Major bug : infantry under fire in wheat or brush

[Hensworth]

I did a search but nothing came up.

Has anyone noticed that infantry under fire in wheat or brush tend to run towards the enemy ?

I noticed it once in a game against the AI. This was long range fire and the wheat was the only cover around. No matter what I did, I could'nt keep my troops to sit still and shoot back. Even if you order them to withdraw, they will still crawl or run towards the direction where the fire is coming from.

Now I've had it again in a PBEM, with disastrous results...

[Doug Beman]

Definitely a known issue. Keep in mind that wheat/brush isn't very effective cover; it offers some concealement, but won't stop incoming fire. Your troops' AI weighed "possible risk of death from staying put" against "possible risk of death of getting the hell out of this exposed area" and decided the risk from the first course of action was greater than the second

DjB

[Hensworth]

Well make up your mind Doug. You start by saying it's a known issue and then proceed to tell me I shouldn't move infantry into wheat or brush in the first place.

I'm not talking about units breaking for the rear. That's perfectly acceptable. I'm talking about units with intact morale who can't be made to keep still.

[The Commissar]

Sgt_Kelly,

Could it be that the nearest cover around was the one indeed occupied by your foe? If this is the case, the rather simple Tac AI seeks the closest cover and makes a mad dash for it. Unfortunately, it doesn't seem to realize that said cover will bring them more death had they staid.

It must be brought up that Ive seen this happen, but only under certain circumstances. If the enemy fire is great for example, and manages to suppres your men, then they will react as you described. However, in a recent PBEM, while my men were few and delivered only meager ammounts of fire, the enemy forces in the wheat manages to remain in the wheat and deliver return fire pretty effectively.

Moral? Don't get caught in the wheat field unless you're sure there aren't many enemies around.

[Hensworth]

Alright. You guys are making this very difficult.

I had a platoon of engineers move up to just below the edge of a small ridge, in some brush. They were not taking any fire as they were in defilade from the Krauts up ahead. Unfortunately they decided not to stop and take cover, like I had told them to, but continued OVER the crest of the ridge and into a galling fire. End of turn.

In the next orders phase I see their morale is nothing worse than cautious, so I tell them to hit the deck and crawl back down the reverse slope. Instead, they continue walking straight into the enemy fire until they break.

So you see I wasn't even expecting the brush to provide cover.

I've used this defilade trick often. It works a treat behind walls for instance. As soon as something shoots at them, they can put their heads down and they're safe as houses. It works even in the open, but not in brush or wheat apparently.

[WWB]

Sgt Kelly,

First, I reiterate what Doug and the Commisar have said. The TacAI will break for what it thinks is the best nearby cover, which sometimes is not your idea of the best way to go and might in fact lead them into a deadly hail of bullets. It is only a computer program, not a human being, and every situation cannot be modeled perfectly.

Moreover, it probably adds realism, as people under fire do not do the most logical thing.

Sorry you lost a platoon.

WWB

[The Commissar]

Sgt_Kelley,

1) Well, to be fair, you didn't tell us the whole story the first time around, so don't go around blaming us.

2) Although your plan might sound good to you, the AI might not realise it quite this way. The brush you chose as cover, protected by the ridge or not, must have seemed too barren a place for your troops to take cover. Or perhaps it did not have the characteristics of the cover you usually use with this tactic.

3) In the military, when you are ambushed, you are taught to immediately charge the enemy.

To quote Peter McAleese, retired SAS veteran:

"You have made the classic mistake: no scout out, approaching an obvious focal point. The enemy ambushes you. Unless you are bullet proof or they are terrible shots, you have to react instantly to stand any chance to survive this one. You need to counterattack immediately, deliver as much fire as you can and get in amongst them. Maximum firepower and maximum aggression are the only ways to survive in this situation."

-Excerpt from 'McAleese's Fighting Manual'

Once again, I realize you probably did not consider the situation you threw your troops in as an ambush, but they saw it as such because they were in relatively open ground when an enemy opened fire on them. So technically, the AI did everything right, albeit with tragic results.

[ 12-09-2001: Message edited by: The Commissar ]</p>

[Seanachai]

<blockquote>quote:</font><hr>Originally posted by Sgt_Kelly:

Alright. You guys are making this very difficult.

