Jump to content

Could this be the first pic of CMx2 WWII ? [Edit] Yes!


Recommended Posts

All good, and hard to disagree with. Except, and on the other hand ...

... introducing these kinds of delays would also tend to quash quite valid 'recon by fire' tactics. Imagine a situation in which you have three platoons advancing to contact in line abreast. The platoon in the centre has the Coy HQ co-located and therefore no minimum delay. The other two platoons are out of command to Coy HQ, but have full platoon-level C&C. None of your men have spotted the enemy.

The platoon on the left (out of company C&C remember) approaches a line of buildings. There is still no enemy seen anywhere, but you decide that the platoon should do a deliberate assault on the buildings, so you set up some overwatch from within the platoon, designate assault groups, give orders to the fire support and ... wait. Wait for the firesupport to kick in because there is no C&C to higher, even though that link is irrelevant in the current context.

It's easy to extend this idealised and simplified situation to more complex and general situations. The basic problem is that there is no unique identifiers for "co-operative behaviour". An HMG can support a section from across the map, while the section right next door might be completly on it's own program. In the sceen cap above, the Stryker right next to the MG firing across the map could be given orders to fire on the building next to the one the MG is firing at, in order to suppor the advance of the troops lined up against the right map edge. What delay should the Stryker have to fire.

Similarly FOs can support units from miles away, without LOS to anyone or thing, or then again they might be firing at exactly the same location for reasons entirely unrelated to what the infantry are doing. How do you determine - as a general case - what delay applies.

We are asking the AI to infer intent and apply delays based on that intent, but intent is notoriously fickle and difficult to nail down. The current solution is less than ideal because it's a compromise, but a different "solution" is in all liklihood just going to be a different compromise, which may be better, worse, or about the same, depending on the specific circumstances in which it's applied.

Then again, having said all that, I do think that increasing command delays for firing when out of C&C (or degraded C&C, or poor quality, etc) would be a Good Thing.

Jon

Link to comment
Share on other sites

  • Replies 297
  • Created
  • Last Reply

Top Posters In This Topic

Other Means,

We've debated Area Fire dozens of times over the past 10 years. There is no viable solution. An arbitrary, completely "gamey" approach (like artificial delays) has been suggested many times before and it's not something we are even remotely interested in implementing. It's far too artificial and completely unjustifiable. It just shifts the problem from one extreme to the other. So don't expect to be seeing it any time soon :D

Guys, there are natural limitations on what we can do to fix unrealistic situations within the game because the game itself is inherently unrealistic. Players enjoy so much more control and precision than can be achieved in the real world that at some point trying to "fix" things becomes a counter productive distraction. Area Fire is one of those things... I think by now all possible solutions have been floated and none are worth pursuing.

Steve

Link to comment
Share on other sites

Jon, you have interpreted my notion as operating at the other end of the spectrum i.e. no more set orders so they just pack up and go home.

No, not "pack up and go home". Rather "pop tents and dig in". In other words, when they get to the end of their chain of orders they stop all further movement, and just defend wherever they happen to fetch up.

You could counter that by having insanely long strings of orders, and in some ways that might be realistic, but it's a pretty fragile way of playing (though it'd be good as a command methodology demonstrator ... which is a remarkably different beast to a tactical game)

I do understand the negativity associated with potentially losing control of perhaps all of your units and the subsequent loss of gaming fun ...

Not at all, or at least not from me. I am down with the idea of introducing chaos and confusion into the game, and often suggest things like longer delays, random delay lengths, etc. However, I'm not in favour of chaos and confusion just for it's own sake, or because of some vague perception of the need for 'more of it'. Losing control of units should be for clearly understood, and controllable reasons, reasons that make sense and survive general case analysis.

Otherwise you're just annoying people for no good reason.

IMO

:)

Link to comment
Share on other sites

... hit/kill info is somewhat unrealistic to provide the player. We've had that discussion a lot over many years, which is why it was made optional in CMx1. It's the sort of feature that people like or don't like, so it should be optional.

Have you considered making it a settable, pre-game-start option that cannot be undone once play commences? Hotkey-hiding it is ok, but depends on mutual honesty if it is to be anything other than a mere personal preference.

