Is the strategic AI the bottleneck of CM game diversity?

[Nabla]

Any CM gamer can name a large number of new command and control concepts (s)he would like to see in CM. Consider, for example, the ability to dismount recon units from light vehichles; the ability to order an AFV to advance, fire one shot, create smoke cover, and retreat afterwards; option to order artillery fire for just 20 seconds; option to order a recon unit to pay special attention to a given battleground area; etc.

A lot of problems have to be solved when implementing such a new feature. I would imagine that once an order has been given in the game, if we discard data structure updating, graphics and other "straightforward" stuff, the game has to implement logic to do the following.

Phase 1. Determine the best set of actions to achieve the order - as in finding out a set of move commands to move an AFV to a given location if there are patches of wood between the start and end points.

Phase 2. Implement trigger conditions and actions which make the unit intelligence for 60 seconds - as in what to do when an AFV targeting an infantry unit has to do when it spots a tank.

The first, if I've understood correctly, is a small part of the operational AI, and the second is the gist of the tactical AI.

Now consider the new, suggested option of being able to dismount recon units from light vehichles. The phase 1 implementation of this feature is basically equivalent to ordinary embark / disembark logic. The phase 2 implementation is very close to ordinary infantry tactical AI, with perhaps some tweaks to prefer retreats and hiding.

That does not sound too bad, does it. If these were the only thing that would have to be considered in implementing a new C&C feature I bet we would see a lot more of them in CM. But my understanding is that the real work is elsewhere: the tweaking of the strategic AI. As I stated above, implementation of a new feature amounts roughly to the implementation of phases 1 and 2, assuming that the order has been given. However, if one side is played by the computer, the order has to be given by the strategic AI.

If I was working in BTS (just a dream ) and I was given the task of implementing the new recon feature, based on my nine years of experience in computer science I'd be glad to implement phases 1 and 2, but I'd be scared to implement the strategic AI part. Not because it could not be done in some way (and it would be very interesting), but because I'd be afraid of customer reaction. As BTS has stated over and over, it is very difficult to implement even simple human-like decisions, and the recon dismount decision is not simple.

So what's the point? Because there's a real danger in saying these things I'm stating first that I understand that these are my opinions, and I understand that there are value judgements behind some of these opinions, so there is no right or wrong here. Please remember this before you hit me with a hammer.

The first point is the following: my understading is, that if there was no strategic AI we would see a lot more command and control options in CM. Ok, this is not a value question, but it's still an opinion. But the second point is even more controversial, so I state it explicitly as my own value: my current understanding is, that CM would become a better game for me if the strategic AI would be given up, and only PBEM and TCP/IP games were allowed. I estimate that this would result in many more advanced features, ones which will never be implemented as long as the game can be played against the computer. I estimate that the joy of having advanced C&C features in the game would greatly exceed the sorrow of having to give up games against AI, especially because the number of PBEM and TCP/IP players is so big that there is no trouble in finding an opponent nowadays. This might not have been possible when CM1 was introduced, but it would be possible now because a large gaming community has developed around CM. I might even go longer and state that some gaming company will do this if BTS will not, because of the popularity of human-human games over games against AI.

This is my opinion, and I'd like to know what the opinions of other members of the CM community are. After reading this message once on the board I came back to add that I do not want to steal a great game from those that have no network / email connection. I want everyone to be able to enjoy CM , and the strategic AI works just great with the current set of actions. But that still does not change my gaming experience. Perhaps we would indeed get the best solution if some other gaming company would make a pure multiplayer CM clone. Who knows...

[This message has been edited by Nabla (edited 03-15-2001).]

[jshandorf]

It is MY feeling that people want TOO much control over each individual unit. So much so I think sometimes they want a "Fart", or "Belch" command.

I am perfectly happy with sending a unit forward and telling it to "Hunt" knowing that it will see, hopefully, the enemy tank, engage it, and then win. I don't want to be telling it which rounds to use, when to pop smoke, when to back up, what target to engage, which way it should be rotating the hull, etc...

I want the unit to figure this out for me. And I know that in letting it do so it can make "mistakes". Fine. I expect mistakes. Sh*t happens. Okay, but what I don't want is blatantly stupid behavior that is repeated over, and over, and over again. I don't believe this happens very often, that is, if at all.

