Small arms casualties - tank crews

[Battlefront.com]

29 minutes ago, Heirloom_Tomato said:

In another thread on halftrack gunners dropping like flies, I ran a few tests and came to the conclusion that the game is having infantry not only spot for enemy targets but also do a threat analysis of the targets they spot. I hypothesised that a halftrack gunner or TC is a higher threat than an infantryman in the open due to the amount of "hurt" they can send in the direction of my men. Therefore, my infantry are more likely to all fire at once at a halftrack gunner or TC out of concern for their own safety, rather than to shoot at an easier to hit target. Is your quote above confirming what I suspected?

Yes.  Think about how unrealistic the game would feel if all infantry opened up full bore on anything it saw all the time *or* lazily plinked away as if they were at a carnival shooting gallery all the time.  That wouldn't seem right at all, would it?  Intuitively, if not intellectually, you expect there to be a difference in behavior based on various conditions, such as threat such as exposure such as range, etc. because that is how things work in the real battlefield.  The best way to simulate this is to have shooting decisions based on threat and payoff.  If something is low threat and low payoff (a single bailed out tank crewman) the shooters behave very differently than if it is a Rifle Squad in open ground at 50m.  As well it should be

As someone that has worked with the TacAI in both CMx1 and CMx2 for almost 20 years now, I can tell you for sure you guys would freak out if you saw how the game works with this sort of logic turned off.  I've seen it back when both CMBO and CMSF were under development as the preliminary TacAI was "fully shoot at anything you see as soon as you see it".  It was like having armies of Terminators with the self control of a 13 year old without the sense to conserve ammo!  Just think of that for a minute.  Entire platoons would be combat ineffective due to no ammo within minutes.  Units would give away their positions blowing through their ammo because one routed guy was spotted 500m away.   Yes, that bad   We didn't even let testers play at this stage IIRC.

Quote

If so, it would seem to me, that players tactics are causing higher death rates rather than faulty game mechanics.

Yes, the player has massive amount of control over results.  It's not just the obvious things like don't drive trucks straight into the enemy lines, but more subtle stuff like making sure that you stay out of your enemy's kill zones as best as possible and if you do have to close with the enemy then make sure you're going in with overwhelming force and lots of prep/suppression fire.  It's basic combined arms tactics 101 stuff, but we all (and I mean *ALL*) make basic mistakes on a regular basis.  We can afford to since it's only a game.  I think if you put most of us experienced gamers out into a real full spectrum battlefield that we'd behave very, very, very differently.  And this gets into some of the inherent realism problems with games, even carefully simulated environments like CM.  Players do stupid/risky things because the payoff potential is there, whereas in real life it rarely is.

Quote

With regard to the OP complaint about additional crew members unbuttoning and being killed off, a tweak to the system to reduce their willingness to do so would be great. I know I would personally be not very keen to stick my head out of a hatch that I just saw my crew mate get wounded or killed in. 

Yup!  It's all well and fine to say to a player "if you don't want this result, then don't do X" when the game is doing what is expected in real life.  By and large if a crew member gets nailed and falls dead or wounded into the vehicle the chances are nobody is going topside at that moment to replace him.  Rare exception being when it appears that there's no other choice.  Currently the game's logic appears to need some tweaking and for whatever reason this hasn't become a big issue until CMFB.  Again, my guess is that the larger number of StuGs in scenarios and the tight terrain is causing this issue to be more obvious.

Steve

[Ithikial_AU]

22 minutes ago, Battlefront.com said:

Currently the game's logic appears to need some tweaking and for whatever reason this hasn't become a big issue until CMFB.  Again, my guess is that the larger number of StuGs in scenarios and the tight terrain is causing this issue to be more obvious.

Could it be vehicle specific / case by case issue? I haven't had much a chance to get my teeth into CMFB yet because of RL, but I did notice playing one of the larger scenarios with Gustav Line (sorry name escapes me at present) that the StuGIII crewman appeared to have a deathwish, and loved popping up to man the external MG34 whenever there was a free chance. Even when I'm telling them to stay buttoned on account of nearby enemy infantry.

Perhaps it sticks out with the StuG more than other vehicles. Lose the first guy not nice but you can cope. Lose multiple and suddenly you don't have enough crew members to fire the main gun.

[Warts 'n' all]

I don't know anything about computer game modelling, but I have often wondered if the game treats Stug mgs as if they were tank mgs, And is therefore unable to take the crewman's vulnerability into account. It would be nice to know if I am right or wrong on this point. In the meantime I tend to have them stand off and make use of their main gun in safety, rather than have them get within infantry rifle range in the first place. 

