An idea for discussion

[pamak1970]

ORIGINAL: Arjuna

It's a strict 300 x 300. So at the most your looking at would be around 432m on the diagonal. Few companies are bigger than 175 personnel. So in general you can have two companies in the same loc without any movement impediment.

I read this in the traffic jams thread and i thought of something else as an idea for possible inprovements of the engine for the feature.

Since a programmer , has the ability to define an area (in this case 300x300) and have the program respond in a certain way adding delays according to the number of units inside it,
could it be possible to give players a similar tool to have the ability to define areas of operations and boundaries between major formations like divisions or regiments?

What i mean is the following.
Until now the player can issue various orders to move his units, without paying any attention regarding violations of boundaries
since this concept is totally absent.
So, you can have in any area many units from different major formations mixed, without any consequence.

What i am thinking is the following.
Imagine that whenever a player has to issue movement orders, he has to do one additional step besides everything he does now.

That is he has to define a certain area of responsibility or operations, for each one of his major formations (regiment and above) he is moving (all or part of it).
All sub units of the major formation, should be inside this area after the completion of the movement.
This area will remain in effect, even after the completion of the movement and until a new one is executed ,in which case the player can again designate a new area of operations as part of the new movement order.

In the same way now the program adds delays if there are many units inside your predetermined zone of 300x300,
it might be possible to have the program add delays based NOT on the quantity of units but on the type of the hierarchy inside the zone that the player determines.

So the player does two additional things.
First he determines a certain zone of operation and second he "matches" this zone to a certain Hierarchy of a major formation.
During the execution of the movements, the program will check if all units inside this zone are part of this predetermined hierarchy.
If it finds that there are certain elements-sub units that are not part of this hierarchy,then it applies a penalty which can be simply a delay.
This delay will not simulate so much traffic jams, but rather the loss of tempo,hesitation,needed time to develop coordination or positive indentification due to the presence of new units which are not part of the original major formation.

So under these conditions, it will be more difficult for the player to order regiments pass easily through the area of other major formations, without any consequencies.

I have additional thoughts regarding how to give the means to a player to execute "pass through" operations,without much penalty as long as he has anticipated and preplanned this type of movement.
The same regarding the execution of counterattacks by sub units of major formations inside the defence area of other major formations.

Still i do not wish to go farther in details before i hear any reactions regarding the basis of my thought and if it workable to program such a behavor.

[Arjuna]

pamak1970,

Thanks for your thoughts on Operational Areas. These are on our wish list and have been since 1996. There are a number of reasons why we have not introduced these to date. Programming the penalties and effects of a unit moving into another force's OpArea is very straight forward. However, once you have OpAreas you also need to have "Passage of Lines" and "Relief in Place" tasks to facilitate planned or coordinated movements through or into another forces OpArea. These would not be trivial to code. We would need to provide the human player with a polygon tool to create the OpArea. A straight rectangle or circle would not suffice as you would want to be able to follow the line of a river for instance. While this is not an onerous undertaking, the real crux is how do you get the AI controlled forces to determine their OpAreas. This requires a little magic and a lot of consideration.

[Rooster]

This is an interesting concept - one I gave some thought to when I created my St. Lo scenario. In reading the War Department's account , it showed how screwed up the U.S. offensive became when they rushed armor into the bridgehead:

The best way, even under favorable conditions, to completely immobilize troops in a small area is to put an armored outfit there too. People think of the infantry as a blue or red line on a map. Actually the infantry has all sorts of activities going on behind that line: supply, wire lines, mortar positions, vehicles, etc. Did you ever try to keep field telephone lines in operation with tanks all over the place?-Well, I don't recommend it. The resulting confusion made it extremely difficult for either the infantry or the armor to get any real effort started, and time which should have been spent by the commanders in working out their own problems frequently had to be spent in arguing with each other about who would do what, where and when, or why not, etc

I can only imagine how difficult it would be to implement in the game.

[pamak1970]

ORIGINAL: Arjuna

pamak1970,

Thanks for your thoughts on Operational Areas. These are on our wish list and have been since 1996. There are a number of reasons why we have not introduced these to date. Programming the penalties and effects of a unit moving into another force's OpArea is very straight forward. However, once you have OpAreas you also need to have "Passage of Lines" and "Relief in Place" tasks to facilitate planned or coordinated movements through or into another forces OpArea. These would not be trivial to code. We would need to provide the human player with a polygon tool to create the OpArea. A straight rectangle or circle would not suffice as you would want to be able to follow the line of a river for instance. While this is not an onerous undertaking, the real crux is how do you get the AI controlled forces to determine their OpAreas. This requires a little magic and a lot of consideration.

Thank you for the reply Arjuna.
Although i do not know many things regarding programming ,i understand your thoughts regarding the nessesity to have a polygon tool and certain options to allow a player to preplan and execute passage of lines , releif in place or even counterattacks
inside another major formation's defensive sector.

I am a little bit confused only regarding the last comment about how to get AI forces determine their area.
I do not know if you think that this is not possible or if it requires a totally new game engine.