Link to comment
Share on other sites

JonS,

Not at all, or at least not from me. I am down with the idea of introducing chaos and confusion into the game, and often suggest things like longer delays, random delay lengths, etc. However, I'm not in favour of chaos and confusion just for it's own sake, or because of some vague perception of the need for 'more of it'. Losing control of units should be for clearly understood, and controllable reasons, reasons that make sense and survive general case analysis.

Otherwise you're just annoying people for no good reason.

Exactly right :) This is what I call a "command level" game and it is a radical departure from what CM is. The market for command level games is very small because the vast majority of players do not want control taken away from them. Not even temporarily. And if control is taken away from them, then their units had better damned well behave in a way consistent with the level of realism the game is supposedly trying to achieve. This means a complex scripting language (far more so than what we have in the Editor now) to have the units be capable of autonomous, coordinated action if the player loses contact.

That autonomous behavior requires more AI programming than any commercial wargame developer could ever muster AND still have a detailed game environment. Or put another way... it's an either or proposition. Either you have a fairly simplistic game system with a very complex AI model to simulate "fog of C2", or you have a fairly simplistic "fog of C2" system and a very complex game system. There is no possibility of having both as far as I can see. Certainly we're not stupid enough to try ;)

Have you considered making it a settable, pre-game-start option that cannot be undone once play commences? Hotkey-hiding it is ok, but depends on mutual honesty if it is to be anything other than a mere personal preference.

With the Play Modes of CMx2 this is easily done. Iron Man, for example, could be prevented from seeing detailed hit/kill information no matter what. In fact, I think that's a good suggestion someone up above had. We could, in theory, have pre-game user set variables (like Fog of War) which can't be changed mid game, but I'm not sure we want to niggle with fairly minor ones like this.

Steve

Link to comment
Share on other sites

Dan, I don't disagree with you in regard to an AI carrying out all or any of those tasks at the same level as a human; that is generally a given. However, I'm not sure that it would have to, only at the same level as it plays the enemy. Would this break the game, as you suggest?

James, I think we already have an indicator as to the answer to this if we look back at all of the CMx1 and CMx2 rants where the Tac AI did the wrong think and get a unit killed :)

The Tac AI is just controlling 1 unit, in one situations (usually only for a few seconds to respond to a threat). If we multiplay that complexity over an entire platoon or even company over minutes of time the problems would be multiplied exponentially.

As I said its a cool idea in theory, personally Id love the option to be able to provide the game with general orders and watch the battle play out. The problem is that AI just cant handle the complexities of such a simulation.

Dan

Link to comment
Share on other sites

If BF go down the MMO route I, for one, will be heading rapidly in the other direction. I'm fairly picky as to who I share my limited leisure time with; the idea of playing on-line with 1000 potential muppets makes my blood freeze!

Hey you sound a bit anti-social :D J/k. You can still pick with which of the 1000 muppets you want to team up. If you havent had the chance to fly in a virtual squadron you haven't experienced true and mature multiplayer. Heck, we even did navigation and formations training, distributed specific roles in the campaigns, (Escort, bombers etc) and most things were done according to plans. And most co-pilots were beyond my age with families and kids. Sure, jerks exist out there too but if you get to know a guy or two and cooperate things become much simpler. No need to arrange anything either if there is a proper lobby to find people to play.

Link to comment
Share on other sites

No, not "pack up and go home". Rather "pop tents and dig in". In other words, when they get to the end of their chain of orders they stop all further movement, and just defend wherever they happen to fetch up.

But then, that's exactly what happened in most armies, at least part of the time. As James Crowley is suggesting, if out of CC they would try to fulfill their last order and take or hold their objective. If they succeeded, they stopped there. If they did not succeed, they fell back until they had a defensible position.

Of course, you could make their order (this is RL I am talking about here) one that says, "Take objective X and keep going until told to stop." But this was not so common. I haven't considered whether giving such an order in the game would be feasible or not.

Michael

Link to comment
Share on other sites

As I said its a cool idea in theory, personally Id love the option to be able to provide the game with general orders and watch the battle play out. The problem is that AI just cant handle the complexities of such a simulation.

That appears to settle the matter...for the time being anyway.

