

For anyone who's interested, this is the raw data from the Panther test. Spotting times in seconds.

261  297  40  93  84  162  19  92  250  366

29  27  10  74  329  38  411  148  85  26

380  76  125  98  30  624  77  216  91  15

32  133  342  74  374  29  37  104  28  115

43  5  48  24  27  26  2  122  50  20

484  101  313 13  279  53  152  101  466  54

146  54  36  188  179  444  1  115  60  116

91  93  212  56  244  200  147  217  323  44

123  71  666  60  80  233  276  35 19  76

33  18  96  153  80  72  281  6  27  317

[BadgerDog]

From the locked thread.

I'm not sure I can provide a definitive answer. Here are my test results.

Thank you partner.... very gracious of you and I really appreciate the effort you've put in to address my question ...

To be honest, regardless of how bad my little virtual electronic units are in this game at spotting opponents tanks, it's nothing as compared to how bad I was in real life, sitting in the cupola of an M4 Sherman trying to spot other Shermans "hull down", or AT guns (even worse) at any distances over 500 yards....

Most of the time I didn't see them until their turrets moved to direct fire on me, or the there was a puff of smoke from the AT gun as it laid the lumber to me. :eek:

So, I should be thrilled I've got better with age, even if it is a virtual battlefield...

Regards,

Doug

[c3k]

Not sure what sort of baseline you are looking for, but here is a save file (2.12 of course) that demonstrates the problem in several different ways.

https://www.dropbox.com/s/zfvho3ef2hyxg7b/That%20lying%20LOS%20tool.bts

Good laugh from the savefile name. Thanks.

[A Canadian Cat]

As for #2, in theory it should work that more eyeballs produces better results. Just like being buttoned or unbuttoned should have an affect, as should moving or stationary. The question should be if the driver/radio operator are contributing too much to the overall spotting picture not if they should be factored out completely. I'm wondering if the internal weights favor information from the driver/radio operator more than they perhaps should.

Certainly one very important factor - and I am glad this is getting attention. The other part is:

How hard it is in absolute terms to spot the still visible portion of the hull down tank I guess would depend on factors

These two clearly interact. How hard it is to spot the turret of a tank vs spotting an entirely visible tank will interact with how many people are trying to find each.

I am glad all this is on the table but please consider that if there is indeed a problem with the relative weighting of the crew's eyeballs for the purposes of spotting don't forget that the difficulty to spot what they are looking for should also not be equal. I should be significantly more difficult to spot a hull down tank than it is to spot an exposed tank even with equal eye balls.

My gut says that the difficulty in spotting a portion of a tank vs the whole thing should totally dwarf the difference between 2, 3 or 5 sets of eyes looking. By that I mean my gut would be that even 5 sets of eyes looking for the hull down tank would be slower than 2 sets of eyes looking for a whole tank.

Sounds like Ken is on the case and looking at the issue. Thank you.

[altipueri]

I said this would drive you mad Vanir.

That Panther data set looks, well, skewed with the standard deviation almost equal to the entire less than average range. (In fact they are all a bit weird; were you smoking something?)

Maybe the answer is Poisson - but sounds a bit fishy to me.

[Sgt Joch]

We also have to remember that we are looking at a very narrow set of variables, i.e. a buttoned up tank suddenly appearing 800 meters in front of a hull down buttoned up tank. At that point, it is just a question of whether 5 sets of eyes looking directly at a small stationary target outweigh 3 sets of eyes looking at a bigger stationary target.

In a more typical scenario, the hull down tank would be unbuttoned and the tank in the open would have been moving either directly towards or across the front of the hulldown tank. In both scenarios, the non moving tank should spot the moving tank first, since "movement" has a big impact on spotting.

That probably explains why this issue has never been noticed before even though the basic code has been there since 2007.

[Battlefront.com]

Very true that CM is not a game that is easily tested because of the wide variety of data available to test. The Sherman is a great example. Anybody that says "Shermans do better than PzIVs" needs to back away from that statement as quickly as they make it There's a huge difference between the different Shermans in terms of how well each spots. Some models don't even have cupolas!

By and large the Shermans with cupolas are roughly equal to PzIVs and Panthers. Vanir's quick tests show that at least one model of Sherman is a little worse off than specific models of PzIVs and Panthers, but not by much. That seems fine to me. Plus, we don't want to get distracted by the minutia of specific vehicles. Pick a couple specific models and test different things with them. Swapping around vehicle models is a really good way to skew certain results. Perhaps to the point of making the test too flawed to be of value.

It is also important to not lose sight of the fact that "clinical testing" often has very little to do with actual game results. Meaning, a significant issue in carefully controlled circumstances might only have a small impact on any one particular battle. Therefore it is important to be cautious about extrapolating a narrow test result with the game as a whole.

Case in point... another poster was saying that the test proves the game is totally borked based on preliminary, and demonstrably questionable, test results. Maybe there is something wrong, maybe not. Maybe it has a wide basis, maybe only narrow circumstances apply. Maybe it's the sort of thing that comes up once a game, maybe it comes up dozens of times. Whatever the case actually is, remember that it is highly unlikely that a single test will do more than scratch the surface of how games actually play out.

For those of you who can't grasp this concept, think about it in terms of dietary sciences. Quick question... is fat bad for you? Do carbs make one fat? Does red wine inhibit cancer? And what do these things mean when sitting down to a specific meal? Anybody that says a definitive yes or no to these questions has only proved how out of touch they are with reality. The same goes for tests in CM as the underlying simulation is sufficiently complex and interwoven that overly simplistic conclusions generally have little relevance on real life games.

Steve

[John Kettler]

Steve,

I've been in a Sherman 76mm recently, and I can tell you, having looked at the world from that cupola, that I see no way at all it can come even close to the view from one of those German drum cupolas. The drum cupolas have their vision openings well above the roof of the tank, whereas the Sherman 76mm has what amounts to a bump on the turret roof. While the Sherman 76mm cupola is certainly a big improvement over having none at all, I'd expect German AFVs with drum cupolas to do significantly better in the spotting department.

Regards,

John Kettler

[Vanir Ausf B]

I said this would drive you mad Vanir.

That Panther data set looks, well, skewed with the standard deviation almost equal to the entire less than average range. (In fact they are all a bit weird; were you smoking something?)

Maybe the answer is Poisson - but sounds a bit fishy to me.

At this point I NEED to be smoking something. I posted the Panther data set because it is the most extreme example, but it is actually not a-typical of spotting data sets in that it is not a normal distribution, and I think you are on to something with the Poisson reference. In all of the data sets the large majority of data points lying more than one standard deviation from the mean are on the same end of the distribution curve. So you have a cluster of data points on one end of the curve with a very gradual decrease in frequency. In a typical group of 10 spotting tanks with a mean spotting time of 120 seconds you see about 3 enemy tanks spotted in the first minute, 3 more in the second minute, 3 more in the 3rd and 4th minutes, and then one crew that takes 8 or 9 minutes to spot anything.

That the Panther tank appears to be somewhat of a dog in the spotting contest when it should be the star of the show is disturbing, but I'm too burnt out on this at the moment to chase it down.