Strangling/suffocating

[Ishtvan]

@Maximus: Supposing this "sweet spot" exists, what are the odds that you're going to hit it every time on guards that are always making slight movements like shifting their weight back and forth, looking around (which moves their head and neck around constantly) let alone walking on patrol? It sounds like you're trying to stab thru both the jugular vein and the windpipe with one blow, so any slight turn of the head will change the angle at which you'd have to stab to pull this off.

 

Like I said a few posts above, a guy who killed a lot of people in prison was interviewed, and asked what the preferred method was. He said the "safest" way to stab someone was thru the armpit and into the lung. The reason he gave was: It's a large target with a predictable result that people go into shock when their lung is stabbed and collapsed. (The catch is you have to grab their arm with your other hand and raise it up to expose the target). The neck on the other hand was ruled out because it requires too much precision in terms of stab angle and location, so any small movement by the enemy could throw it off and turn a fatal blow into a non-lethal one that lets the enemy cry out.

 

Even if you stabbed the enemy in the neck and destroyed their ability to cry out, I would think someone with a hole in their windpipe and jugular would still have some time to struggle before they die. (I don't really know though, maybe you go into shock, I'm just basing this on the fact that people with their throat slit don't die right away). An armored guard stomping their boots and flailing their arms around should make a fair amount of noise, even if you hold onto them so they don't fall to the ground.

 

Realism questions aside, we want killing to be rather loud for gameplay reasons. That is the negative consequence of killing in our game. If you want to make an assassin mod where you can kill everyone silently, it will be very easy to lower the "death cry" sound volumes, or change the AI scripting so that they don't make the death cry if you kill them a certain way. Killing isn't the focus of our mod tho, so I don't think we'll be doing that ourselves.

[Maximius]

No, you are quite right, as I remember it its a very pinpoint attack, you have to slip the knives between some muscle groups or something. In fact, Steven Seagal uses the move in one of his horrible movies Executive Decision. In the beginning, Seagal is sneaking up on some Eastern Euro. mobsters hideout. He sneaks up behind a guard and punches his knife into his neck. This is the move I described. Im no expert, the only thing I would attack with a knife would be a nice plump chicken leg. I just wanted to throw a bit of information into the mix not argue contra something you were saying.

 

I prefer the no killing rule myself, with the occasionaly exception for some dirtbag who needs it. I think killing should be loud as hell too, MoF I think any guard hearing any really loud noise should kill the mission. I think BJing a guard should start a clock counting down to where his buddies say OK whats going on with Benny and start looking, seriously and unrelentingly, after say fifteen minutes.

[Pyrian]

Even chopping someone's head clean off may leave their corpse twitching or even thrashing on the ground.

 

It's possible to muffle a sap - although that also makes it less effective(!). Still, BJ'ing someone with a loud smack and having them crash down to the ground in full armor, their sword ringing against the flagstones, with another guard two feet away who didn't hear a thing was always kind of silly.

 

I'd just as soon the game rewarded ghosting as much as possible and made the "clear the level" style of play less plausible and certainly less of the path of least resistance.

[obscurus]

1) Can SC also do dynamic shadows (which are a big drain)? 1) Does it do sound propagation? 3) How many things can the AI potentially do?

1) Yes, 2) Yes, and 3) AI is pretty smart most of the time, so lots. And as far as I know dynamic shadows are almost entirely handled by the GPU anyway, so they should leave the CPU with plenty of clock cycles for the other stuff (that is the whole point of having a separate graphics chip). SC has a lot of scripted bits so I don't know how much of that is scripting and how much is AI, but the AI in a lot of games is getting better and better these days, and most engines are using dynamic shadows etc now (D3 seems to do this better than some engines, but game engines aren't that easy to tell apart any more)....

 

I know you probably need a team of full time programmers getting payed big bucks to expect fully optimised AI, so in the case of DarkMod, I can imagine you don't have the resources to go to the level of detail some of us would like

 

But my point about multiple threading was more that the trend is for separate chips handling separate streams - this is implemented in the Xbox 360, and physics processor units are about six months away, Creative has just announced a new audio chip that handles 3d sound propagation and occlusion, and has considerably more processing power than a 3.5 GHz P4 (!) And Id games is likely to add support for these to D3, as they become available, through patches... So when graphics, physics and sound are taken care of by separate processors, the CPU will have a lot more cycles free for gobsmackingly smart AI (or so is the industry plan).