Michael

Link to comment
Share on other sites

But then, that's exactly what happened in most armies, at least part of the time. As James Crowley is suggesting, if out of CC they would try to fulfill their last order and take or hold their objective. If they succeeded, they stopped there. If they did not succeed, they fell back until they had a defensible position.

There are orders, and then there are Orders.

In Real life , prior to moving out commanders at all levels will give orders to their subordinate commanders or their men. The format would typically be some variant of G-SMEAC (Ground, Situation, Mission, Execution, Admin&Log, Command&Sigs). This would provide the people receiving those orders enough to be going on with for anywhere from 1 hr to 1 month, depending on the nature of the task. During the conduct of the activity the commanders will be giving fine tuning orders to cope with certain contingencies, but at all time the overall orders (and especially the Mission) are primary. And yes, once the Mission encapsulated in those orders has been acheived, untis will tend to sit down and defend in place (if that's what the mission specifies - they may equally return to base, or commence some other activity). It's possessing this mission (which encapsulates the commanders intent) that allows units - real units - to think for themselves and carry on when out of C&C.

In CMSF there is no way to do this. You, the player, have to hold the mission orders inside your head and enact them through the fine-tuning orders you give during the course of play. The fine-tuning orders are the only ones the AI knows about. Let's assume the AI can carry out the fine tuning orders flawlessly, nevertheless, at the end of that sequence of orders it has nothing else to work with, so it will sit the units down and they'll defend in place whereever they happen to be, and regardless of what the mission is.

It's the higher level of orders that is missing, and why the AI can't - IMO - just be left to run things by itself. To go back to your example, you cannot give an order along the lines of "take objective X", and then overlay that with the fine tuning micro orders necessary to carry it out. All you can give are context-free micro-orders.

Link to comment
Share on other sites

Personally, I think JonS has given some of the most reasonable and well-phrased arguments in this post. I concur 100%. with you, JonS. I won't quote, but whoever didn't notice this, I urge you to go back and read his stuff in context with the posts he was referring to, and let his words sink in a little.

Add to this what Steve and Kwazydog and a few others (you know who you are) have said and I believe you will see a complete answer to the question of how the player and the AI can interact in a game like CMSF. And remember that literally none of us have ever really played with the AI as stipulated by several people in this thread. Anyone who plays SP (most folks, I believe..me for sure :) ) has all of his experience with the AI from playing against it, and most of us, well, generally, we win. And when we do something stupid and lose, we reload or start over, and we learn from our mistakes on a tactical and strategical level, and we try not to make those mistakes again. The AI does not, and in the scope of CM, very probably never will have, this capability.

Go figure. AI as your coop will never, ever make you completely happy. It's hardcoded and inflexible, no matter how good it is. :)

It's the player's capability to act as an overlay in the sense of training, common sense, wider strategic goals, unit self-preservation, situational awareness and a host of other important motivations that only a human can understand, above and beyond the raw orders received at the immediate briefing (or given to the AI in the editor), which makes the individual in-game challenges surmountable and the game fun to play for us human players. Only a human can continually interpret the underlying meaning of his objectives on the fly and lead his forces to victory by use of all the resources that are naturally available to a RL commander (be it on a battalion, company, platoon, squad or individual level).

Link to comment
Share on other sites

What it all boils down to is you can not convey "Commander's Intent" to the AI without a really powerful scripting tool which, in turn, is connected to some fairly complex AI sub routines. For a command level game to work the way the player expects, you'd need to be able to tell your subordinate commanders very specific things that fit into the context of a bigger plan. For example, you probably need to instruct B Company to do one thing while instructing C Company to do something completely different like...

"B Company Commander, you are tasked with providing support for C Company's attack and seizure of Hill 193. You are in a supporting role and are to avoid direct contact by default. The plan is for you take your forces around to the north of Hill 193. Do not travel down Route Fluffy because C Company needs to move its armor up and the road is incapable of handling more traffic than that. Once you reach Objective Tweety debark your infantry and set up a base of fire covering the top of Hill 193. Hold your vehicles nearby, but not within LOS of the hill since it is believed there are AT assets up there. The FIST is assigned to C Company so that it can lay down suppressive artillery on the hill at 08:20. The bombardment will last approximately 10 minutes. 8 minutes into the barrage move your vehicles so they can fire at the ridgeline and get your infantry up to Objective Sucker as quickly as possible. Once the artillery has let up remain in position as C Company starts its advance from Objective Mealworm. Do not move from your positions unless C Company requests aide or if the enemy appears to be attempting to move off the hill to the east..."