[sttp]

2 hours ago, Heirloom_Tomato said:

I hypothesised that a halftrack gunner or TC is a higher threat than an infantryman in the open due to the amount of "hurt" they can send in the direction of my men. Therefore, my infantry are more likely to all fire at once at a halftrack gunner or TC out of concern for their own safety, rather than to shoot at an easier to hit target. Is your quote above confirming what I suspected? If so, it would seem to me, that players tactics are causing higher death rates rather than faulty game mechanics.

I agree with this somewhat, but on the other hand, the simplest tests one can devise do pretty quickly start to hint that something more than "faulty tactics" is at play here. We can pretty much remove "tactics" from the equation.

This evening, I took Rokko's scenario and removed the brick wall. The German soldier therefore had no cover whatsoever -- it was just that 1 guy with his Kar98, 100 yards away from 2 Americans with Garands. Flat, open terrain. 40 trials of that. Then, 40 trials of those same 2 Americans vs. a German halftrack, the same distance away. And the halftrack was facing them.

 

Average number of shots needed to kill the German 100 yards away who had no cover : 30.

Average number of shots needed to kill a much better protected halftrack gunner: 19.

Rate of fire was the same whether the Americans were targeting the halftrack gunner or the lone soldier. (Both were right around .6 rounds per second.) So the " TacAI willingness to shoot" that's been mentioned appears to be minimized as a factor

 

Vs. the halftracks, it took the two infantrymen only 5 or fewer shots in almost 25% of the trials, 10 or fewer shots in nearly half of the trials, and 20 or fewer shots in almost 75% of the trials. It was one shot-one kill 10% of the time. These kinds of numbers may reasonably reflect reality, Not sure. But they must be contrasted with the following....

Vs. the German on foot, in the open... well, these two infantrymen could take him out in under 5 shots only 17% of the time, and under 10 shots about 20% of the time. The median # shots it took to kill the HT gunner was 12. The median number of shots to kill the infantryman out in the open was 26.5. (Note that vs. the halftrack, the mean is more than 1.5 times the median. This reflects the fact that it was basically only a few extreme trials that pulled the average so "high" up to 19... i.e., 12ish is a much more typical result than 19ish.)

These kinds of numbers are surely not conclusive, and we shouldn't read as much into them as their precision implies. We really need to test more infantry types and vehicles. Absolutely. But, in the meantime, how can such results not be at least initially alarming? Some people call tests like these "apples and oranges" ... and some say it's merely "poor tactics" by players plus their confirmation bias making it seem like infantry are much more accurate vs vehicle gunners. More numbers are needed, yes, but... it's a start, and in my opinion even these have a lot to say.

(Oh, and by the way, yes, I do have a full spreadsheet and all trials have been saved.)

 

Edited April 12, 2016 by sttp

[Bulletpoint]

11 hours ago, Rokko said:

Also, to me this whole thing is not at all about whether the gun shield works as intended, it's about finding out if there is some technical issue that makes the accuracy calculations more, well accurate when applied to shooting at personell in vehicles compared to shooting at infantry in the open.

Maybe it would be better to take the gun shield out of the equation and use a US halftrack instead? The American crew sits with half the body exposed, like the infantry behind the low wall.

[Rokko]

7 hours ago, Battlefront.com said:

Targets come with different "incentives" for enemies to shoot at it. By "pecking order" I meant that infantry is more incentivized to fire as fast and as much as possible against something like a flamethrower than a truck driver.  How much of a difference it makes in this artificial setting... I have no idea, but it's something to take into consideration.  You can get high priority target such as a single flamethrower dude into the game by using the pre-battle casualties to bump off the assistant.

Ok, but this test is so simple there is really not much room for prioritisations. You've repeatedly written as if there was a whole squad of 11 guys with a variety of weapons firing in this test, but it's only two guys with Garands. And my and @sttp's observations so far seem to indicate that there isn't really a change of ROF of something alike.

And since we are talking about accuracy, wouldn't a higher ROF further decrease the accuracy, thus increasing the number of shots it takes to take out a high priority target. Unless there is something in the Tac-AI that makes guys shoot less accurately than they possibly could because the target is just a truck driver and they don't really want to kill him I don't see the relevance of this. How could two guys with M1s "shoot harder" at a guy because he has a flamethrower or an MG that they could at a truck driver?

