Infantry Don't Bennfit From Low Walls

[Childress]

Downloaded your file, sburke. It was kind of a trip. You're a wild and crazy guy.

Yes, it appears something is amiss. Or at least, counter-intuitive.

[Battlefront.com]

Some good stuff here. From what I see there's ample evidence that low walls do, in fact, reduce casualties. At least sburke's tests on the previous page show better results from being behind a wall than no wall at all. Is it enough or making sense in all situations? Not sure yet.

But I think the tests here need to have more variables isolated. As AKD said a few pages back, walls are HIGHLY situational. The best results behind a wall come from when the enemy is the same height or lower. Put the enemy higher up and cover drops depending on height and distance.

1. If you're using the same types of units on both sides (i.e. MGs, Rifle Squads, etc.) make sure you play Blue/Blue or Red/Red with the same exact type of unit on both sides. This eliminates variables such as rifle grenades, Panzerfausts, different numbers of shooters, different mix of weapons (though there is some variation possible even within the same type), etc.

A pretty typical problem with tests is forgetting to account for unrelated variables. Here's how I would set up a test:

2. Make sure both sides are using the same "soft factors", such as Experience, Leadership, Morale, etc. Variations can cause major differences in outcomes. We recommend REGULAR for Experience and the middle settings for any of the others. Don't use TYPICAL because that varies things more than selecting a specific level. Setting too high or low will skew the data.

3. Make "test lanes" within a single map so that units from one can not engage anybody other than the intended target. This is extremely important because weight of numbers is variable that has to be kept out of this equation.

4. Play "hotseat" so you can rule out TacAI issues and have play back to count and better check cause/effect.

5. Play with in any Mode other than Basic Training. This ensures FoW isn't an issue.

6. Make sure to isolate any HQs so no unit has the benefit of C2. If you do have C2 then make sure all units on both side have the same C2.

7. Put 10-20 Test Lanes per test. This allows you to run 10-20 examples each go through. Do this at least 5-10 times so there is decent sample to draw from (ideally you would do it 100 times, but lets be realistic ).

8. Run all tests for a set number of minutes, not until a unit is eliminated or what not. Pick a time which is appears to be sufficient enough to see something interesting, then run it a little longer. Run the test a few times just to figure out when this time is and then run the tests for real using that time.

OK, now for two different ways to test the open vs. wall component. Each of these should produce a slightly different end result, but either should show what walls are doing:

Even Odds

In this test you would have a Squad in the open and about 200m-300m away another Squad behind a Low Wall. Make sure both are facing each other and not able to engage, or be engaged, by other units. This shows how much a difference the wall can make in a fictional, but common, combat situation.

Isolated Odds

Put one Squad behind a wall and about 200m-300m away put a HMG Bunker. Make a second test scenario with the walls removed, but no other changes. Run the latter first and figure out how long to run the test for. Then use that time for BOTH scenarios. This will keep an apples to apples comparison of roughly x number bullets from the HMG cause roughly y amount of casualties.

Steve

[Battlefront.com]

Yes, I remember Steve stating that skylining was not simulated although a quick search came up empty.

Correct, Skylining is not explicitly simulated. However, there are natural conditions which could produce some Skylining effects. This would simply be a factor of angles and terrain, not anything explicitly simulated.

Steve

[Vanir Ausf B]

I then ran the test for 5 rounds.

You're tests are interesting. The only issue I have is I think you haven't run nearly enough iterations to get meaningful numbers. In my own tests I saw a lot of variation in results from test to test. It wasn't until I had generated nearly a thousand casualties per side that a pattern began to emerge that remained consistent.

[Vanir Ausf B]

2. Make sure both sides are using the same "soft factors", such as Experience, Leadership, Morale, etc. Variations can cause major differences in outcomes. We recommend REGULAR for Experience and the middle settings for any of the others. Don't use TYPICAL because that varies things more than selecting a specific level. Setting too high or low will skew the data.

In the new test I'm setting up I am going to make everyone fanatical. The reason is that I am testing vulnerability specifically. Cowering or hiding at the base of the wall outside LOS will invalidate the results, IMO, so I need those guys behind the wall to stay up and shooting back.