So on and so forth. And that's just for B Company and even then perhaps only the first half of the operation!! Now, in real life this is the sort of instructions that would be laid out and if you, the Battalion Commander, got fragged or stuck in the porta-potty without a radio link, the attack would go on just as described to the best ability of B and C Company Commanders. B Company's forces would not just sit around in the woods short of their objectives because nobody had yet issued them specific Commands to move up the hill, nor would they sit around forever in those positions doing absolutely nothing in the event C Company got into a pickle. Why not? Because not only did the Battalion Commander (the player) give some specific instructions and time tables, but he also told each Company how they should generally behave so that they can adapt their actions to best suit the conditions without buggering up the operation.

So on and so forth.

The thought of how much time we'd have to put into AI programming to get this to even work half-assed some of the time, and fall flat of expectations most of the time, is mind boggling. And the truth is that even if we put all of the time and effort into it, and made a near perfect AI, I bet 90% of our customers would turn it off and play as God. Why? Because more players want to be God than not. Which is why we're not going to touch this one with a 10000m pole ;)

Steve

Link to comment
Share on other sites

Ummm... what about the recondo platoon? Surely they should be ahead of Coy B with eye's on reporting the conditions of Objective Tweety and if possible observation of Hill 193? Would Coy B not have some mortar support to put down smoke if they got caught out and attacked while approaching and finding a way to bypass Route Fluffy?

These things matter to the grunts at the fronts. I think the AI would have to look at support and FUBAR fixing on the fly, Steve. :)

Link to comment
Share on other sites

Ok ok - I get it! If this was a straw poll for AI sub-command then I guess the nays have it.

Perhaps my logic is flawed. I thought that if the existing AI can do a reasonable job with handling the opposition, then it should be able to do the same for out-of-command friendly units as well. Not the perfect, all-singing and dancing job that Jon and Steve are suggesting, just a get-by.

If a get-by, as Steve suggests, is unacceptable to most players because it is not top-notch then, presumably, the current AI as opponent is not top-notch either and 90% of players don't play solo?

I think most people just accept the shortcomings and get on with it.

And 90% of players don't use full FoW/extreme FoW in CMx1 or Veteran/Elite in CMx2? Hmm... my perception of CM players is that they will embrace any reality-enhancing features that are available, most, if not all, of the time.

I can accept that resources may not allow development in that direction. I can equally accept that the producers are not inclined in that direction. I don't accept that it isn't developed because players wouldn't use it. But what do I know?

Link to comment
Share on other sites

Regarding area fire:

Maybe a simple compromise would already reduce unrealism a lot? What about an optional general delay of a few seconds for area fire orders - say 15 or 20 seconds? (could be dependant on troop quality, and vision equipment)

And if it is easily implementable, certain modifiers of the command-chain could additionally reduce it: is the unit in command of a HQ?

Another idea:

Make area-fire a bit fuzzy-logic controlled: a certain chance, that the unit does not fire exactly where you order it to fire (another group of trees nearby, adjacent window/house). Appearance of this faulty behaviour weighted by troop quality, equipment, HQ-command,...

Link to comment
Share on other sites

I can accept that resources may not allow development in that direction. I can equally accept that the producers are not inclined in that direction. I don't accept that it isn't developed because players wouldn't use it. But what do I know?

James I took several courses in ai development during my degree so I have a reasonable understanding of the subject when I suggest that you are likely expecting far too much from what an AI can handle whilst still providing they player with a realistic and enjoyable experience.

As I mentioned, look back over cmx1 and cmx2 posts on the difficulty of having the tac ai react appropriately to just a single given situation as an example. If you then multiply those problems by 100 you'll just be starting to scrape the surface of the complexities and the amount of work involved in such a system.

Dan

Link to comment
Share on other sites

James I took several courses in ai development during my degree so I have a reasonable understanding of the subject when I suggest that you are likely expecting far too much from what an AI can handle whilst still providing they player with a realistic and enjoyable experience.