7 hours ago, Battlefront.com said:

We've had discussions about accuracy many times before.  Sometimes it's better to be out in the open as it makes aiming for a specific spot harder.  Hitting a specific spot, on the other hand, is a separate issue.  And this is where the dude bouncing up and down has a big impact.  When the truck driver ducks down the riflemen remove their rifle from the aim point, which means each time the truck driver comes back up they have to target fresh.  At best they might be adjusted for range and shooting conditions compared to the unbroken engagement against the HT driver which allows for better accuracy of follow up shots.  Now, in a real world situation the difference might not be all that huge because soldiers targeting the truck driver might keep their weapons trained on that position.  CM, however, does simulate this because 99% of the time it's not what happens in combat.  Target goes down, soldiers start to look for other things to target because the world is a hostile place and target fixation, especially if you think you got the guy, isn't a healthy practice.

I agree with all of the above, there is just nothing I as a regular player can do about that while testing. I guess using a single sniper or a single Garand shooter could help to decrease the cowering due less volume of fire and suppression but not entirely rule it out. One could also abort the test once the truck driver cowers for the first time but I suppose this would make comparing the test results almost impossible.

7 hours ago, Battlefront.com said:

Cowering is a "state" and therefore changing the animation won't change the behavior.

But we know that the 3D model and its position in the 3D space has an impact on hit calculations. The question here is, whether the simple state change to cowering is giving some sort of extra survivability bonus distinct from the bonus that comes from the lowered stance and thus smaller target area. If you say this is so than this idea is moot, of course.

7 hours ago, Battlefront.com said:

In real life I would expect the chances of placing a shot onto the shield of the HT would be easier than getting a shot similarly placed relative to a guy kneeling behind a wall.  At least in terms of combat type shooting.  Whether it's more effective, however, is a second question.

 Huh, why so?

Besides, the issue does not seem to be with shots hitting the shield, but with shots hitting the tiny visible area of the top of the helmet or the tiny slit where the shield is open.

7 hours ago, Battlefront.com said:

In real life I would expect the chances of placing a shot onto the shield of the HT would be easier than getting a shot similarly placed relative to a guy kneeling behind a wall.  At least in terms of combat type shooting.  Whether it's more effective, however, is a second question.

Again, no squad at all present. Just a two-man team of M19A4 ammo bearers with M1 Garands and alot of ammo to fire. Snipers would be an option, though, but I wanted things to be as regular as it gets.

7 hours ago, Battlefront.com said:

That's what I've been trying to get you to realize... you are not comparing apples to apples.  You are comparing one guy who is predisposed to standing up and taking on enemy fire (because he assumes he can fire back) to another guy who is instructed to take cover.  This alters the chances of hitting and that skews the data.  You can not ignore this any more than you can ignore factors such as the angle a target tank is facing in one situation compared to another.

Put another way, because we know these factors DO have an impact on the results, you can't use comparative results without minimizing the differences.  It is just not possible to do, therefore look for another way that doesn't run into this problem.

Yes I do realize there is a difference, totally. I just took offense to you claiming the difference between a bag of meat and bag of meat that ducks out of sight was as siginificant as the difference in armor protection between a truck and a Tiger 

7 hours ago, Battlefront.com said:

It would be pretty close, probably yes.  But we can not turn off the TacAI and therefore you have to find another way to do a comparison.  Arguing with me on this point won't further this discussion, so move onto something that will.

To summarize my suggestions for a new test, I'd do the following:

Make 12 "lanes" with Tall Walls separating each.  For this test I'd go with something like 100m wide and each lane a minimum of 600m deep.  Which means you could do a map that is 1200x600 and get what you need.  Run a strip of different dirt every 100m from "left to right" to mark the distances.

Lanes 1-3 and 7-9 put a SPW 250 with Regular or Veteran Crew and neutral modifies.  For the first group put them at 200m and for the second group put them at 300m

Lanes 4-6 and 10-12 put depleted flamthrowers (i.e. 1 man) at the 200m and 300m marks respectively.  Place a short wall between them and the firing line.

On the firing line for each late place a depleted sharpshooter (i.e. 1 man) with Regular or Veteran experience and neutral modifiers. 

Give each Target a short Cover Arc, give each Shooter no orders at all.  Hit GO! and note the following information for each lane:

1.  Shots fired before Target becomes casualty or ducks down or becomes a casualty.  Note which is the case.

2.  If a Target pops back up, note it and repeat the same tabulation as if it is a brand new example.

3.  Continue until all Targets are eliminated.

4.  Tabulate how many shots were needed to affect a state change for the Target and how many rounds were necessary to make it happen.  Group these results by state change (Duck, WIA, KIA).  Treat

Note that in my methodology I am not tabulating the data to have a total shot count for Targets that bounce up and down.  Instead I am treating each as a separate incident.  With the data recorded that way I can examine it individually or combine it for Target totals.  Which is a reminder that HOW you record the data is as important as the data itself.