I believe BTS has corrected most of the real errors in CMs tactical AI, and I am very happy with it, and therefore i don't see a real need for finer granularity in the control of units in CM. Let them fight it out! That is the exciting part! That is, never knowing exactly what is gonna happen next.

Jeff

[Terence]

I like to play against the AI, not because it is always the most challenging opponent, but because it is always ready to play, never argues with me and does its turns really fast. If this game were multiplayer only, I wouldn't have bought it.

I play to relax after work and don't need the additional hassle of lining up an opponent who may or may not want to fall in line with my plans for the evening. PBEM and TCIP (which I have never tried ) are the icing on the cake for me -- really really fun wonderful things but they will never be the majority of my CM play.

[Croda]

I think that people tend to forget exactly what the game is simulating.

When you play CM, you are taking over the role of the Bn, Co, or Platoon commander. You give orders, and the soldiers (played by the AI) carry out those orders.

It seems to me that what people really want is to merge this with an FPS type game and take greater control over the units doing the actual fighting. That's not the game we bought.

CM measures your ability to make tactical decisions, not your accuracy with an M1. You give orders, the AI carries them out. It's that simple and it shouldn't change.

This is a game for Lieutenants, Captains, and Colonels, not Privates. The game models command structure only, which is why we don't see detailed combatant information like health and stamina that you see in FPS games. In CM it is the unit that is of utmost importance, and not the individual.

So if you're wondering what the hell I'm talking about, and most likely you are, I'm saying I like it as it is. I want the AI to do what it is doing. I want to call the shots and let the AI carry them to fruition.

[Vergeltungswaffe]

Originally posted by Terence:

I like to play against the AI, not because it is always the most challenging opponent, but because it is always ready to play, never argues with me and does its turns really fast. If this game were multiplayer only, I wouldn't have bought it.

I play to relax after work and don't need the additional hassle of lining up an opponent who may or may not want to fall in line with my plans for the evening. PBEM and TCIP (which I have never tried ) are the icing on the cake for me -- really really fun wonderful things but they will never be the majority of my CM play.

Amen!

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

Give a man fire, and he's warm for a day. Set a man on fire and he's warm for a lifetime.

[Terence]

Originally posted by Vergeltungswaffe:

Amen!

minor correction. I have tried and like PBEM very much. I have never tried TCPIP.

[Herr Oberst]

Forget more complicated, extended orders, except perhaps for a "Hunt" or "Advance Until Engaged" order for infantry.

Highest on my list of wants, and probably the most difficult to implement for Charles et al. would be giving the various units "memory" from turn to turn. The amnesia that strikes AFV commanders between turns makes for some, shall we say, "interesting" considerations to gameplay.

From the standpoint of implementation musings, for each unit, you would need a way to serialize the turn-ending "knowledge set" of each unit. What the unit saw, where it saw it, and how lethal it seemed to be, lethal being a relative term subject to storms of debate.

The start of each turn would consist of de-serializing the knowledge set and reconstituting each unit in its entirety prior to kicking the strategic or tactical AI routines. That way, a CM turn ceases being a "break" in the action, and becomes a snapshot "pause" in the action at a given point in time.

The downside? Turn files are big, big, big.

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

To the last I grapple with thee; from hell's heart I stab at thee; for hate's sake I spit my last breath at thee...

[Maximus]

Now is this coming from the Inner or Outer Croda?

But anyway, well said.

I think the issue lies in that people are really not aware of how powerful the Tac-AI really is. What I tell RTS gamers about the Tac-AI is that it does what a RTS gamer has to do manually. And that is like you said, pop smoke, reverse back down a slope out of view, switch targets, etc. IN an RTS game, the human has to do all of this manually which is why most RTS games boil down to click fests. It is because the human is constantly having to micromanage every little action that his cyber troops do. In CM the Tac-AI does this for us. Which in turn creates a much more realistic battle. I mean just the action of the CM troops going prone to duck bullets in itself is pretty amazing.

Even Steve said once that the game was unplayable without the Tac-AI because the troops would just sit there if they didn't have a direct command to fire or what-not.

