Lack of abstraction and its implication on realism

[Northman]

I read a couple of posts a few days ago that stated, if I remember correctly, that the modelling of cover and concealment was WYSIWYG except for a small bonus that is applied based on the underlying terrain. As I've understood it this means that if a bullet's trajectory passes through someone then that someone is going to get hit, just like in any FPS. But if this is so, then wouldn't that mean that the realism of the game is heavily dependent on the AI of the individual soldier, i.e. his ability to find cover, the correct path, facing etc.? I've only logged about 5-6 hours with this game so far, but already I have the feeling that this game is not so realistic as one would expect, and it seems like the 1:1 representation in place of CMx1's abstraction is the main reason.

Any thoughts?

(Sorry if this has been brought up before; did a search but couldn't find anything specific about it.)

[Capt. Toleran]

Well, it would seem it is also heavily dependent on your skill to manage your troops and keep them out of the way of bullets. Their AI will do what it can, but if you send them charging out into a field, or on a rooftop surrounded by other hostile rooftops, the calamitous results of that decision will cost them, often before they can react.

Honestly, this is the most realistic CM game yet.

[Kineas]

I'm not sure if the CMx1 fire effects were more realistic. Remember the LMG effectiveness? Try now advance in the open in front of an LMG team.

In this world a lot depends on the skills of the AI soldiers, more precisely on the implementation of these skills. But that's another topic.

[Northman]

Originally posted by Capt. Toleran:

Well, it would seem it is also heavily dependent on your skill to manage your troops and keep them out of the way of bullets. Their AI will do what it can, but if you send them charging out into a field, or on a rooftop surrounded by other hostile rooftops, the calamitous results of that decision will cost them, often before they can react.

Ah, sorry, forgot that. Thanks for pointing it out.

[Northman]

Originally posted by Kineas:

I'm not sure if the CMx1 fire effects were more realistic. Remember the LMG effectiveness? Try now advance in the open in front of an LMG team.

Well, I play CMx1 daily and I think the LMG's effectiveness is pretty realistic, and I think the main reason for that is that the movement of each individual soldier is abstracted, so that the pathfinding problems and other AI issues which all computer games have to cope with does not affect it.

[Capt. Toleran]

Sorry Northman, I guess that came out wrong. I guess I mean more of what Kineas said, some of the stuff in CMx1 was highly abstracted and never felt right. You also have to remember that the spaces in CMx1 were significantly bigger, as the abstracted icons also implied abstracted terrain as well. This is more like a FPS with CM rules and management overhead.

Ever play a realistic FPS like America's Army, or any of the other semi-realistic ones? Someone breaks out into the open when you've got your SAW ready within 300 yards, he's toast. Very easy to lay down heavy surpressing fire, with a LMG, or pick off individual soldiers from similar ranges with the new scopes everyone seems to have. It's like everyone is a sniper now, not firing over open sights like in the WWII CMx1 timeframe.

Cover is very hard to come by in the early scenarios so far, but if you make it into a stone building, you'll be ok. Rooftops seem to be a risk -- better angles of fire, but much more exposure to being picked off. I usually just put a squad with Javelins on some far off rooftop to provide overwatch.

[CRourke]

I think I see northman's point. Imagine you have a squad taking cover in a stand of light trees. In CM1, they automatically gain cover. In CMx2, any advantages in cover are relative to the AI's ability to get each individual to take a proper position (ie, one that puts tree between you and incoming fire). CMx2's method should be more realistic, but only if the AI is good at its job. If not, you may see little to no benefit from your 'covered' terrain.

[Northman]

Originally posted by CRourke:

I think I see northman's point. Imagine you have a squad taking cover in a stand of light trees. In CM1, they automatically gain cover. In CMx2, any advantages in cover are relative to the AI's ability to get each individual to take a proper position (ie, one that puts tree between you and incoming fire). CMx2's method should be more realistic, but only if the AI is good at its job. If not, you may see little to no benefit from your 'covered' terrain.