As for the Experience settings, I chose neutral "game average" so the results are more likely to be comparable to common game situations.  Making them Elite and Fanatical means you're testing for something that isn't seen on a regular basis and is therefore not as comparable.

I can't think of anything significantly wrong with this approach, so I think it's a pretty solid way to go.

Steve

Seems like a potentially sound set-up to me. Except that I don't think several firing lanes in one "scenario" are a good idea. If a guy in one lane gets hit it still affects the "global morale" and thus the events are not totally secluded from one another. It's faster to test but I think it introduces more variables into the whole thing.

Would you then only consider the "Hit" events and discard all "Cower" events? So only count the shots that occured between either the scenario start or a guy getting back up and the same guy getting hit? If thats the case I'd probably prefer to abort a test once a traget ducks and restart, so the effect of improved aiming at target that has been aimed and shoot at before doesn't come into play. 

Edit: Regarding the Elite/Fanatic setting I want to make clear that those were only the case for the TARGETS, all shooters were Reg/Normal/+0. My assumption was that Experience/Motivation/Leadership didn't matter when being shot at, since I didn't assume more experienced guys were able to somehow dodge bullet. If you're hit you're hit, doesn't matter if you are green or elite.

In some cases I've tried with lower settings for the halftrack the crew could rattle and loose its covered arc or start backing away, the elite/fanatic setting prevents that pretty reliably, at least I haven't seen it happen with those settings so far.

Edited April 12, 2016 by Rokko

[Rokko]

4 hours ago, sttp said:

I agree with this somewhat, but on the other hand, the simplest tests one can devise do pretty quickly start to hint that something more than "faulty tactics" is at play here. We can pretty much remove "tactics" from the equation.

This evening, I took Rokko's scenario and removed the brick wall. The German soldier therefore had no cover whatsoever -- it was just that 1 guy with his Kar98, 100 yards away from 2 Americans with Garands. Flat, open terrain. 40 trials of that. Then, 40 trials of those same 2 Americans vs. a German halftrack, the same distance away. And the halftrack was facing them.

 

Average number of shots needed to kill the German 100 yards away who had no cover : 30.

Average number of shots needed to kill a much better protected halftrack gunner: 19.

Rate of fire was the same whether the Americans were targeting the halftrack gunner or the lone soldier. (Both were right around .6 rounds per second.) So the " TacAI willingness to shoot" that's been mentioned appears to be minimized as a factor

 

Vs. the halftracks, it took the two infantrymen only 5 or fewer shots in almost 25% of the trials, 10 or fewer shots in nearly half of the trials, and 20 or fewer shots in almost 75% of the trials. It was one shot-one kill 10% of the time. These kinds of numbers may reasonably reflect reality, Not sure. But they must be contrasted with the following....

Vs. the German on foot, in the open... well, these two infantrymen could take him out in under 5 shots only 17% of the time, and under 10 shots about 20% of the time. The median # shots it took to kill the HT gunner was 12. The median number of shots to kill the infantryman out in the open was 26.5. (Note that vs. the halftrack, the mean is more than 1.5 times the median. This reflects the fact that it was basically only a few extreme trials that pulled the average so "high" up to 19... i.e., 12ish is a much more typical result than 19ish.)

These kinds of numbers are surely not conclusive, and we shouldn't read as much into them as their precision implies. We really need to test more infantry types and vehicles. Absolutely. But, in the meantime, how can such results not be at least initially alarming? Some people call tests like these "apples and oranges" ... and some say it's merely "poor tactics" by players plus their confirmation bias making it seem like infantry are much more accurate vs vehicle gunners. More numbers are needed, yes, but... it's a start, and in my opinion even these have a lot to say.

(Oh, and by the way, yes, I do have a full spreadsheet and all trials have been saved.)

 

Thank you for joining in on this, I see you have the same concerns as I do and your observations seem to match up nicely with my own.

I had thought about removing the stone wall, the effect is that the truck driver would never kneel and always be lying down. From the POV of the Garand guys this would mean the target is smaller and thus actually more comparable to the size of the visible area of a halftrack gunner. But it would also mean that shots would be fired while the truck diver is cowering due to the (light) suppression. While cowering doesn't reduce the actually visible target area, the mere "state change"  does give some "soft" protection bonuses which mean that rounds, that otherwise would be counted as hits are counted as misses. At least that's how I think it works.

If the stone wall is removed, I'd also replace all the short grass with pavement or something similar thats absolutely even and does nothing to conceal a lying down target.

[Rokko]

1 hour ago, Bulletpoint said:

Maybe it would be better to take the gun shield out of the equation and use a US halftrack instead? The American crew sits with half the body exposed, like the infantry behind the low wall.

