"Run to mamma and get killed" bug still not fixed in v1.1 ?

[Spook]

Folks, I think I know exactly what is going on here, (as relating earlier experiences with "test" scenarios on infantry TacAI), but I first have to ask the following of tero:

1) Were the Allied troops in cover? Or rather, was the closest cover, as might have have been seen by the German FO, in the direction towards the enemy troops?

2) If the German FO was indeed running from the building towards covered terrain in the direction of the enemy, how far away was this cover from the building?

[Hakko Ichiu]

From Tero:

Even when the back of the house is out of LOS and there is no way for the AI to start thinking it is surrounded when it has been monitoring the approaches which are all within relatively unobstructed LOS ? This early in the game to boot.

The fire started on the wall facing my troops and the FO exited the house from that same side. Fire blocking all exits is feasible if they had to jump out of the second storey and the back wall was facing a slope. But why did they jump from the side facing my troops ? Why not through a window on the two walls that would have allowed some coverage from LOS once they were out.

In CM, graphical fire is an abstraction. From its graphical depiction, you have no way of knowing where in the building the fire started. It is unclear to me whether you recognize this basic aspect of the game.

I don't know, and perhaps only Charles does, whether the game engine models the position of the fire relative to the occupants of a building when the TacAI is determining which action to take. If it does model this, then the FO was taking precisely the correct action: get out of the burning building ASAP and take your chances. If the game engine doesn't model this, then you're still operating from a data set of 1, which leaves a heck of a lot of room to say "stuff happens".

Personally, I have not seen wholesale panic flight toward the enemy in any game I have played, and that is not for lack of routed troops. I play Axis and Allied troops about equally.

It is my considered opinion that you are inferring way, way too much from one data point.

------------------

Ethan

-----------

"We forbid any course that says we restrict free speech." -- Dr. Kathleen Dixon, Director of Women's Studies, Bowling Green State University

[Spook]

Originally posted by Hakko Ichiu:

Personally, I have not seen wholesale panic flight toward the enemy in any game I have played, and that is not for lack of routed troops. I play Axis and Allied troops about equally.

Actually, I can create the circumstance of "wholesale panic flight TOWARD the enemy" quite easily in a test scenario. But whether troops will panic towards the enemy versus just ducking down or routing away is strongly driven by relative distances in this same scenario.

This is why I've posed the earlier question as to what "distances" are in effect for tero's case point.

[Guest Germanboy]

Originally posted by tero:

[b>It ahs to assume someone has got a bead on it, and then it should bug out before the big stuff coming in confirms that suspicion.

Yes. But at what point ? Is the up to the player or the AI to decide ?

To the AI. If it occurs late in the turn you can override it, if early, I want the guy out of there. Waiting until the end of the turn, then factoring in command delay is just no good.

------------------

Andreas

Der Kessel

Home of „Die Sturmgruppe“; Scenario Design Group for Combat Mission.

[Guest Germanboy]

Originally posted by Horncastle:

...the only decent cover is a building...

Horncastle - no offense intended. This is the flaw in your reasoning. They are not decent cover. Houses are the places attacked first. If you have an isolated house/farmstead, and no cover nearby, avoid it and go somewhere else. In towns and cities the situation is different, because you have enough houses to split your troops. Peter is spot-on about that. When the Germans used single houses to defend their border in 1945, these houses were fortified. Even the houses in the villages south of Caen were fortified before they were seen fit as defensible positions. CM does not model fortified houses, and that is why a single house is a death-trap.

There has been a huge discussion about this in the thread that led to the code-change for automatic withdrawal. I still stand by what I said then.

------------------

Andreas

Der Kessel

Home of „Die Sturmgruppe“; Scenario Design Group for Combat Mission.

[CavScout]

Anything that stands out from the surrounding terrain is not a good place to seek cover. This could include a building or a clump of trees. Sommething that is obvious to the defender is going to be obvious to the attacker. A building can provide good cover and concealment provided it doesn't become the target of every gun the enemy has.

[jshandorf]

I played a TCP/IP game the other day with my friend and we both watched his FO, that was in his rear area, taking HEAVY HMG fire run forwards from cover then back to it. He did this for about 2 full turns until he figured out that running back towards friendly lines was the way to go.

I thought it was really odd behavior and would have no "real life" explaination for it.