I tend to beleive that you mean the second, not because i know programming but cause i have seen one case where AI seems to understand regions determined by the player.
I do not know if you are familiar with the mission editor of Steel Beasts game.
Before i continue i have to make clear that i do not intent to do any type of comparison between two totally different games that i equally respect and enjoy.
I just try to understand if it is workable to have AI "understand" the issues we talk about here and i try to fill the "gap" i have in programming knowledge by observing AI behavor of other games.

Anyway and returning to my comment regarding the mission editor of Steel Beasts , let me mention some things.
First the editor can be used for both making scenarios as a designer, and issuing pregame-orders as a player.
In this type of editor ,the player or the scenario designer, has the ability to draw regions of any shape and name them.
The player has the option to determine conditions under which certain orders will be triggered.
For example one type of such an order is in the form of
"embark on route from waypoint A to waypoint B IF a certain condition is met during the scenario.

So by having a player for example define a region named "trap", the condition of the above sentence can be " .......IF enemy units are observed inside region "trap".

or it can be "..........If friendly forces are located inside region "trap".

or it can be even more specific like
"..........If at least three enemy units are inside region "trap"

Or ".......If the first platoon of the third friendly company is inside region "trap".

So ,at the moment these conditions are met , AI triggers a certain response according to the designer's or player's intentions
In this example the reponse is "embark on route between waypoint A and waypoint B".

The above details give me the impression that AI does understand the concept of a region and if certain units are inside or outside of it.

What i do not know is if it is more difficult to have this AI ability during play, compared to have it only during the use of a mission editor .

[Mark Weston]

I am a little bit confused only regarding the last comment about how to get AI forces determine their area.
I do not know if you think that this is not possible or if it requires a totally new game engine.

I think the problem is writing an AI capable of generating Operational Areas for its own subordinate HQs as a battle progresses. There's no way that pre-designated OpAreas are going to be relevant 8 days into a 10 day campaign, especially in a battle as non-linear as Arnhem or Crete. I actually can't even begin to imagine how you would do this!

[MadScot]

Why would you NEED operational areas for the AI? It's something of a 'cheat', but is there any reason why a 'reality constraint' put in place to make the game seem more realistic to the player needs to be implemented on the AI in anything like the same way. If the AI performs well as an opponent, who cares if it has it's own subtly different rulesset.

For example, you could have all the operational area stuff apply to the player, but have no rule that stopped him mixing his units up hopelessly - he'd just have a horrendous oparea management problem.

While you could code the Ai to be VERY reluctant to disperse a given brigade - while not actually having an oparea restriction at all.

In essence, you can control what the AI does directly - you don't need a carrot-and-stick. While to 'make' the player act historically you need realistic constraint on his play, but you leave him free to 'break the rules' if he's willing to take the penalty.

It would be analogous to the following:

Let's say you find you can't get the Ai to do a good job of managing fire support; it keeps wasting half it's ammo on useless targets. You can either write some horrendously complex routine to try to mimic the player's decision making process, to make the Ai smarter at using it's arty. that's hard.

Or you can just artificially give the Ai twice the 'normal' ammo loads and supplies. Since half the rounds it fires are duds anyway, the effect on the game is the same.

Similarly, if the Ai was really bad at guessing player intentions, you could relax fog-of-war for the AI only, to give the computer player a bit of help. It's not 'fair' but I don't want 'fair' - it's not like I prove my intellectual superiority over silicon by beating a PC; what I want 9and I think we all want) is a good game that feels right. It's like printing a photo in the newspaper; look closely and it's all dots. But if it looks right, who cares.

my 2c

[Mark Weston]

No, I strongly disagree. Because when you remove game-lmitations designed to create a good simulation, you distort the "reality" of the game, and in practice that distortion affects the player whether you intend it to or not. Lets look at your examples.

Relaxing fog-of-war for the AI means that the player loses half the value of winning the recon battle; defeat the enemy recon and he still knows where you are. A carefully screened night-time redeployment is completely wasted; the AI knows you're coming. Securing the start-line for your main attack is pointless; you're going to get hit with artillery anyway.

Double the AIs artillery ammo loads, and on average half of it may miss, but when a scenario focuses down to one key battle at one key objective where the situation is so obvious even the AI can't make the wrong targetting decision, the player will still find himself being hit by twice as much artillery as was historically available.

What happens is that the player is no longer rewarded for making historically and militarily appropriate decisions, because the AIs responses - while they will make the game harder to beat - will be completely unrealistic. The only rational response from the player will be to learn gamey tactics that bear no relation to the military history, but exploit whatever weaknesses in the AI behaviour that the player finds.

[MadScot]

I agree that ideally the AI engine should be capable of a challenging and realistic game without crutches from handicaps.

However, I'd hate to forgo part of the command experience as a player simply because they couldn't come up with an AI which could make intuitive decisions as well as a human.