I agree, it could improve the comparability because the visible target area from the shooter POV would be more similar for each test case, at least if we keep the kneeling target behind a wall as it is now.

I am thinking of redesigning the test so, that the target is lying down all the time but also visible all the time without any concealment by the short grass (as little as that may be). In that setup a Sdkfz 251 gunner would be more similar I think 

[Bulletpoint]

1 minute ago, Rokko said:

 

I am thinking of redesigning the test so, that the target is lying down all the time but also visible all the time without any concealment by the short grass (as little as that may be). In that setup a Sdkfz 251 gunner would be more similar I think 

 

Maybe use pavement instead of grass, as that would give a smaller microcover bonus for the target.

It seems to me the entire thing is probably just about the micro cover not being applied to vehicle crew.

[Rokko]

This is what I was trying to migitate by having the target kneeling. My assumption was that the microcover bonus from short grass (a few centimeters high, essentially football field grass height) for a kneeling target was zero. Either way using pavement should rule out the guess work.

[Bulletpoint]

24 minutes ago, Rokko said:

This is what I was trying to migitate by having the target kneeling. My assumption was that the microcover bonus from short grass (a few centimeters high, essentially football field grass height) for a kneeling target was zero. Either way using pavement should rule out the guess work.

I think grass has a tiny cover bonus to represent tiny terrain bumps that the engine cannot properly simulate.

But the grass is not really the main issue. It's the wall. I believe the game adds a strong microcover bonus to troops being behind cover such as low walls or windows in buildings. Some shots are physically deflected, especially at range, but others penetrate, and I often see lots of bullets pass through soldiers inside buildings without causing damage. I believe this bonus also applies for the part of the soldier that pokes up above the cover.

In any case, as you say, pavement without any wall will take any such modifiers out of the equation.

Edited April 12, 2016 by Bulletpoint

[c3k]

It sounds like a "retarget" occurs if the target ducks out of LOS. It seems that "retarget" means there are the usual range estimation errors, etc., induced. As opposed to firing at the same target and correcting each shot as needed until a more accurate solution is found. IF (big "if") that supposition is correct, then the guy ducking down behind the wall introduces a new "retarget" algorithm each time. That could be what Steve is trying to get at.

My .02.

Interesting discussion on many levels.

Ken

[A Canadian Cat]

Yes, interesting.  I feel like we might be on the edge of finding something really interesting. I look forward to more information.

[[MyIS] Buffpuff]

On 4/10/2016 at 6:44 PM, Battlefront.com said:

Quick note... Chris just ran a couple of small tests, 30 iterations each.  Conclusions?  "Feelings" are not a good tool to use for making game design decisions

He ran 30 tests at 200m and 300m with a single US Rifle Squad (1x SMG, 11x Garand) up against a single Panther with exposed TC under perfect/ideal conditions for the US.  Here's the basics:

300m Test = 2x WIA, 28x Buttoned.  About 50% of the US Rifle Squads did not engage at all, despite spotting the tank facing it.

200m Test = 1x WIA, 18x KIA, 11x Buttoned.  All US Rifle Squads opened fire.

Overall I don't see anything wrong with these results.  200m should be a slaughter fest under these ideal TC plinking circumstances.  If someone is taking massive TC casualties the first thing I'd do is see what the range is and if it's 200m or so then you should adjust your tactics to something that more closely resembles real life (i.e. buttoned!).  If you're over 300m and seeing your TCs dropping like flies, then set up a test scenario to replicate the conditions (as best you can) and run the test a whole bunch of times (30 minimum) to see if there's any trend.

To sum this up, here's what we're examining:

1.  Reduce the threat level before a TC will button

2.  Reduce the chances a TC will unbutton if the player has explicitly instructed it to be buttoned

3.  If the crew is under strength it will by default revert to buttoned status

We are not examining anything other than this unless someone puts forward a test, or tests, which indicate there is a problem which needs addressing.

Steve

 

 

 

Interesting thread (I've certainly had fun reading it). But I have to say Steve I find it hilarious that Chris ran a test on button/unbuttoned tank commanders. Have you watched his Twitch/YouTube streams? NO tank commander is safe when Chris is behind the wheel (hey I used a car analogy there!!! LOL!).

[Rokko]

Ok so I've made a revised version of the test based on some of the suggestions posted here.