Dan

Dan, firstly thank you for taking the time to reply. I have always been appreciative of the effort that the little gang at BF (and previously BTS) have made to discuss all matters CM with their customers. I hope my suggestion and comments are not seen as overly critical or judgemental.

My knowledge of programming, let alone AI development, is non-existant so I readily accept your explanation, which is broadly in-line with what other developers elsewhere have said in the past.

I suppose one always hopes that things can and will be improved over time, in much the same way as computer graphics have over the last ten years. However, the latter has been more of a product of ever faster processors than anything else and I get the feeling that games based AI has not kept pace with the bounds in computer technology, in general.

Perhaps it is because it follows the laws of diminishing returns, as Steve has suggested; too much time and effort for too little obvious return.

A shame because that is leading, IMO, to a situation where many (the majority?) of games come with a sort of generic, vaguely passable AI , damned by faint praise even by their own designers and the usual byline of 'AI is ok but nowhere near as good as a human player'

One can see a point being reached where AIs will be dropped completely and the only option will be online gaming. And yet according to many polls and assorted statistics, the majority of computer gamers play solo. Curious.

I have played many PBEM games and a smattering of HtH and Lan and, yes, a human player can often add a far greater challenge than an AI. Is that good? I suppose it depends on the format of the game. I have taken part in a couple of the old RoWs and from a competitional, gaming-to-win perspective it absolutely is.

But I often prefer to play from a historical/experiencing point of view and that can often be 'subverted' by 'clever' humans doing things that were not historical and, ultimately, not realistic. So, for me, the best way to get that experience is to play against the AI which, ok only sometimes, manages to stymy my best efforts (one or two who may read this will say I am just a crap player).

Well, I'll put this one to bed for the time being.

And congrats on that excellent Tiger, which is where this all started!

Link to comment
Share on other sites

With the Play Modes of CMx2 this is easily done. Iron Man, for example, could be prevented from seeing detailed hit/kill information no matter what. In fact, I think that's a good suggestion someone up above had. We could, in theory, have pre-game user set variables (like Fog of War) which can't be changed mid game, but I'm not sure we want to niggle with fairly minor ones like this.

Steve

Steve,

Are you talking here about a "custom" difficulty level setting that would allow the player to toggle various options on and off? I think this would be a good step forward for the game series, and in line with most other high-fidelity wargames and sims.

Link to comment
Share on other sites

James,

Yeah, definitely each time this discussion comes up the bulk of the people here definitely express two things:

1. An inherent distrust of the AI to do anything other than the wrong thing :D

2. The "unfunness" of having control of the battle wrestled away from them with no certainty that it will return to their control.

I've said in the past, and I'll say it here now, that theoretically I would LOVE to have a command level game. I'm one of those guys that this appeals to, at least in theory. But I don't think it's practical to overcome problem #1.

Perhaps my logic is flawed. I thought that if the existing AI can do a reasonable job with handling the opposition, then it should be able to do the same for out-of-command friendly units as well. Not the perfect, all-singing and dancing job that Jon and Steve are suggesting, just a get-by.

There's a huge difference... the AI Player doesn't have to consult with anybody... it does whatever it feels like doing and has no restrictions on being able to do it. Just like the Human Player :D What is needed, to make a command level game playable, is an interface between the Human Player and his subordinate AI Leaders. That in turn requires a scripting language which would necessarily be quite complex. We can in fact do that, but the amount of time it would take rules it out from a commercial viability standpoint.

Or put another way, if some department/ministry of defense gave us a pile of money and enough time to work with, we could definitely do it. But the pile would have to be very big because we could instead do a lot with our time in the commercial market.

If a get-by, as Steve suggests, is unacceptable to most players because it is not top-notch then, presumably, the current AI as opponent is not top-notch either and 90% of players don't play solo?

I think most people just accept the shortcomings and get on with it.

But they aren't really shortcomings if you have an excellent plan and, through bad luck or some other circumstance, you lose control of your subordinates and they do nothing but sit around NOT carrying out your plan. Or worse, being stuck in an extremely precarious spot and by the time you reestablish contact they're combat ineffective. I'm 100% convinced this would get old really, really fast.

And 90% of players don't use full FoW/extreme FoW in CMx1 or Veteran/Elite in CMx2? Hmm... my perception of CM players is that they will embrace any reality-enhancing features that are available, most, if not all, of the time.