Take the fog of war case: a skilled human player may be able to deduce intentions from contact information that no practical AI can. The only options remaining are then:

a) play with no 'fog of war' for either side : unrealistic player experience
b) play with full 'fog of war' for both sides : AI is unable to provide a realistic challenge because it can't take the disparate threads of contact info and deduce an enemy plan (it's as if they have no combat int function; plenty of contact reports and DATA but no INTELLIGENCE)
c) give the AI more info than the player would for the same situation

It doesn't have to be perfect intel; just perhaps a little better than the player would have. So a successful night move with NO contact might still be 100% effective; but if perhaps only one unit blundered into an enemy outpost, instead of a vague contact the AI might get a better report, which enables it to react to the contact IN A SIMILAR MANNER to a human.

Obviously making enemy troops supermen is a stupid balancing decision, and would imply a horrendously weak AI; but overestimating AI force effectiveness by, say, 5-10% to compensate for slightly worse defensive positioning by the AI would be perfectly fine in my book.

No simulation can EVER be a perfect model; it must consist of compromises, and in some cases counterbalancing errors may exist.

For example, who's to say the effect armour penetration of a 17pdr is right? But it doesn't matter, as long as the effective armour values of the German tanks correspond so that the chance of brewing up a Panther is correct (ish). I don't care if it should be 75mm of penetration against 60mm of armour, and the game considers it 90mm vs 75mm. As long as my At troop takes out the Panthers in a reasonably historically correct fashion, who cares?

[Arjuna]

MadScot,

I appreciate your point of view. It's the same view adopted by most designers. The line of least resistance if you like to achieving the effect you want. I refer to this approach as the "design for effect" approach. Personally I eschew this, in part for the same reasons Mark Weston mentioned but also because I can't stand playing against a computer player that "cheats". How I hate that. [;)]

The main reason however is that to support a hierarchical command structure the Player has to be able to rely that his AI controlled subordinates will do a reasonable and realistic job of managing their assigned resources to achieve the objective/orders you have set. If you knew that the AI could cheat you would soon deduce that you would be better off letting it run the whole battle. That's not my cup of tea. So we don't let the AI cheat. We try very hard to have it reach the same decisions a human player would using the same info that would be reasonably available to a human player and no more.

Getting back to OpAreas, so once you introduce these for the human player you do need to come up with a means of generating these on the fly for AI controlled forces. They need to be tailored for the current orders, forces, terrain and conditions. Yes it is a difficult subject but then again so was getting the AI to conduct complex attacks, delays and secure tasks. Difficult but not impossible. Rome wasn't built in a day, nor ten years, which is what we are coming up to for our game engine. OpAreas will get modelled I'm sure. The obvious time for their inclusion is when we do Team Play. [:)]

[Golf33]

A brief note on Steel Beasts: there is no AI in this sense in Steel Beasts. The movement of all non-human-controlled units is totally controlled by routes and conditions established by the scenario designer at design time and by the player at run time. If the scenario designer does not plot routes for the enemy into a particular area, they will never go there. 'AI' in Steel Beasts, such as it appears to the player, is simply the result of the scenario designer providing the computer-controlled units with a number of movement options which are triggered by various situations during the game (which are also all specified by the scenario designer). If the scenario designer didn't take into account a particular possible approach, then the computer-controlled units will never react to that approach except at the lowest level (by firing on units as they are spotted). They will not manoeuvre in response.

In HTTR, things are totally different, since the AI looks at its objectives, the terrain, and the enemy, and develops a plan based on those. It then periodically reassesses the situation and will modify the plan if necessary. The scenario designer doesn't have to do as much work in this case: it's more a matter of assigning objectives and priorities to the AI to influence its decisions rather than telling it exactly where to go under exactly which circumstances (which is in any case not possible in HTTR).

So I'm not misunderstood as attacking SB (which I have and greatly enjoy), this isn't a problem for that game; it's a deliberate design decision to limit any AI to very low-level tactical decisions like finding a hull-down position along the line of march, or deciding which target to fire on first (something which SB actually doesn't do very well at all).

Regards
33

[MadScot]

Ah, I forgot friendly AI; I agree that having an AI handicap when control of friendly units can switch from AI to player and back again is very difficult to implement - taking the arty ammo example, you'd be adding and subtracting ammo from units every time they got direct orders or went back to Ai control.....friendly AI kills off AI design-for-effect when it's switchable....

However...

If you knew that the AI could cheat you would soon deduce that you would be better off letting it run the whole battle.

Since we're discussing improving the AI so it handles a new game rule, can I change that to....

If you knew that the AI was as or more competent than the player you would soon deduce that you would be better off letting it run the whole battle.

Being in a position where the AI is as likely to succeed as the player is not, in itself, a bad thing. The most efficient way for me to play chess against Fritz would be to use a second copy of Fritz, unless I happen to actually be a grandmaster-class player!

Ideally, the AI will become so good that I will not have to issue Coy level commands (just like we don't have squad-level commands right now). The absence of squad level tactics and commands is a 'design for effect' decision, it's abstracted away because it doesn't affect player behaviour or the experience of the game.

To use another example: reviewers get all giddy about 'X-plane' because it uses strip theory to calculate the vehicle aerodynamics, and this is deemed 'more realistic'. However, the aircraft industry doesn't do it like that at all: commercial simulators are very much 'designed for effect' not from aerodynamic first principles. If I were to suggest we used the X-plane approach I'd be laughed out of my job.