Oh yeah.. And one more thing... Why is it that troops still don't see foxholes as good cover? I have had troops run into a scattered trees where foxholes are present only to have them hit the dirt about 5-10meters from them. Personally if I was being shot at and I saw foxholes about 15-30 feet away I would move like a raped ape to get into that cover.

Jeff

------------------

First of all, David, you stupid sot, if names were meant to be descriptive, everyone would have the, culturally appropriate, name of, "Ugly little purple person that cries and wets itself." -Meeks.

[This message has been edited by jshandorf (edited 01-29-2001).]

[Vanir]

Originally posted by tero:

When talking about the Allied units and their behaviour people SEEM to be more eager to look for solutions outside the "**** happens" cathegory.

I think this is off base. There is no difference between the behavior of Allied and German units. They use the exact same AI. Had that FO in the house been American he would have done the same thing.

------------------

You've never heard music until you've heard the bleating of a gut-shot cesspooler. -Mark IV

[Guest Big Time Software]

PeterNZ wrote:

I agree it would be nice for them to have a little 'think' before moving.

I can assure that it does this already. Unfortunately, AI can NEVER be as smart as a human. The FO either thought that was the best direction (i.e. TacAI made an outright error) OR was Panicked enough that it didn't select the best available choice presented to it (i.e. TacAI knew a better way, but the unit didn't "listen"). I can't tell you which happend in this case, but either are possibilities.

As annoying as it is to see the TacAI select the wrong path, there is very little we can do about this problem on the whole. There are too many special cases than we can code. There are even more restrictions on how much CPU horsepower we can invest in figuring out which is the best way. While even the dumbest human can instantly look at the map and say "this way bad, this way good" the same is not true for the computer. It might have to chug away for a full minute to consistantly come up with the same answer a human would produce after about 1 second.

So things will remain the way they are unless we find a particular pattern of behavior that can be nailed down to a VERY specific set of circumstances that comes up very regullarly.

Thanks,

Steve

[Sirocco]

Originally posted by tero:

An other FO is another matter.

Well, I couldn't possibly comment on that, of course, but as tero's opponent in the aforementioned game, I should answer some questions raised.

1) tero is attacking up a slope

2) the FO was on the upper floor of a light building

3) the building is at the top of the ridge, with my positions falling away behind it

4) the FO was shaken, not panicked

5) the nearest cover was a hedge in front of the building

The TacAI simply chose the nearest cover. I would be more disappointed if I didn't know just how hard it is to write good AI that processes information accurately and quickly. A balance does have to be made, as Steve pointed out.

Another issue was raised about rushing back and forth, and I've noticed that, too. I think the only solution is to give all units some kind of memory, whether the units are squads, AT guns or tanks. I'd like to think that's under consideration for CM2 as I think it would solve a number of different problems.

------------------

"He belongs to a race which has coloured the map red, and all he wants are the green fields of England..."

- Joe Illingworth, Yorkshire Post War Correspondent

[PeterNZer]

Thanks for the clarification Steve

PeterNZ

------------------

"Patriotism is the virtue of the viscious" - Oscar Wilde

"Don't F*CK with Johnny Cash!" - Chupacabra

[Guest tero]

>1) Were the Allied troops in cover? Or rather, was the closest cover, as might have have been seen by the German FO, in the direction towards the enemy troops?

There is a T shaped hedge in front of the house in front of the forest my troops are emerging from.

>2) If the German FO was indeed running from the building towards covered terrain in the direction of the enemy, how far away was this cover from the building?

The bit of the T shaped hedge that gives cover is some 30-50 meters downhill from the house and the FO was heading for that cover, got stopped by some fire and turned to run laterally in front of the house ie. the "run about undecided" routine.

NOTE: the FO was able subsequently to save itself. by running behind the hill, no doubt with express orders to do so.

I see where you are driving at with these questions: the unit AI looks for the cover with the biggest cover value in the front arc of its LOS. Breaking LOS to the firing units by going backwards is not an option, perhaps because it does not "remember" it can break LOS by going backwards. I have seen units go over the ridge to get into cover eventhough the right thing to do would have been to move backwards a few meters to break LOS.

[Guest tero]

>In CM, graphical fire is an abstraction. From its graphical depiction, you have no way of knowing where in the building the fire started. It is unclear to me whether you recognize this basic aspect of the game.

"Abstraction" is the term I love to hate when talking about CM game mechanics.