Thanks, that was my point exactly. I can't understand how BF thought they could make an AI good enough to drop this abstraction almost completely. The AI is not bad compared to other games, but, you know, it's an AI... there are stragglers, soldiers that take detours, that lay down in the middle of the road with their backs turned against the enemy etc.; everything you would expect from such an AI. And when everything else in the game is modelled realistically(as is the case with CM:SF) then of course that sort of behaviour will get them killed, just as it would get a bunch of half-blind retards killed if one were to send them out in a fight in real-life.

I'm pretty disappointed with this game. The more I play it the more I feel that realism has been sacrificed for 1:1 representation. I haven't played a scenario so far without having several men killed simply because they are AI-soldiers. If I remember correctly the manual says that this sort of thing should not have an impact on the overall realism of the game, but I can't see how it doesn't.

[Guest Guest]

deleted per user request

[PFMM]

Ah, the old argument that if one can't see it, it's good, but if one can see it, and one doesn't like what one sees, it's bad.

Bunk, twaddle and **** for brains.

[Bruce70]

When the 1:1 rep was first anounced there were several people who were concerned about this (link)and rightly so IMO.

My personal opinion has always been that the 1:1 rep done right would be great, but done poorly would be worse than the abstraction.

I've only played the demo so far, but from what I've seen I think it's pretty good. The thing that makes it work is the "hotspots". As far as I can tell a map designer doesn't have to designate these, so I'm assuming it's done automatically. It may be that on some maps the hotspots are not well defined, or it may be that with no hotspots (e.g. in open terrain) the AI is fairly poor, but I haven't played enough to form an opinion.

[thewood]

This was the exact arguement I used to use for not having a 1:1 representation for squads. CC, GIC, and TOW all did it and had the same problems. The AI didn't know how to get an entire squad in cover. There was always a lot of fitzing around to get everyone in cover, if you ever could. CM didn't have that issue. I am hoping that that is not the case with SF.

The fitzing around thing may be tolerable in WEGO, if you know how the abstractions work, but in RT, it can be frustrating to have to spend minutes moving squads around, reissuing orders, and trying to make it work. Meanwhile all hell is breaking loose because the T72 you spent 5 minyutes getting into hull down was is getting toasted because of the same issue.

[Northman]

Originally posted by AdamL:

I think the answer might be simply that it's not trying to simulate everything, that there is still quite a bit of abstraction going on. If you actually test it out you'll see that the infantry don't die as quickly as you'd expect if they really had as little cover as is represented. Partly this is also because I think the shots themselves are abstracted. I think they are calculated first (with an abstract model), then represented second, like the way AP rounds were in CMx1. So you see a lot of shots going astray, but maybe that is the visual representation of "misses" when the abstract cover is factored in. Similarly, when rounds hit trees, they don't just happen to hit but the system has already determined that and simply gives you some eye candy to go along.

I could be wrong on all this, I'm just guessing.

I've been thinking exactly what you describe here, but unlike you I haven't been able to fit this reasoning with what I've seen in-game so far. I hope you are right, though. I think it was one of the devs who actually said the game was WYSIWYG with a small abstraction modifier.

[panzermartin]

I think Company of Heroes has a very flexible cover system for infantry and its a piece of cake to use. You can easily put your men in a line behind a wall or pack them behind an abandonded car. Things I really miss in CMSF.

[Northman]

Originally posted by PFMM:

Ah, the old argument that if one can't see it, it's good, but if one can see it, and one doesn't like what one sees, it's bad.

Bunk, twaddle and **** for brains.

Not quite. If one can't see it but the results are good, then it's good. Similarly, if one can see it and it looks bad, but the results are good, it's good. It's the results that matters, not how things look. I don't mind having my grunts spinning around in the middle of the street as long as they don't get themselves killed just because they do so. I think this philosophy is mentioned in the manual, at the beginning if I remember correctly.