To me, all of these requests for more micro-managing orders, are not necessary. Because if you have a good understanding of the orders we do have and how they work, you can get the units to do whatever you wanted them to do with the requested specialized orders.

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

For your dream car click here.

For a Close Encounter click here.

Hey look! I can see my house!

And for all you Hamster Lovers out there, check this out! Kitty, this one's for you!

[dalem]

Wow. I can agree with Jshandorf and Croda on the same thread. Kinda makes you think, don't it?

-dale

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

I LOST A BLOOD HAMSTER BATTLE TO PETERNZ. MY SIG FILE IS NOW HIS AND I AM HIS SLAVE. PLEASE ABUSE ME AS YOU SEE FIT AND CALL ME A DOO-DOO HEAD. IT IS MY PLACE IN LIFE, I WILLINGLY ACCEPT IT.

[David Aitken]

What Shandorf and Croda say is very valid. The one thing I will add is, consider how you gauge success in an FPS compared with CM. In an FPS, every health point you lose matters. If you die, that is the ultimate disaster (reincarnation aside). You strive to fight on without taking any damage.

If you try to apply FPS logic to CM, you'll always be tearing your hair out. Read any history of any major battle, and you'll realise that in war, things go wrong. Always. So when you order a platoon to attack and they are massacred, don't blame the game – don't even blame yourself – blame the realism of the simulation. That kind of thing happens all the time in real wars.

What also happens is that things happen which you, as the commander, can see happening, but are powerless to prevent. Your viewpoint in CM is that of high-ranking commanders – the people watching the battlefield through binoculars or receiving reports from reconnaisance aircraft or whatever. You see things which are about to go wrong. If you use FPS logic, you demand to be able to avert disaster. You want facilities which would have allowed you to save your men. But if CM is to remain realistic, it won't give you all the facilities you want. It will force you to sit back, issue orders, and watch as things go according to plan, or horribly wrong.

Bradley knew that Omaha beach was far more heavily defended than SHAEF had previously believed, but he couldn't call off the assault. He had to let his men go in, and let 3,000 of them die, and hope that at the end of the day he would have some kind of result. That's war.

David

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

All true Americans love to kill deer, wade into them, spill their guts, shoot them in the belly. When you put your hand into a gob of goo, that moments before was your best head of garden lettuce, well, you'll know what to do. – Bruno Weiss

[Nabla]

Reading the first replies and rereading my opening message I see that I've directed the discussion to micromanagement which is not the point here. My examples were not as good as they could have been. Micromanagement is not what I want, and I don't even want it to be enabled, because then micromanagement might be required to win in the game. The examples I stated have been expressed on this board by gamers, so I used them as examples of command and control concepts that CM gamers might like. But I do think there are lots of things that are not micromanagement which I would like to see.

Ok, so let me state some examples of these things I'd like to see which I do not think are micromanagement.

1. The ability to dismount recon units (the example I described in the opening message).

2. The ability to call artillery fire for only a period of time (or number of rounds).

3. The ability to hold artillery fire at a given position, and release the hold (avoid delay when target is known in advance).

4. The ability to call for moving artillery fire barrages (hope this is the correct expression).

That is, I'm talking about things that are impossible with the current game system. By doing micromanagement you want to avoid certain things. I myself don't want to prevent light armor from attacking a tank (although some people do). But in the current system recon units will never dismount, arty will always fire the whole minute, and will never move its target during that minute.

Hope this clarifies things.

[Croda]

Gotcha. You want to be able to issue slightly more detailed orders to the AI. Makes sense to me. All your ideas are good valid ones that would enhance the game with little to no detriment. Perhaps the engine rewrite will allow for some of these things.

[Jeff Heidman]

Funny how 95% of the psots in this thread just completely missed the point.

Jeff Heidman

[Croda]

Funny how you had nothing beneficial to add and were less congenial than the original misunderstood poster.

[Nabla]

Originally posted by Jeff Heidman:

Funny how 95% of the psots in this thread just completely missed the point.

Badly chosen examples in the opening message - my mistake .