Shooter party:

The shooting party is now an MG ammo bearer team with its headcount set to 50%, so it's only one guy with a Garand. The total ammo count at the beginning is now 1000 rounds instead of 1008 for some reason! This makes things a lot easier though and you don't have to pay attention for mid-air shots. The shooter now has a stone wall in front of him. This was done to make sure he always shoots from the same stance, since in some tests I've run the shooter started and fired from the prone position and in some cases he'd start and fire from the kneeling position. The fact that the headcount is at 50% somehow gives the shooter a -1 leadership modifier, nothing I could do about that.

Target party:

The targets remain largely the same. They're still fanatic but their experience is set to regular. No particular reason for the change, mostly to prevent any complains on that part. The truck driver has the stone wall in front of him removed, so he's lying on the ground and can not duck out of sight.

Terrain:

The short yellow grass between the shooter and the target is now pavement, so there should be no significant abstracted micro-cover bonuses anymore.

Everything else remains unchanged, the denotation standard is also unchanged, so here's the results of my first 10 test runs each:

vs HT:
001 - 02(K)
002 - 10(K)
003 - 03(K)
004 - 22(K)
005 - 11(K)
006 - 03(K)
007 - 15(K)
008 - 05(K)
009 - 12(K)
010 - 26(K)

vs Inf:
001 - 18(K)
002 - 05(K)
003 - 07(K)
004 - 98(K)
005 - 05(K)
006 - 97(K)
007 - 04(W)
008 - 17(K)
009 - 21(K)
010 - 16(K)

From what I can tell Steve was right and I greatly underestimated the effect of having to zero back in on a target that has ducked out of sight before. In none of the new test runs the shooter lost his spot on the target. So far I can't establish a clear trend, the visible target area of the HT gunner is still significantly smaller than the prone truck driver. Shooting at infantry also seems to be more susceptible to outliers (the two instances were it took nearly 100 shots).

https://www.dropbox.com/s/fq1hfh4xm7qoafu/Accuracy%20Test%20Revised.zip?dl=0

Edit: Almost all of the test runs for the first test (vs HT) were over in under one minute, mostly due to the HT (and its gunner) beeing spotted almost immediately.

The shooting vs Inf took longer on average, up to 6 minutes I think in the two outlier cases, but spotting also took much longer on average, between 30 seconds and one minute I believe. I didn't really note those informations because I don't think they matter in this test, just FYI.

Edited April 12, 2016 by Rokko

[Battlefront.com]

10 hours ago, Ithikial_AU said:

Could it be vehicle specific / case by case issue? I haven't had much a chance to get my teeth into CMFB yet because of RL, but I did notice playing one of the larger scenarios with Gustav Line (sorry name escapes me at present) that the StuGIII crewman appeared to have a deathwish, and loved popping up to man the external MG34 whenever there was a free chance. Even when I'm telling them to stay buttoned on account of nearby enemy infantry.

Perhaps it sticks out with the StuG more than other vehicles. Lose the first guy not nice but you can cope. Lose multiple and suddenly you don't have enough crew members to fire the main gun.

The TacAI behavior isn't vehicle specific per se, however there could be a specific problem with a specific vehicle model.  For example, StuG III G (Late) might have some sort of glitch in the coding of the gun shield which effectively neutralizes its defensive properties.  It's unlikely, but that sort of thing is technically possible.  I don't think it's a factor in this instance.

1 hour ago, c3k said:

It sounds like a "retarget" occurs if the target ducks out of LOS. It seems that "retarget" means there are the usual range estimation errors, etc., induced. As opposed to firing at the same target and correcting each shot as needed until a more accurate solution is found. IF (big "if") that supposition is correct, then the guy ducking down behind the wall introduces a new "retarget" algorithm each time. That could be what Steve is trying to get at.

Yes, that is exactly what I'm saying.  Which means that each time the Target goes down prone behind the wall the count for that particular test should end.  When the guy pops back up again a new count should start .

IIRC the shooter's TacAI does have the ability to "remember" targeting information if the target stays out of sight only for a couple of seconds, but more than that it rests.  Note that this is separate from units remember what they were last shooting at and giving that target preference if it appears again.  The time for that is longer, though again not all that long (still measured in seconds).

Steve

[Battlefront.com]

9 hours ago, sttp said:

I agree with this somewhat, but on the other hand, the simplest tests one can devise do pretty quickly start to hint that something more than "faulty tactics" is at play here. We can pretty much remove "tactics" from the equation.

This evening, I took Rokko's scenario and removed the brick wall. The German soldier therefore had no cover whatsoever -- it was just that 1 guy with his Kar98, 100 yards away from 2 Americans with Garands. Flat, open terrain. 40 trials of that. Then, 40 trials of those same 2 Americans vs. a German halftrack, the same distance away. And the halftrack was facing them.

 

Average number of shots needed to kill the German 100 yards away who had no cover : 30.