[sburke]

You're tests are interesting. The only issue I have is I think you haven't run nearly enough iterations to get meaningful numbers. In my own tests I saw a lot of variation in results from test to test. It wasn't until I had generated nearly a thousand casualties per side that a pattern began to emerge that remained consistent.

I wouldn't disagree with you there. My problem is I did notice an abnormality that had come out rather quickly and every test I have run has consistently shown that effect. The degree has varied somewhat, but only slightly. I am more than happy to run a bunch more iterations, but figured if the discrepacy was so exteme it was worth providing to better minds than mine to see if maybe something in my testing is inherently flawed or not. If it is not flawed it is demonstrating a condition that may have a fundamental impact on the question of whether a wall provides as much benefit as expected based on conditions I would not have even thought to apply. If it is flawed then I am unnecessarily spinning my wheels when I could be getting back to Courage and Fortitude.

[Flanker15]

From the tests I want to field a theory that:

(in the game) Being prone makes you harder to hit than a wall offers cover from rifle bullets. Accentuated by the problem of guys reloading while exposed at the wall (and windows) and not popping up and shooting then dropping out of los again. So the other guys can get repeat shots (hence better aim) on kneeling troops (which are easier to hit than prone guys).

The wall has more non-fatal casualties too which is probably from bullets going through the wall and doing less damage.

[sburke]

From the tests I want to field a theory that:

(in the game) Being prone makes you harder to hit than a wall offers cover from rifle bullets. Accentuated by the problem of guys reloading while exposed at the wall (and windows) and not popping up and shooting then dropping out of los again. So the other guys can get repeat shots (hence better aim) on kneeling troops (which are easier to hit than prone guys).

The wall has more non-fatal casualties too which is probably from bullets going through the wall and doing less damage.

I am not sure I have data to support that. Here are the results of 10 additional consecutive tests.

The numbers are in the open dead/wounded and behind wall dead/wounded. The final figure is the number of casualties suffered by the squads w/o the wall behind them along with the percentage of overall casualties. What I seem to be finding is there may be something else in play here that can either make the units behind the wall suffer significantly higher casualties than those in the open or significantly less. I don't think you can actually say in a simplistic sense that a wall is better or worse in and of itself.

39/35 33/51 56 66%

35/38 37/47 55 65%

33/37 31/52 53 63%

31/30 28/56 58 69%

37/39 31/48 58 73%

27/20 48/53 61 60%

31/23 50/45 67 70%

35/24 43/56 67 67%

27/28 38/42 60 75%

19/37 41/58 74 74%

Totals

314/311 380/508 609 70%

Proportionally it does appear units in the open suffer a high proportion of dead versus wounded, but that could simply be they are still exposed and are being hit by additional bullets versus the guys behind the wall who when wounded are falling behind cover.

[Vanir Ausf B]

From the tests I want to field a theory that:

(in the game) Being prone makes you harder to hit than a wall offers cover from rifle bullets. Accentuated by the problem of guys reloading while exposed at the wall (and windows) and not popping up and shooting then dropping out of los again. So the other guys can get repeat shots (hence better aim) on kneeling troops (which are easier to hit than prone guys).

The wall has more non-fatal casualties too which is probably from bullets going through the wall and doing less damage.

The wall that I have been placing men behind does not appear to be one than small arms could penetrate. If they are penetrating then the graphic is misleading or there is an issue with the ballistics of small arms.

Guys not ducking down while reloading is an issue, I think. One other possible issue is that guys at the wall do not use the wall as a gun rest to steady their aim as they would in reality.

[Battlefront.com]

Correct, rifle fire does not penetrate the walls. these are thick rock walls which would easily deflect/defeat a rifle round. .50cal might get through, but I'm not sure about that.

The theory about firing opportunities and what not is one reason to be careful about matchups. If you're comparing the results of A vs B directly, they should have as few variables between then as possible. Open vs. behind wall, open vs. open are the the two that should be at the heart of these tests.

Steve

[sburke]

Correct, rifle fire does not penetrate the walls. these are thick rock walls which would easily deflect/defeat a rifle round. .50cal might get through, but I'm not sure about that.

The theory about firing opportunities and what not is one reason to be careful about matchups. If you're comparing the results of A vs B directly, they should have as few variables between then as possible. Open vs. behind wall, open vs. open are the the two that should be at the heart of these tests.