[Leonidas]

Originally posted by Croda:

Gotcha. You want to be able to issue slightly more detailed orders to the AI. Makes sense to me. All your ideas are good valid ones that would enhance the game with little to no detriment. Perhaps the engine rewrite will allow for some of these things.

Hooray! Nabla may have won yet another convert to the cause of more detailed orders in CM. My list of detailed orders would be a little different, but at least we can agree on the basic principle more detailed orders are needed.

[CavScout]

Giving more order options is not always "micromanagment". Hell, in fact giving extra order options gets rid of some of the "micromanagment". For example, a "hull down" command. Is it "micromanagment" to use a "hull down" command as those against claim or is it "micromanagement" to zoom it and inspect terrain and emplace individual tanks yourself?

[Guest Michael emrys]

Originally posted by Nabla:

I understand that these are my opinions, and I understand that there are value judgements behind some of these opinions, so there is no right or wrong here. Please remember this before you hit me with a hammer.

[picks up hammer]

Gee, I hate to do this. You seem like such a nice guy.

[shrugs; puts hammer back down]

Since you have been generous enough to share your opinions, I'll respond in kind.

I never PBEM or TCP/IP. The reasons are twofold. One is the kind of games I care to play. I play to try to reproduce the historical "flavor" (though not any specific historical action, the reasons for which I will leave for some other thread), which means that they are inherently unbalanced. It is hard for me to imagine a human player so masochistic as to actively seek to play under those conditions. Secondly, the time I have available to play is so limited and so irregular, that a PBEM opponent would need a double dose of masochism. TCP/IP is completely out of the question.

Sooooo...for all its faults and shortcomings, the game AI/computer opponent is a real make or break item for me, probably even more so than Terence. If it were to be thrown out altogether, I would regretfully have to take my cash elsewhere.

That might not mean, however, that the strategic portion of the AI could not be disabled for person-to-person play if, as you say, it is preventing certain kinds of orders from being implemented that many people find desireable. Although how that would make the two versions of the game stack up against each other, I hesitate to say.

I would go conservative here and keep things as they are rather than lose the computer opponent, but if both can be done, then fine.

Michael

[This message has been edited by Michael emrys (edited 03-15-2001).]

[jshandorf]

Originally posted by CavScout:

Giving more order options is not always "micromanagment". Hell, in fact giving extra order options gets rid of some of the "micromanagment". For example, a "hull down" command. Is it "micromanagment" to use a "hull down" command as those against claim or is it "micromanagement" to zoom it and inspect terrain and emplace individual tanks yourself?

You have a point since I was one of those people a while back.

I would stare and suffer over which exact point my tank should go to be hull down and have the most visibility to the targets or his FOF, but regardless of my past experience I would still NEVER want a hull down command.

The reason for this is that now, since I have been playing the game for quite a while, I can put any tank of mine into a hull down position in about 5 seconds. I just zoom into the terrain take a look and issue the movement order.

So, in essence my experience in the game is paying off, and thus it works vice versa.

Also, being able to hold your arty target is not only probably unrealistic it would also be a crutch for the more inexperienced players.

I feel that all these commands, asked for above, only make the game easier for the less experienced player, and then therefore takes the advantage of being a veteran player away.

When I first started playing CM I was horrible at fire missions and predicting my enemies advance and defensive positions, and therefore my arty was really ineffective. But in time as I gotten better, and learned more, I have had more and more success.

I believe the way CM is designed right now rewards the player with more experience in battle, which is as it should be. Adding in all this extra command functionality only "levels" the play field and therefore takes away the advantage a more experienced player has over a beginner.

Jeff

[Nabla]

It seems that the micromanagement discussion which my original message created attracted a larger audience than the actual subject.

Well, in this implicit bump let me just remind that what I'm saying does not apply to just command and control features. The same argumentation can be used when adding other new concepts with strategic importance.

Consider, for example, the addition of trenches, or complex sewer systems. Once again, phase 1 and 2 implementations for existing units and actions in these terrains seem much easier than designing strategic AI so that it knows when to send troops into the sewer system, or when to create a defence outpost using trenches.

Also, after sleeping over this thing, I would myself like to add that of course I hope that BTS will be able to implement all these things with the strategic AI. I'm just wondering if this is going to be very likely.