[H.W. Guderian]

Originally posted by Northman:

</font><blockquote>quote:</font><hr />Originally posted by CRourke:

I think I see northman's point. Imagine you have a squad taking cover in a stand of light trees. In CM1, they automatically gain cover. In CMx2, any advantages in cover are relative to the AI's ability to get each individual to take a proper position (ie, one that puts tree between you and incoming fire). CMx2's method should be more realistic, but only if the AI is good at its job. If not, you may see little to no benefit from your 'covered' terrain.

Thanks, that was my point exactly. I can't understand how BF thought they could make an AI good enough to drop this abstraction almost completely. The AI is not bad compared to other games, but, you know, it's an AI... there are stragglers, soldiers that take detours, that lay down in the middle of the road with their backs turned against the enemy etc.; everything you would expect from such an AI. And when everything else in the game is modelled realistically(as is the case with CM:SF) then of course that sort of behaviour will get them killed, just as it would get a bunch of half-blind retards killed if one were to send them out in a fight in real-life.

I'm pretty disappointed with this game. The more I play it the more I feel that realism has been sacrificed for 1:1 representation. I haven't played a scenario so far without having several men killed simply because they are AI-soldiers. If I remember correctly the manual says that this sort of thing should not have an impact on the overall realism of the game, but I can't see how it doesn't. </font>

[Guest Guest]

deleted per user request

[Battlefront.com]

Oh, there is no question that the LOS/LOF system in CMx2 is vastly superior to the way it was in CMx1. The overly abstracted sytem in CMx1 is in no way more realistic, better, or whatever compared to CMx2. That's not because CMx2 is a perfect 1:1 system, rather because CMx1 was so heavily abstracted. It was fantastic for its day, and I still think it worked out really well, but I also feel that CMx2 is a vast improvement.

I'm not saying this just because it's our game. I'm saying that because I honestly feel that. I've also been playing this thing a LOT longer than any of you, so I also suggest that I've got more than a first impression opinion about this

Guys straggling, taken wrong paths, etc. is actually far more realistic than it is not. Sure, occasionally guys do some things that they probably shouldn't (there are a few cases we are working on), but overall their behavior is pretty darn'd spot on. Some of you might not see that yet, but I think you will if you really give it a chance.

As for LOS/LOF, there are some abstractions that are necessary to make the game work. LOS is somewhat abstracted based on the underlying 8x8m tile structure. In CMx1 LOS was not abstracted, but the terrain was extremely simplistic compared to CMx2. So what you lose in LOS fidelity in CMx2 you gain in terrain resolution. The tradeoff is absolutely a net benefit to the quality of the simulation. Just think about it guys, in CMx1 could you:

Have a house, trees, bushes, walls, road, variable ground type, and another house just for the heck of it all within a 20x20m space?

Nope. All you could have is a single house with grass, or grass with trees, or grass with a wall and trees, or road with trees and grass.

So... think about that a bit more. It's a huge improvement in realism, and a massive improvement in map design flexibility. It also happens to look a lot nicer too

This terrain resolution also makes MOUT warfare possible. MOUT in CMx1 was very poorly modeled because the terrain resolution was simply too blunt for the highly compact terrain found in urban areas. Heck, even CMx2's 8x8m terrain is "limiting" in some areas, even though it is far more flexible than CMx1. So I would say that not only is CMx2's new terrain better to have from a simulation standpoint, without it we couldn't do contemporary warfare.

The tradeoff for this is that we can't have exact LOS from a random sub-meter to another sub-meter like we could in CMx1. But since the terrain and units were so heavily abstracted I honestly don't see how it can be said that having an exactly LOS in CMx1 was worth much. In fact, we had to force the exact LOS to be more generalized because units aren't on heads of pins. So I guess it is fair to say that the exact LOS in CMx1 wasn't very exact either as far as the game system was concerned

Now, does this mean the CMx2 system is perfect? Heck no! It isn't. It's just a lot better than CMx1.

Also, there are probably still some circumstances where there is a LOS bug or two in the game. The massively complex terrain combination possibilities has meant a lengthy debugging process. We're only Human so it is entirely possible some got missed or were unfixed. We'll stay on top of those and get them fixed as we ID them and can code around them.

Steve