Steve

Well I may try some other alterations to see if I can figure out if something in my tests is flawed or if I can isolate some other way what is happening. I followed your earlier notes and I think my test does comply. However I don't really have a goal other than to understand the effect. If somehow there is something going on that ends up replicating a silhouette effect, that is fine by me. I don't think my results are inaccurate from a realism standpoint, I am just trying to understand why it is producing them. I certainly don't want it taken away. Maybe HAL has become self aware and added it in.

[BigJ62]

I'm thinking not silhouette effect but proximity to another wall gives a slight bonus.

Well I may try some other alterations to see if I can figure out if something in my tests is flawed or if I can isolate some other way what is happening. I followed your earlier notes and I think my test does comply. However I don't really have a goal other than to understand the effect. If somehow there is something going on that ends up replicating a silhouette effect, that is fine by me. I don't think my results are inaccurate from a realism standpoint, I am just trying to understand why it is producing them. I certainly don't want it taken away. Maybe HAL has become self aware and added it in.

[womble]

I'm thinking not silhouette effect but proximity to another wall gives a slight bonus.

I very much doubt that this is the case. Cover from features is supposed to be modelled on a pixel-by-pixel basis and a wall behind a target will never (barring ricochets) have a bullet path intersect it before it intersects a ptruppe's tender pixels. Looking at the screenies, the 'background' wall is at least a couple of Action Points away from the troops behind the wall anyway.

[Vanir Ausf B]

Finished the test ck3 suggested. I also just lost the post explaining it all because somehow pressing some button on my keyboard made the page back up to the main forum, erasing everything I had wrote. So here's the quick and dirty:

14 isolated firing lanes, at one end a German MG bunker with a squad of Germans in it, at the other a squad of US armored infantry with a Pl HQ. Test was run 60 times each for wall and no wall. Test was 1 minute long.

With wall

killed: 866

wounded: 951

total: 1817

average: 30.3 / 19.6%

Without wall

killed: 1114

wounded: 1141

total: 2255

average: 37.6 / 37.6%

These results are about the opposite of my previous tests. Why? I ran into a couple of problems with this test. The first is that the infantry in the open were about 5m closer to the bunkers than the ones behind the wall. It couldn't be helped. The wall goes down the middle of the action spot, so any infantry behind it are stuffed into the back half of the action spot. When the wall is removed the infantry spread out with most of them tending to go toward whichever way they are facing, which in this case was toward the bunkers. It's not a big difference, about 198m vs. 193m on average, but it probably made a small difference.

The bigger issue is that despite making the US squads fanatic they did cower. A lot. I did put some immobilized trucks next to the bunkers for them to shoot at but they ignored them and only shot at the bunkers. I went back and reran my first test to see if the fanatics ever cowered in that one. They did. So why the different results? Because the 2 tests test 2 very different situations, one where cowering is a Good Thing the other where cowering is a Bad Thing.

In this test the US squads were being fired on by something that was nearly invulnerable to their return fire. So the only smart thing to do was to get out of LOS/LOF. The guys behind the wall could do that, they guys in the open had nowhere to hide. In my first test the squads were in a firefight with an infantry force of equal strength. In that situation cowering allowed the other side to gain fire superiority.

So for me the uptake of this is that if you want your guys to hide, a wall is a good thing to be near. If you want your guys to shoot at something when there's a strong chance something will shoot back at them, you are better off in the open.

[BigJ62]

What is intended/revealed and what it actually does can be 2 different things. I’m more inclined to believe a bug or abstract modifier than a very cool but unknown feature. Then again it could just be some strange coincidence.

I very much doubt that this is the case. Cover from features is supposed to be modelled on a pixel-by-pixel basis and a wall behind a target will never (barring ricochets) have a bullet path intersect it before it intersects a ptruppe's tender pixels. Looking at the screenies, the 'background' wall is at least a couple of Action Points away from the troops behind the wall anyway.

[womble]

What is intended/revealed and what it actually does can be 2 different things. I’m more inclined to believe a bug or abstract modifier than a very cool but unknown feature. Then again it could just be some strange coincidence.