Average number of shots needed to kill a much better protected halftrack gunner: 19.

I'm not surprised by these results.  It would be MUCH more difficult to hit a prone soldier in the grass than one standing up behind a shield.  At 200m the prone soldier is almost invisible, the soldier in the HT is effectively standing behind a huge sign that says "shoot me here".

Quote

 

Rate of fire was the same whether the Americans were targeting the halftrack gunner or the lone soldier. (Both were right around .6 rounds per second.) So the " TacAI willingness to shoot" that's been mentioned appears to be minimized as a factor

Good to have confirmation that the RoF is not different.  My guess is that the RoF is regulated by turning soldiers on/off.  Meaning, with so few shooters the TacAI keeps them all "on", but if there were 12 it might turn half of them off and have the other half firing occasionally.  In this case that isn't an issue.

Quote

Some people call tests like these "apples and oranges"
 

(raises hand) I'm one of those people

This thread is an excellent example of why making credible tests is so freak'n hard.  There's so many different ways for the tests to be "flawed" and wind up being "invalid" because of something you didn't take into consideration.  And that's why I keep asking you guys to NOT do tests until we've discussed the variables first.  That is part of the scientific method, BTW.  You peer review the test parameters before engaging in the actual tests so you have consensus (or at least some form of agreement) that the test is going to be valid BEFORE the test is conducted.  It also saves a lot of time and energy because what's the point of investing time into a test only to have me (or someone else) kick it in the teeth AFTER the test is conducted?

Steve

 

[c3k]

Lots of cool "behind the curtain" info being discussed, e.g., the retarget and how the TacAI "remembers" the last target, amongst others.

Makes it seem that that this is a pretty detailed sim.  Oh, wait: it is.

[Rokko]

Range 100, maximum zoom in each case

[Battlefront.com]

5 hours ago, Rokko said:

And since we are talking about accuracy, wouldn't a higher ROF further decrease the accuracy, thus increasing the number of shots it takes to take out a high priority target. Unless there is something in the Tac-AI that makes guys shoot less accurately than they possibly could because the target is just a truck driver and they don't really want to kill him I don't see the relevance of this. How could two guys with M1s "shoot harder" at a guy because he has a flamethrower or an MG that they could at a truck driver?

Yes, accuracy decreases, but chances of hitting increases overall because more lead flying downrange is more lead to hit with.  Not proportional per shot, but more than not shooting at all.  As for the RoF issue, see previous comment about the TacAI controlling RoF by determining how many guys are shooting, now how much each guy shoots in a given period of time.  Now that I think about it I'm pretty sure that's correct. Which is why having 1 or 2 guys is definitely better than having a dozen for this sort of test.

5 hours ago, Rokko said:

I agree with all of the above, there is just nothing I as a regular player can do about that while testing. I guess using a single sniper or a single Garand shooter could help to decrease the cowering due less volume of fire and suppression but not entirely rule it out. One could also abort the test once the truck driver cowers for the first time but I suppose this would make comparing the test results almost impossible.

Since you can't control the TacAI behavior you have to adapt what you are testing to the circumstances you do not have control over.  This gets to "what are you trying to test for".  If you are trying to test for "how many shots does it take to get a state change of the target" then it works fine.  If you are trying to figure out how many shots it takes to get a KIA/WIA result then it doesn't work because of the reasons I stated.  And your subsequent test, more-or-less along the lines I suggested, confirms that.

5 hours ago, Rokko said:

But we know that the 3D model and its position in the 3D space has an impact on hit calculations. The question here is, whether the simple state change to cowering is giving some sort of extra survivability bonus distinct from the bonus that comes from the lowered stance and thus smaller target area. If you say this is so than this idea is moot, of course.

I suspect there are things going on that are separate from the animation, though you are correct that the physical position does matter.  It's just that there are other things going on besides the physical position.

5 hours ago, Rokko said:

 Huh, why so?

Have you shot at things at even 100m distance?  I can assure you even a moderately competent shooter can get a tighter hit grouping on a shield than a Human silhouette.  It has to do with how well your brain is sure of what it is shooting at.  This is why soldiers do not wear bright colored clothing and instead wear camouflage or dark single colors.  The easier it is to see a target the easier it is to hit a target. And a large piece of vertical metal is a lot easier to see than a dark colored Human figure at 200m.

5 hours ago, Rokko said:

Besides, the issue does not seem to be with shots hitting the shield, but with shots hitting the tiny visible area of the top of the helmet or the tiny slit where the shield is open.

