Case For New Type Of Chain of Command AI

[James]

At the moment we have the stratAI & OperationalAI which is a very rough abstracted chain of command with rough abstracted results: mixed up and scattered platoons, generalised inflexable attacks ect.

I propose a new approach to the AI in CM one that realistically simulates the chain of command. Starting at the top the say Batalleon AI/CO views the battle from his perspective and gives orders to his companies useing the SOP of his force. Each company then interprets its orders according to the situation and gives orders to coordinate the moves of its platoons as a real company CO would useing SOP and intitative. Like wise the platoons act on their orders and coordinate the moves of their squads. As the situation changes each level of command fights the battle from its perspective. Each formation should also have the intitative to adapt its orders to the situation with a view to preserving its numbers and cohesion. The quality of the HQ unit also directly affects the intellegence and initative in its interpretration and giving of orders.

This Chain of Command AI also gives the player the OPTION to give orders through any of its HQ units useful for those big senarios and those wishing to fight the battle from a different perspective.

So instead of 2 generalised AIs taking on too much, you'll get lots of little AI's working together greater than the sum of its parts.

I know this is a little late for a patch or its inculsion in CM2, but what about CM3? What do you all think of it? Has Steve & Charles thought of it, would they consider it?

[ 05-05-2001: Message edited by: James ]

[James]

Just bumping this back up. Any thoughts on this subject, anyone?

[JasonC]

CM is a strategy wargame. It is not a sim. It is not a roleplaying game either, but a strategy game. The appeal of strategy games is that the outcome depends on the interaction of match wits of the opposing commanders. Thus Go and Chess are successful strategy games, despite the fact that they simulate little or nothing.

The actual realistic impact that a single higher level commander has on a large, grand-tactical battle, in real life, is very small. The number of decisions between real alternatives (as opposed to just not screwing up), that one commander makes in the course of a 30 minute firefight, can be counted on the hands. If he is lucky a majority of them may actually be carried into effect; it can easily be the case that the majority have no effect for hundreds of local reasons.

One commander's decisions are not a rich enough set to create a challenging strategy game. Real officer's lives just aren't interesting enough. This is not a criticism of them; it merely reflects the reality that hundreds of individuals provide the relevant inputs, and the scale of impact of single commanders is necessarily limited in realistic fog of war situations.

A truly realistic simulation of life of one marine on Guadalcanal, would be torture not entertainment. A perfect sim would be a movie, and not a strategy game, because almost everything the commander tried to do would have little impact on unfolding events, driven almost entirely by interacting forces beyond his personal control. Action movies present a different picture precisely because they are romantic and unrealistic.

The way in which military wargames create the conditions for good strategy games, is that they allow the commander to make decisions that in reality occur at several levels of the military hierarchy. The impact of the combined decisions of a dozen sergeants, a handful of lieutenants, and one captain, on a 30 minute firefight, can be decisive enough (indeed, almost always will be), to result in an outcome that depends on matched wits rather than fate and fog-of-war.

Efficient military units manage to produce highly coordinate actions despite the decisions occuring in parallel at many levels. Strategy games take a real process that occurs outside the level of one man - the "corporate" "meshing" of reactions and commands with one another, in an experienced military unit - and puts that piecing-together inside the player's head. It give him conscious control over what is in reality a distributed process of cooperation.

It is that cooperation problem that contains the real strategy richness present in tactical warfare. Anything that carries the level of "sim" past it, below it, abstracting that cooperation process into deterministic or random game system outcomes, drains the result of its strategy interest. The "player" becomes a cog. The cooperation problem is hard-coded, according to the game designer's personal sense of what is a "realistic" way for it to go. The player feels like he is watching a movie the game designer has choreographed or directed. The outcome depends not on his wits and his opponent's, but upon fate or chance.

The only outright sims that are sufficiently interesting and challenging to be games, are ones in which a single individual has a vastly higher control over relevant outcomes, than in ground warfare. Thus, for a fighter pilot it can work. For a ship commander it is possible, with all the subsystems under him attacking as so many robotic multipliers of his decisions.

Even in those cases, trying to handle battles with higher numbers of ships or planes, tends to encounter the problems already mentioned. Thus 2 or 4 plane sims work, but flying 1 plane in a multiple-squadron dogfight still leaves the "chance-fate" outcome.

Other tactical sims, as shooters and RPGs, simply adopt the action movie formula, and handle the problem that one person's decisions do not realistically have much effect, by magically pretending otherwise. But that is not realism, obviously.

Game design is not the same thing as sim design. It is a different job, and a much harder one. Literalism, engineering, programming, choreography, do not solve the problem of game design, and in practice they are its greatest (because most common) pitfalls. The job description of a game designer is to -design-, a -game-.