<hr></blockquote>

Actually, I think they're making you define with some degree of accuracy what actually occured, which is essential to deciding whether it's a 'bug' in action, a weakness of the Tac AI, or circumstances/player stupidity.

If you want to report a 'bug', and want it registered as such, then all other explanatory factors have to be accounted for and dismissed. If you want to merely report that the fact that you got your arse chewed up means there's a bug in the game, than by all means carry on. Don't be surprised if people continue to refine down the question of 'why'. For that matter, you should appreciate their efforts to help determine exactly what was at fault.

Your final, more detailed explanation at least made me consider that it might be a bug. Your earlier posts made me picture the times when my troops came under fire, and moved towards the best 'cover', as perceived by the Tac AI, and suffered horribly for it.

I just had a squad in a game against JD Morse, one of the evil lawyer types, run across a road, come under fire, and, rather than moving another 5 meters to a point (albeit in 'open terrain') where none of the opposition could have fired upon them, they turned, hit the ground, and headed back the long way through enemy fire towards perceived 'cover'. They died, of course.

[Seanachai]

The Commissar and I are in agreement. All hail Slavic perspicacity!

I imagine he'll still punch me in the nose, given half a chance...

[ 12-09-2001: Message edited by: Seanachai ]</p>

[Affentitten]

But hey, at least your men did something. In Close Combat your men would have hit the deck, crawled 5 meters towards the enemy, 5 meters to the left, 2 meters back, one meter forward and so on until dead, all the time ignoring the tree cover 9 meters away.

[History Buff]

Has anyone thought that maybe the few men that ran towards the enemy were fanatic?

:confused:

Just a thought.

[Captain Wacky]

I don't think so. The manual only states that fanatic troops will hold their position no matter what, will rarely cower or surrender. I don't know how fanatic troops perform in an offensive role. I would imagine they fulfill their orders no matter what kind of fire they're under.

[Hensworth]

I beg forgiveness for the rather amateuristic way in which I am going about this. Up until now I've mainly enjoyed the game tremendously without harping on about shortcomings which are inherent to a computer simulation.

I've been playing against human opponents for ages and am not surprised to lose the odd unit through unpredictable behaviour or my own stupidity.

I believe to have stumbled upon something though which cannot be explained by either.

Before I waste your energies further, I'll try and setup a test to show what I mean.

[JasonC]

It is a real issue, and I can explain why you are seeing it and being frustrated in the reasonable things you are trying to do.

The root issue is that the tac AI, which makes decisions during each turn, does not understand LOS line based forms of cover. It only understands the "% exposure" ratings created by the actually terrain each unit is in. This is caused by the inability to calculate every possible unit pairing on the fly during the turn resolution, which restricts the things the tac AI can "look at".

Instead of checking LOS from A to X Y and Z, and repeating that for every possible location it could move to (down to 2m location resolution), and trying to maximize some utility function over this complicated "phase space", the tac AI only "looks" at the ground under the men's feet, which means only one variable for each unit. Much simpler, thus much faster. The problem is it leaves out the following cover effects - walls, hedges, sufficient distances of trees to break LOS, elevation changes.

The tac AI doesn't know that it can achieve 100% cover by retiring 20m. Sometime the unit will run just to get away from people, reach a location that no one can see, and stop there simply because it is not taking fire. But it didn't stop because it got out of LOS. It stopped only because it hit scattered trees or whatever, and it stays because it is not taking fire.

This can sometimes lead to the "blind chicken's run to dead ground", whereby a broken unit tries this or that spot of cover, leaving each in turn if shot at, until finally halting in a spot where no one can shoot at it because it is out of LOS. But this was not a calculated move to a dead ground location. It was just a messy process of trial and error ending up there.

What happens as a result of this "look at your feet only" cover check, is open ground is avoided even when it is the safest spot around. "Open" in this sense includes open, brush, and wheatfield - even with foxholes present as well. All of which leave % exposed in the 40-80% range. The cut off line is scattered trees, which is around 30%. Rough also works (25%), as do buildings (light 20%, heavy 10%), woods and dense pines (15%, 10% with foxholes added).