That is a separate issue and must be tested for differently.  There could very well be a problem with the shield under performing, but I'd have to research it more.  For example, I don't know how well the armor deflects a 30.06 round at 200m.  For all I know there's a chance rounds penetrate.

5 hours ago, Rokko said:

Seems like a potentially sound set-up to me. Except that I don't think several firing lanes in one "scenario" are a good idea. If a guy in one lane gets hit it still affects the "global morale" and thus the events are not totally secluded from one another. It's faster to test but I think it introduces more variables into the whole thing.

There's no command structure or LOS between units and the test is happening very quickly, therefore your concern isn't likely to have any negative impact.

5 hours ago, Rokko said:

Would you then only consider the "Hit" events and discard all "Cower" events? So only count the shots that occured between either the scenario start or a guy getting back up and the same guy getting hit? If thats the case I'd probably prefer to abort a test once a traget ducks and restart, so the effect of improved aiming at target that has been aimed and shoot at before doesn't come into play. 

It seems you figured out what the right answer is to this.  Count all the bullets necessary to get the guy to "change state" and stop counting for that particular instance.  You can elect to count the "cower" or not later on.  But if you don't gather the data or separate it then you can't.  In fact, I've sometimes found some pretty interesting and unexpected trends from looking at data I wasn't specifically interested in.

5 hours ago, Rokko said:

Edit: Regarding the Elite/Fanatic setting I want to make clear that those were only the case for the TARGETS, all shooters were Reg/Normal/+0. My assumption was that Experience/Motivation/Leadership didn't matter when being shot at, since I didn't assume more experienced guys were able to somehow dodge bullet. If you're hit you're hit, doesn't matter if you are green or elite.

Yup, I understood that is what you did.  However, Experience/Motivation/Leadership does make a difference when being shot at because it determines how likely the soldier is to duck and how quickly it will expose itself to more danger.  Which may have an impact on perception when you're trying to compare it to actual in game results.  Which is generally why I use Regular because Regular is the most common situation in an actual game setting.

5 hours ago, Rokko said:

In some cases I've tried with lower settings for the halftrack the crew could rattle and loose its covered arc or start backing away, the elite/fanatic setting prevents that pretty reliably, at least I haven't seen it happen with those settings so far.

Yup, but unfortunately it also makes the crew more likely to stay upright.  Either way, keeping the Experience/Motivation/Leadership stuff the same for all Targets is generally a good idea.

Steve

[sttp]

A guy's 100 meters away from you in this flat and empty field of short grass.

He yells over and says "hey, I'm gonna start firing my 30 calibre rifle at you. For 60 seconds. So you'd better either get into that halftrack, behind the gun, or you can just stay out there in the open in whatever position you choose. Your choice. Either way, I'm gonna fire once every two seconds."

Which option do you take?

 

Edited April 12, 2016 by sttp

[Bulletpoint]

18 minutes ago, sttp said:

A guy's 100 meters away from you in a flat, empty field.

He yells over and says "hey, I'm gonna start firing my 30 calibre rifle at you. So you'd better either get into that halftrack, behind the gun, or you can just stay out there in the open in whatever position you choose. Your choice. But either way, I'm gonna fire once every two seconds."

Which option do you take?

 

"Do you feel lucky, punk? Do ya?"

Well no, I don't feel lucky.. I'll take my chances in the halftrack please, behind that big cosy gun shield 

Edited April 12, 2016 by Bulletpoint

[JoMac]

24 minutes ago, sttp said:

A guy's 100 meters away from you in a flat, empty field.

He yells over and says "hey, I'm gonna start firing my 30 calibre rifle at you. So you'd better either get into that halftrack, behind the gun, or you can just stay out there in the open in whatever position you choose. Your choice. But either way, I'm gonna fire once every two seconds."

Which option do you take, and why?

 

LOL @ Bullet,...Yeah, I would go into the HT. I just feel threatened in the Open if someone is going to shoot at me, and getting into the HT gives me a better perceived survival rate...You know, until another GI with a Zook 25 meters away, says, "You shouldn't have done that" 

Edited April 12, 2016 by JoMc67

[JoMac]

8 minutes ago, Bulletpoint said:

"Do you feel lucky, punk? Do ya?"

Well no, I don't feel lucky.. I'll take my chances in the halftrack please, behind that big cosy gun shield 

And It's..."Do You Feel Lucky, Punk...Well, Do You?"

You have a lot to learn,'Grasshopper, but you will soon Master the Art of the 'West Ponders'...

[Bulletpoint]

I've never been in a real halftrack, so I don't know for sure. But for you who tried it, if you're at the gun and looking down the sights, where would your eye level be? And where would that place the top of your head - above or below the edge of the gun shield?