Which means, he determines what controls to place in the hands of players, and how the interaction of inputs from those controls shall lead to a space of outcomes. The outcomes can vary randomly, in addition to their variation from player inputs. But the player inputs must be by far the most important factors in the outcome, or the game will not have any interest, because the decisions it presents will lack weight.

The situations presented must be of a practically indefinite complexity, and variety. The interactive effects of move and counter, must not be fully foreseeable or subject to exhaustive analysis (else, the right decision is determined). But they must be partially forseeable and subject to some level of analysis (else the player could improve his play by flipping coins for his decisions. It is less predictable, which is an edge in avoiding -being- analysed).

And the result must be playable, and possess enough immersive realism to attract and hold interest, if the game is to be any "improvement" (as an additional game option, obviously, not a replacement) on Chess or Go.

These constraints are more serious than many who don't design games for a living suppose. There are not that many ways of handling player inputs, level of control, etc, that meet them. And many of them were found (by trial and error) long ago.

Most wargames made, fail. Most that do not fail, follow formulas that have been known for quite some time, with minor improvements or tweaks. This stability is not the result of any lack of imagination or programming skill to make more complicated games or more literalist sims. It is the result of the fact that those things are not improvements - in playability, in depth of gameplay, in replay value, in competitive interest.

[Stalins Organ]

Whatever CM is, it is most certainly NOT a STRATEGY game!

nothing at Bn or lower level even remotely qualifies as strategy!

[JasonC]

Don't be daft. Strategy is just greek for "what a general does", and when it modifies "game" it means not the level of military aggregation, but merely that decisions of commanders decide outcomes. Chess is not a strategy game because pawns "are" "divisions". Tactical, operational, or strategic levels of military operations, can and are all made into strategy games.

[aka_tom_w]

<BLOCKQUOTE>quote:</font><HR>Originally posted by Mike the bike:

Whatever CM is, it is most certainly NOT a STRATEGY game!

nothing at Bn or lower level even remotely qualifies as strategy!<HR></BLOCKQUOTE>

OK

crazy as this might sound I thought it was a TACTIC strategy game same as Tactics II was a tactical strategy game.

Is Chess not a Game of Strategy?

How about that Silly game Stratego? I thought that was a strategy game.

What about Star Craft is that not a strategy game?

just curious?

-tom w

[Fred]

<BLOCKQUOTE>quote:</font><HR>Originally posted by James:

At the moment we have the stratAI & OperationalAI which is a very rough abstracted chain of command with rough abstracted results: mixed up and scattered platoons, ...

[ 05-05-2001: Message edited by: James ]<HR></BLOCKQUOTE>

This is one of the main problems when the AI attacks (or even when shifting postions during a defense); the lack of unit cohesion.

It is more or less a coincedence, if squads of the same platoon are near each other and/or in command.

At least on the platoon level, the AI should handle the squads and HQ as one "tactical unit".

This is a case for the "Operational AI", and with some luck it will be better in CM2

Fred

[James]

Jason C: The main thrust of my proposal was that the AI SIDE uses the chain of command system to play a more challengeing coordinated and realistic game. I only added that the human player could use it as an OPTION if he wanted a different experience.

The way a CM player uses the system as it stands is exellent. I dont want to do away with it! I just believe that an AI simulating the chain of command realistcally would give a better AI opponant to us players.

[Warren Peace]

James,

I like your analysis and idea. I would like to have the ability to not micromanage every unit on the board in large scenarios.

Warren

[James]

Is there any chance steve & Charles will get to read and comment on all this?

[JasonC]

"I just believe that an AI simulating the chain of command realistically would give a better AI opponent"

Fair enough. To me it is an empirical question, whether it would or wouldn't. Off the top of my head, I can see it going either way. Maybe better coordination, or maybe more rigid, seperated, less flexible actions. The way to find out would be to build it and see.

[Username]

<BLOCKQUOTE>quote:</font><HR>Originally posted by James:

I know this is a little late for a patch or its inculsion in CM2, but what about CM3? What do you all think of it? Has Steve & Charles thought of it, would they consider it?

[ 05-05-2001: Message edited by: James ]<HR></BLOCKQUOTE>

There once was a thread similar to this. My contention was that computer platoons should use "drills" instead of the piecemeal antics that individual squads, mgs, etc now use.

To me, the computer player is modeling a novice CM player all the time. Cover, concealment, covering fire, etc doesnt come into his method of play.

In retrospect, the coding should have just had evryone rush the nearest enemy as soon as they found them. It works for me.

Lewis

[James]

I beleve it would work. AI structured around the chain of command has worked very well in the past eg: the battleground serise and the simulations of Dr Turcan. Both gave challenging and realistic results.

[Berlichtingen]

<BLOCKQUOTE>quote:</font><HR>Originally posted by JasonC:

Thus Go and Chess are successful strategy games...<HR></BLOCKQUOTE>

Chess is a game of tactics... not strategy