Why suspect an abstract modifier when they've done their damndest to remove such things? Or not put them in the first place, rather. Why would a high wall ever have an abstract modifier associated with it when it's a binary "Behind it: safe from small arms/In front of it: no protection" situation? Especially one that extends out for 2 or more APs? That makes no sense whatsoever. And the 'very cool' feature isn't "unknown"; Steve has said in this very thread that "...there are natural conditions which could produce some Skylining effects. ..." How cool is that?

[BigJ62]

After running some tests I’m even more convinced that when the firing unit is within 2 grids of a wall to its rear it gets significant bonuses or whatever you want to call it.

10 USA rifle squads versus 10 USA rifle squads. Each lane is 100m wide separated by a high wall and distance is 210m. Each squad is regular, fanatic and 0 leadership. Hotseat, fow veteran.

1st Test no walls both sides, 3 runs of 10 turns

Allied wia 54, kia 81, total 135

Axis wia 129, kia 164, total 293

2nd Test Allied no wall Axis back wall, 3 runs of 10 turns

Allied wia 168, kia 190, total 358

Axis wia 20, kia 15, total 35

3rd Test Allied back wall and Axis no wall, 3 runs of 10 turns

Allied wia 33, kia 28, total 61

Axis wia 121, kia 224, total 345

So for a total of 6 runs for 10 turns we get 703 casualties for the side that has no wall and 96 casualties for side that has a wall behind it with an empty grid between it and the wall.

[SlapHappy]

Wow, so what explains the difference between troops who are directly behind the wall, and those who are actually up to 24 feet away and on the wrong side of it? Why would this "modifier" effect be less pronounced at the actual action spot?

[dieseltaylor]

Interesting I think. Worrying.

[Childress]

It means that if you're ever sentenced to death by firing squad be sure to demand, as your last request, a CM™ wall.

[xian]

Perhaps BF used a programming shortcut, with an algorithm such as: if a unit is within a certain distance of a wall (no matter the direction of incoming fire) it receives a protection bonus.

Could this really be the case?

[Childress]

Or some code parameters were simply reversed.

[xian]

indeed. ouch - glad it's been discovered.

[limbo696]

BigJ62, your test is perfectly reflective of reality: What is clearly happening here is that the rifle grenades are skewing this result by ricocheting off of the opposing wall, back over the enemy troops, and landing back right back on top of the shooter! Your test is completely flawed--need to get rid of those rifle grenades!

<sarcasm on>

Big thumbs up to the people that are running some very basic test/experiments trying to find out how different types of cover may not be working correctly or giving very counter-intuitive results that many of us are experiencing in our own games.

Big thumbs down to some of the "arm chair generals" here who are criticizing the validity and design of the tests trying to examine a SINGLE variable without running any of their own.

[sburke]

In response to the suggestions that it could somehow just be a bonus modifier from an additional wall structure I ran some additional tests. The results from a single test (I have a number of things I wanted to test so only had time for a spot check) indicated that a low wall behind actually made things worse. My suspicion is it gave troops a place to break to that in actuality offered no more protection and in the process of withdrawing they exposed themselves even more. The units with a low wall behind them got cut up pretty bad.

With no wall ---------wnd/dead 36/25

with low wall behind-- wnd/dead 65/73

number of casualties w/ no additional wall-56

number of casualties w/ additional wall----82

I then tried it again with a tall hedgerow figuring if it is a silhouette effect the hedgerow might create it as well

With no wall ----------wnd/dead 28/33

with hedgerow behind--wnd/dead 70/71

number of casualties w/ no hedgerow-68

number of casualties w/ hedgerow----73

Next test I extended the firing range to move the large wall quite a bit back from the low wall

With no wall ---------wnd/dead 21/22

with low wall behind-- wnd/dead 62/75

number of casualties w/ no additional wall-68

number of casualties w/ additional wall----69

What I think I am finding with admittedly a small sample pool is that having that large wall within a couple hexes is creating some additional bonus almost as it it is considered part of a structure. Any other material simply gives the troops a thought they can retreat to cover and they just get chopped up. Once the wall is far enough away the effect is no longer created and they behave just like the troops with no wall.

I will run a couple more and see if I can tell when distance kicks in (if it does) for a change in behavior.

Meanwhile all the tests with no wall, an additional low wall or a hedgerow all do show an advantage to some degree of prone troops in the open. In all of my testing the troops with simply a low wall have fared poorly against prone troops in the open and that is with approx 200 iterations.