As to no way of knowing where the fire started: ALL my assets, including the 2" mortar which most propably started the fire, aimed at the wall facing my troops. Given the physics of the situation: house on a hill and the mortar firing from a position considerably lower to the ground floor of the house I would assume the round that started the fire hit 1) the wall 2) on the floor quite near to a window on that wall.

>I don't know, and perhaps only Charles does, whether the game engine models the position of the fire relative to the occupants of a building when the TacAI is determining which action to take.

IF the houses are modelled like movie sets and fires are modelled as having a single generic starting point then I it must follow that the FO had only one possible exit out of the house.

Incidentaly: the graphics of the house showed the door to be on the safe side facing the center of the village and not my troops.

>If it does model this, then the FO was taking precisely the correct action: get out of the burning building ASAP and take your chances. If the game engine doesn't model this, then you're still operating from a data set of 1, which leaves a heck of a lot of room to say "stuff happens".

True.

>It is my considered opinion that you are inferring way, way too much from one data point.

But I am basing the premise on the fact that v1.1 readme said the tendency to take cover by running towards cover near enemy troops was toned down. I just to know how much and by which criteria.

[Guest tero]

>To the AI. If it occurs late in the turn you can override it, if early, I want the guy out of there. Waiting until the end of the turn, then factoring in command delay is just no good.

Agreed. But what if you order it to run like hell to a location that is good cover but the AI overrules the player command and gets itself killed ?

[PeterNZer]

quick point:

The graphical depictions of the houses make no difference to the engine at this stage. So door or no, they walk in and out of them with ease. The doors and stuff are just a nice image laid over the top

PeterNZ

[Guest tero]

>There is no difference between the behavior of Allied and German units.

Not in my experience.

>They use the exact same AI.

And that is often the problem. It seems the AI assumes for example that all tanks have a stabilizer. That is why the Allied tanks have an advantage when the Allied Shermans can start engageing a target without having to disprupt its movement orders while the PzKw-IV is forced to complete its movement order before it can start engageing the target. If there was a separate AI for the German AFV's they would overrule the movement command and take appropriate action like make a sudden turn at full current speed to turn the gun towards the enemy. Before v1.1 they had to complete the movement order and only then they started turning the (ultra) slow turret (like a good Allied tanker with a ultra fast turret would have done) instead of turning the entire vehicle. I have still to see how effective the Stug has become in AT work now that in v1.1 tanks supposedly use hull turning more.

>Had that FO in the house been American he would have done the same thing.

Propably.

[Guest tero]

>quick point:

The graphical depictions of the houses make no difference to the engine at this stage. So door or no, they walk in and out of them with ease. The doors and stuff are just a nice image laid over the top

I am aware of this. Somebody just brough the issue up so I checked the placement of the door.

[Horncastle]

I agree it would be an improvement if the AI had some rule that acted as a memory.

If an action( get into nearest cover due to enemy fire)failed, due to a building being full or running across 50m of open ground covered by HMGs causing them to run back to their original position; the AI would not then try to get back into that same cover but seek something elsewhere.

It would most probably add many AI rules and be pretty complex to code accurately but its something to put in CM2 maybe.

[Guest tero]

Another thing that occured to me:

It would be nice if instead of the chain gang style panic now modelled the panicked unit would automatically split into the headless chickens section which runs into its death and the more level headed section which seeks cover more intelligently.

Or the entire panicked squad just disintegrates and disappears from the map for say 5 - 10 minutes and then reappears (IF it can) near a command unit regrouped and composed.

[Valamir]

I was wondering if perhaps the easiest solution wouldn't be to add a Fall Back position marker similar in form to the Ambush Marker.

Routed units could then be "encouraged" to head in the direction of this marker using a similar kind of logic structure as what is used to prioritise fire targets. Meaning even when you manually target an enemy the AI uses some logic to evaluate when switching targets is a better idea. A similar routine could be used to determine when other cover is a better idea.

An added level of realism would be that if a unit takes fire in a Fall Back position before recovering its morale, that morale effects are extra detrimental (even the safe place isn't safe).

[Guest Germanboy]

Originally posted by Valamir:

I was wondering if perhaps the easiest solution wouldn't be to add a Fall Back position marker similar in form to the Ambush Marker.

Routed units could then be "encouraged" to head in the direction of this marker using a similar kind of logic structure as what is used to prioritise fire targets. Meaning even when you manually target an enemy the AI uses some logic to evaluate when switching targets is a better idea. A similar routine could be used to determine when other cover is a better idea.