Basically, the simply test is "will I take more than 1/3rd of the incoming FP if I stay here?" If the answer is yes, the ground is effectively open. There is no calculation of where the enemy FP can reach, and where it can't. Even a unit with heads down behind a wall, thus 0% exposed, will consider itself endangered and in "open" ground, if that is the terrain behind the wall.

What then happens to units shot at while in the open, especially moving in the open (including brush and wheat)? The answer is they go into a "stress" move routine. In which there are three states - not in danger (very low incoming fp only), in danger, and panic.

Panic and worse, the unit will beeline for the nearest available cover but will not close the range with the enemy. If there is practically no cover around, you will see the unit turn and run straight sideways, clear off the map if necessary, because it can't move towards the enemy once in panic. You will see similar behavior from "pinned" or (more rarely) "shaken" units, without ceasing to respond to orders - they break for cover.

"In danger" includes all the moral states from "alerted" to "cautious". The unit reacts to the fire by a "resolute" move toward cover, the enemy shooters, or its assigned waypoint. They accelerate to run to reduce their time of exposure. They are acting as though they are ambushed in the open - they charge to eliminate the fire or get out of the kill zone.

They will "stick" at "cautious" morale state, even under very heavy additional fire. The instant they reach cover, they will go to ground, heads down, and often their morale state will drop precipitously - to shaken, pinned, panicked. They are actually in a special intermediate "moral state", effectively in between "cautious" and "panicked".

Sometimes this is indeed the behavior you would want. If a platoon "moving" (walking) across an open field is suddenly ambushed by a machinegun at the far side of it, this would be SOP. Stop walking, run, get out of the field, do not go to ground, charge the enemy rather than remain under his fire. That is what the behavior is meant to simulate - the tendency of experienced troops to "push on to cover" instead of letting themselves get pinned in the open. When the men really have no cover, this is often just what you'd want them to do.

The difficulty you are seeing comes from the interaction of this special "open risk" behavior on the one hand, and the limitations the tac AI has in understanding LOS-line based forms of cover on the other hand. Because you wouldn't put the men where you did, or want them to stay there, unless it really was reasonable cover or at least the best available to fight effectively from. But the tac AI does not know this. It thinks it is being ambushed in an open field.

Within the present game system, there really is no remedy for this except to adapt to the AI's limitations and make less use of the LOS forms of cover. This is a pity, because CM actually does a pretty darn good job of modeling those forms of cover at LOS-check and shot-resolution time. (There are minor problems there - e.g. peeking over a crest never gives foxhole-style cover bonuses, when realistically it would, against direct fire anyway).

You basically can only count on your men standing still to fight in scattered trees, woods, pines, rough, or buildings. Even the excellent cover provided by walls will not be used correctly, unless there are scattered trees behind it too.

For CMBB, several aspects of all this are going to be changed. The "run under fire" behavior is going to be modified extensively. A new form of movement - "assault" - is going to be added, simulating use of available cover, leapfrogging, which rapid movement. "Run" will in turn become much more dangerous and running units will not fire as well as they do now (or at all, perhaps). Fire will more often drive men to ground instead of making them accelerate in determined charges. Special "human wave" rules will allow such charges in particular cases.

All the details are still sketchy - I am just retailing statements BTS has made on the issue. Suffice it to say they know the issue exists, and one of its CM byproducts - "bum's rush" tactics, especially with SMG infantry - will be changed significantly in CMBB. For the final model, we will have to wait and see exactly how they handle it. I doubt they will address the actual cover-checking mechanisms to include LOS based, unit-pair calculations. But the behaviors that result from "being shot in the 'open'" will change.

I hope this helps. In the meantime, stick to the trees and buildings.

[M Hofbauer]

I was about to post something with the same bottom message (CM does not account for LOS-type cover, just terrain-type cover) as JasonC'S post, but of course way less profound, detailed and articluated, but with his post mine is superfluous now.

bottom line: Sgt_Kelly, I think the answer to the problem you described lies in the CM "bug" (or rather: "routine") elaborated in JasonC's post.

[Hensworth]

Considering I haven't been able to reproduce the odd behaviour (probably too much incoming fire, but I'm not sufficiently obsessed to go fiddling around with this) and Jason's exhaustive coverage of the subject, I'm willing to let it go until such time as it costs me another platoon and I'll be back raving and ranting.

(shuffles off stage mumbling to himself)