[Hurricane]

I also would have never bought the game if you couldn´t play against the computer. I haven´t tried PBEM or TCP/IP, even though I will when I get a bit more familiar with CM (only had it for about a month).

What I would like to see is an AI which is dynamic, i.e. can adapt to the situation and which learns from its mistakes. I´d love to see an AI built on a neural network, not on a few thousand lines of if-elses. Throw in a simulated memory, and the AI could become a very hard enemy to beat. I know it is nest to impossible to program such an AI, but maybe in CM 10?

[Olle Petersson]

WRT the title of this thread:

NO

It's not the strategic AI, but the tactical AI, that sets the limit.

I'd definately love to give more abstract platoon orders, and let the AI take it from there...

Unfortunately it's very difficult to implement suitable SOPs to be used by the AI for every situation that might come up.

A friend of mine is working on a game system that's supposed to work in that manner. All unit actions are supposed to be based on historical doctrine, and issuing orders is what takes time.

Micromanagement is possible but not adviceable, as the time to get orders through is generally too long to make them obsolete.

This game isn't computer based though, and it still require some tactical level combat simulation to solve the actual outcome of any combat situation. (CM might be an option...)

Cheers

Olle

[Nabla]

Originally posted by Olle Petersson:

WRT the title of this thread:

NO

It's not the strategic AI, but the tactical AI, that sets the limit.

I'd definately love to give more abstract platoon orders, and let the AI take it from there...

Unfortunately it's very difficult to implement suitable SOPs to be used by the AI for every situation that might come up.

Now we're getting to the subject. Thank you for your post

Ok, I gather that these standard operating procedures your talking about are the same thing as the trigger conditions and corresponding actions I mentioned above. In CM I'd say that the primary task of these procedures is to provide short-term intelligent action. That is, keep the unit alive and kicking for those nut-crunching one minute periods.

In CM you have to remember that we are talking about just one minute. If the time span is longer, then it's a whole different ballgame. The current SOPs the units are using are not that intelligent, are they. And they need not be.

Consider the current set of fallback conditions for a rifle squad that is targeting the enemy. I don't think that under heavy opposition the tactical AI would ever tell the rifle squad to try to get to the enemys flank, or to take a whole different attack avenue to a victory flag. Furthermore, I don't think it will tell another squad to come and help the unit that's facing the opposition. (Although I'm not entirely sure about this: friendly tanks do support infantry units in distress with smoke.) But the strategic AI has to make decisions like this, and the outcomes of such decisions are determined much later.

What the tactical AI currently probably does in the situation above is to keep firing, unless the unit takes too much damage. If it starts taking damage, try to take cover and perhaps retreat. If you compare these decisions with the larger decision above (flanking, different route, assistance from other units), which ones would you say are more difficult to program?

I hope the discussion continues now that we've had a good start

[Jarmo]

If we had these (or other) extra commands, would it be so terrible if the AI couldn't use them? I believe it could still do a good job anyway, it's not like they would totally change the way the game is played.

One exception being the infantry hunt. It would be unbalancing if the player could hunt and the AI couldn't.

Example of skipping an AI ability is the assault boats. AI can't use them. So it wouldn't be a first time AI is cut short.

[Olle Petersson]

I'm not limiting my thinking to the one minute turns, but to the response of one order.

Take for example a platoon of tanks that I want to advance along some route, seeking enemies and when found engage in combat (without advancing further).

* As it is today I'd have to plot leapfrog Hunt commands for the tanks, trying to find hull down positions or other cover when possible.

Once an enemy is encountered they either stop immediately and keep shooting 'til the enemy is gone, or they continue moving along the plotted course.

* My whish is to give an order in style of; Platoon, along this route, engage in combat, then wait for further orders.

Then the AI line up the platoon in a suitable formation and start bounding movement with overwatch.

Once enemies are encountered all tanks fire one shot, go for covered fighting positions, fire another shot, change position, ..., until the enemy is down. One section might try flanking movement and so on.

Then they wait for further instructions.

I think you get the picture...

In short: If I'm to play the role as company or battalion commander I don't want to act like a platoon or squad leader.

Cheers

Olle