An added level of realism would be that if a unit takes fire in a Fall Back position before recovering its morale, that morale effects are extra detrimental (even the safe place isn't safe).

Err, correct me if I am wrong, but isn't that already in the game? Broken/routed units flee towards the next available HQ OR the friendly map edge. If e.g. a broken squad receives fire while recovering, chances are it will be routed (or the other way round, can never remember which is which).

------------------

Andreas

Der Kessel

Home of „Die Sturmgruppe“; Scenario Design Group for Combat Mission.

[Guest Big Time Software]

Tero wrote:

And that is often the problem. It seems the AI assumes for example that all tanks have a stabilizer.

Not true. The TacAI knows full well which vehicles are good for firing on the move. But it can only act within the parameters you set up. A Move order is for moving. Hunt is for fire and move (usually means stopping for Allied vehilces, always stopping for German ones). Fast means don't stop for anything. Since certain Allied vehicles can fire on the move more effectively than German vehicles, they tend to fire while moving more often. That is realistic.

That is why the Allied tanks have an advantage when the Allied Shermans can start engageing a target without having to disprupt its movement orders while the PzKw-IV is forced to complete its movement order before it can start engageing the target.

Use the Hunt command instead. Hunt is designed for just this situation, while Move is telling the TC that getting from A to B is more important than firing. Think about it Tero. If we had the TacAI stop vehicles whenever they thought a shot was possible, tell me how common would it be to see a vehicle move from A to B without being stopped a half dozen times in a combat situation? The TacAI can't read your mind, so that is why there are three different movement orders with very clear logic behind each. People would be yelling bloody murder on an hourly basis if we followed your suggestion about allowing the TacAI to stop a Move order JUST like a Hunt order Again, we put in the different orders for a reason.

Horncastle wrote:

It would most probably add many AI rules and be pretty complex to code accurately but its something to put in CM2 maybe.

Not only too many rules, but too many CPU cycles to resolve them. It just isn't possible to make the AI as smart as a human. Period. As it is right now the AI GENERALLY does the right thing MOST of the time. That is far better than other AIs out there, and unfortunately it is about the best we can do right now.

Steve

[Valamir]

Originally posted by Germanboy:

Err, correct me if I am wrong, but isn't that already in the game? Broken/routed units flee towards the next available HQ OR the friendly map edge. If e.g. a broken squad receives fire while recovering, chances are it will be routed (or the other way round, can never remember which is which).

Well, I've also seen very odd behavior from routed units. One that stands out in particular reminded me of a B-Western where the banditos open up at some cowboys feet crying "dance gringo dance"

I had two squads and a machine gun hidden near the edge of a field. A platoon of krauts rushed out of some woods on the other side attempting to cross the field. When my units opened up, most of the enemy squads hit the dirt behind a wall on their side of the field. A squad that had been in the lead was caught in the middle of the field.

Rather than head back to the saftey of the wall where his CO was, or to the shelter of a nearby farm house he remained in the middle of the field. He'd first run one way for about 10-20m. Then, hit by several bursts of fire, he'd reverse direction and run another way. Sometimes he'd drop the ground briefly and get back up and run yet another seemingly random way. This squad did the "funky chicken" for a good 40 seconds in the middle of the open field while his man power dwindled to 2 even though his CO and a nice safe wall were a mere 20-30m behind him.

I'm willing to put it down to "some people just act stupid under fire" and not worry too much about it (I'd probably be one of those people). But since it seems to bother several persons, I thought I'd just throw out an idea I had for improving the AI decision making in these situations.

[Guest tero]

>Not true. The TacAI knows full well which vehicles are good for firing on the move.

Exactly. That is not quite what I am contesting.

>But it can only act within the parameters you set up. A Move order is for moving. Hunt is for fire and move (usually means stopping for Allied vehilces, always stopping for German ones).

In my experience the Germans UNDER HUMAN COMMAND move all the way to the last way point first, and only then start to engage the target. If you have set the Hunt order too far the German vehicle will NOT stop and engage the target (at least in pre-v1.1 games I have played) before it has reached the last waypoint set.

>Fast means don't stop for anything. Since certain Allied vehicles can fire on the move more effectively than German vehicles, they tend to fire while moving more often. That is realistic.