Remember, perceptions based on this Forum are always skewed towards the type of player you described. But even then, the kind of system you're picturing is inherently unrealistic and therefore is reality breaking, not reality enhancing. That's because units must be able to operate autonomously and yet in concert with a higher plan in order for reality to be preserved. And that's the sort of coding stuff that could take us years to do even moderately well.

I can accept that resources may not allow development in that direction. I can equally accept that the producers are not inclined in that direction. I don't accept that it isn't developed because players wouldn't use it. But what do I know?

Players wouldn't use something inbetween in any large numbers, therefore anything inbetween is not worth the effort. Even a multi year development project to get the game beyond that point would only appeal to segment of the realistic minded wargamers because the reality of losing control of the battle, so utterly, is likely to cross a line with many people who play with all the currently offered realism settings set on full.

I suppose one always hopes that things can and will be improved over time, in much the same way as computer graphics have over the last ten years. However, the latter has been more of a product of ever faster processors than anything else and I get the feeling that games based AI has not kept pace with the bounds in computer technology, in general.

Perhaps it is because it follows the laws of diminishing returns, as Steve has suggested; too much time and effort for too little obvious return.

This addresses why AI hasn't improved as much as things like graphics. Graphics are easily improved and have a known impact on sales. AI, on the other hand, is EXTREMELY difficult to produce, the end result will not be known until development ceases, and the point of diminishing returns (in terms of sales potential) is hit extremely early on in the process. And that's the way us creative guys think... the big game companies don't even want to talk about AI designs because just the time to talk about them cuts into their bottomline :D

A shame because that is leading, IMO, to a situation where many (the majority?) of games come with a sort of generic, vaguely passable AI , damned by faint praise even by their own designers and the usual byline of 'AI is ok but nowhere near as good as a human player'

Leading? It's been this way since computer games started and it isn't likely to change :D If you want to boil it down, in broad commercial terms, "graphics sell games, AI doesn't". Which is why you find game publishers who spend more money on graphics for a single game than Battlefront has collectively spent on all of its development of all types in the past 12 years. Not to mention how much we spent on AI development :)

One can see a point being reached where AIs will be dropped completely and the only option will be online gaming. And yet according to many polls and assorted statistics, the majority of computer gamers play solo. Curious.

There is definitely a move towards this, but as you say... people still pretty much play solo. But they've also been playing solo on computers for nearly 30 years, therefore they know that AI isn't the thing that keeps people playing games solo.

To sum up... the idea you have, James, isn't a bad one. It's just an impractical one from our standpoint.

Steve

Link to comment
Share on other sites

AKD,

Are you talking here about a "custom" difficulty level setting that would allow the player to toggle various options on and off? I think this would be a good step forward for the game series, and in line with most other high-fidelity wargames and sims.

No, I wasn't suggesting customized settings but having it built into one of the existing ones. Now, I've always wanted to have toggle options for realism settings (have a nice dusty design from 2003 sitting right here, in fact :)), but it's never been high enough up on the priority list to get included. It still isn't, unfortunately.

Steve

Link to comment
Share on other sites

Guys, we're reaching the magical 300 post mark. We close threads up at that point because they tend, like this one, to shift around too much and focus gets lost. I invite any of you guys to create new topics for anything you want to continue discussion or to raise issues that haven't been raised already. I think we'd all rather have 10 topics with 5 pages instead of 1 topic with 50 :)

Steve

Link to comment
Share on other sites

Hi James

I suppose one always hopes that things can and will be improved over time, in much the same way as computer graphics have over the last ten years. However, the latter has been more of a product of ever faster processors than anything else and I get the feeling that games based AI has not kept pace with the bounds in computer technology, in general.

James I think Steve answered all of your points as well as can be. The bottom line is that unlike computer graphics, which is a problem of processing power, AI is more of a problem of concepts and the time to implement them. The theory behind those concepts has change very little in the last decade or so.

We have already seen a move away from AI in certain games...the Battlefield series for instance has a very basic single player AI that I find almost unplayable as do many RTS games. Personally I doubt we will ever lose it completely though as single player capability will always be an important part of the market.

Dan

Link to comment
Share on other sites

Guest
This topic is now closed to further replies.

×
×
  • Create New...