Please quote your source where you picked up the fact that the Allied tankers fired (and hit) targets on the move more than the Germans did. All the sources I have ever read quite cathegorically state that ALL tankers, regardless of nationality, during WWII stopped the vehicle to fire its main gun unless it was a REALLY tight spot.

>Use the Hunt command instead.

But I do. I just have found that if you plot the Hunt moves too far the German tanks "overshoot" and lose the tactical initiative in a fashion that is less than historically accurate. And that has nothing to do with the vehicles being buttoned up. The spotting is done OK but the vehicle moves to the waypoint before it starts the target engagement cycle (ie, line the turret up, load up with appropriate ammo etc). It is as if the TacAI is disabled because it is blocked because there is no stabilizer in the vehicle and the vehicle has to physically stop before the TacAI can kick in.

>If we had the TacAI stop vehicles whenever they thought a shot was possible, tell me how common would it be to see a vehicle move from A to B without being stopped a half dozen times in a combat situation?

It depends on the situation of course. Which is more important: the movement or the firing in any given tactical situation ?

Since there is no priority target selection possible I can see your point. But since there SHOULD be a distinction between a rookie and an elite crew it should extend to their tactical eye as well as physical performance.

>The TacAI can't read your mind, so that is why there are three different movement orders with very clear logic behind each. People would be yelling bloody murder on an hourly basis if we followed your suggestion about allowing the TacAI to stop a Move order JUST like a Hunt order

Could the hunt order be made more like a suggestion than a real movement command ? Or could there be an additional command like "Free Hunt" where the tank is in effect under TacAI command until told otherwise by changing the orders by assigning an explicit target or making the command a regular Hunt or Move order. The TacAI can execute its movements towards the waypoints set by the player in the manner it sees fit and it would be more free to engage targets of opportunity at will.

[Guest Big Time Software]

Tero wrote:

Please quote your source where you picked up the fact that the Allied tankers fired (and hit) targets on the move more than the Germans did.

This was the whole point of the gyrostablizers. For more information, including some veteran reports from M18 tankers, use the Search to look for "gyros" or something like that.

All the sources I have ever read quite cathegorically state that ALL tankers, regardless of nationality, during WWII stopped the vehicle to fire its main gun unless it was a REALLY tight spot.

Not true. Soviets fired on the move for disrruptive effect, even though they had little chance of hitting anything. Allied tankers did fire on the move, but the general method for EFFECTIVE firing was to stop and fire stationary. The Germans had a SOP of only firing while stationary.

Note that we did tone down accuracy for Allied vehicles firing on the move. I think that was in 1.1, but perhas it is in the yet to be released 1.12.

But I do. I just have found that if you plot the Hunt moves too far the German tanks "overshoot" and lose the tactical initiative in a fashion that is less than historically accurate.

This is a factor of spotting, nothing more. The same exact thing happens for the Allies. But since this is a largely random/circumstantial type of a thing, you can form different opinions based on selective sampling. But I can tell you for sure that there is absolutely NOTHING that hinders the Germans in this regard any more than the Allies. So in short, I am saying your observations are flawed and that your point is not valid.

It depends on the situation of course. Which is more important: the movement or the firing in any given tactical situation ?

That is why we have different orders for Move, Hunt, and Fast. You are instructing the TacAI which is more important -> firing or moving.

Since there is no priority target selection possible I can see your point. But since there SHOULD be a distinction between a rookie and an elite crew it should extend to their tactical eye as well as physical performance.

This assumes that it is possible to define the conditions that would logically (realistically) have a vehicle halt its movement against orders and perhaps good sense. I say that it is not possible to come up with an acceptable pattern of behavior, so this is out. If you want your vehicle to move around in a combat zone and stop to fire, use Hunt. It is that simple.

Could the hunt order be made more like a suggestion than a real movement command ?

With the existing game engine, no. What you are asking for is some sort of SOP behavior, determined by the player and exectuted by the TacAI. That is far too complicated to implement at the moment (i.e. until we rewrite the game engine).

Or could there be an additional command like "Free Hunt" where the tank is in effect under TacAI command until told otherwise by changing the orders by assigning an explicit target or making the command a regular Hunt or Move order. The TacAI can execute its movements towards the waypoints set by the player in the manner it sees fit and it would be more free to engage targets of opportunity at will.

I am not sure I understand what you are asking for. Using Pause, Cancel, and Hunt I see the same behavior. So either I am missing your point or you are looking for control that is already in there.

Steve