[FishFace]

I'd rather make the mission real difficult than failing - failing forces you to load. Making it difficult means it's possible to complete - you still have the option of living with the consequences.

[obscurus]

I prefer the no killing rule myself, with the occasionaly exception for some dirtbag who needs it. I think killing should be loud as hell too, MoF I think any guard hearing any really loud noise should kill the mission. I think BJing a guard should start a clock counting down to where his buddies say OK whats going on with Benny and start looking, seriously and unrelentingly, after say fifteen minutes.

I think it would be hilarious if, on stabbing a guard, instead of dying straight away, he runs through the castle bleeding and screaming "help I been stabbed, aaaaaarrrrrrgh!!!, get the taffer!" and then drops dead in a really obvious, brightly lit spot. I think guards should definately notice if their buddy who was just patrolling with them a moment ago has suddenly dissapeared, and should start looking for them...

 

I also hope that blood in Darkmod looks a bit more gory and a lot less like a lump of wax than it did in TDS :lol:

[Maximius]

No kidding obscurus, TDS's blood splatters looked like a graphic from the first Resident Evil.

 

I had an idea for T2 a while back where the AIs would have differing levels of "intelligence" to respond to events. For example, the Thief fires a noise arrow to distract a group of guards. Now if the guards are all privates, raw recruits or whatever, they all go running yelling and carrying on. But lets say there is a sergeant in the crowd. The sergeant AI doesnt know whats up, but he knows something isnt right. Hes a little bit brighter than the goon squad. So rather than running off, he starts to sniff around, not red alert only yellow, but not green either. Maybe if he is still sniffing when the crew gets back, they may go to yellow for a short time too.

 

But now lets say the Captain is on duty tonight. She hears the noise arrow and immediately remembers her Captains Training manuel where it warns of arrows that pop or click used to distract guards. She goes to high yellow/red (if those distinctions exist) and furthermore sends out a general alert. The whole place starts shaking up but rather than following the arrow they start tearing apart their general area. Heck, firing a noise arrow with a Captain around on Ghost could kill the game.

 

There could be a priority list for different events.

 

 

Noisearrows: Grunts follow it Sarge= yellow Capn= red/rallies troops

 

Explosion : All go Red Rally---------------------------------------------->

 

Buddy missing: Grunts= yello 2 red Sarge= yellow Capn=doesnt notice

tilltroops do

 

Too many torches G=grumble S= yellow C= red

out: relight----------------------------------------->

 

Sounds of fighting: All go Red Rally----------------------------------------------->

 

Gigantic moss carpet

suddenly appears: Yellow yellow red

 

you get the idea. I dont think it would require recoding or anything, those would be those particular AIs scripted characteristics so you would only change them once.

Edited May 19, 2005 by Maximius

[obscurus]

That sounds like a realy good idea Maximius.

 

I have been making maps for unreal/UT (Thievery) for a while now, and while I am not sure how it works in in Doom 3, in Unreal x you can tweak an individual AI's properties quite a bit, and I have always made levels wherre there are no two enemy AIs that are alike - that way you can't assume that just coz one guard reacted one way, so will all of them. They are all individuals with different personalities, strengths and weaknesses, but the player doesn't know that from the start, they have to find out the hard way.

 

This is better than having different classes of guard, because rather than recognising what class a guard belongs to and knowing how it will behave, you will have to treat all individuals with caution.

 

this is not to say that having classes based on rank etc is bad, not at all, i would just take it even further than classes...

When I am designing a map, I will actually go so far as to give individual names and short personality descriptions about each AI I place in a level. I am possibly a bit mad, however

[Domarius]

For me personally, more realism = more fun...

I remember when I used to say this too

Edited May 19, 2005 by Domarius

[Maximius]

Sounds like my kind of madness. Im new to editing, I got as far as the first floor of a family home with T3ED but since I discovered that GTK Rad is available for free plus VTMs Ive given up on TDS. Ahh dreams...

 

Little touches like the ones you describe can make a big difference. I had a similar one on TTLG about recurring characters. e.g. I want to make a vampire character, one that the Thief encounters in different FMs, in different contexts. First FM, Introduction, two Masters of the Dark meet and briefly clash. A few FMs later, they cross in the night, a cameo appearence to keep the relationship err, warm. Then later a real humdinger, a mission where the two are allied, maybe against a Hammerite jihad against the undead and anything else creeping around. Doesnt have to be vampire, maybe a Drept type or another Thief, some big shot wizards or royalty.

 

The characters could be made available for the community. Also, the characters history could be posted on a website for updating, the local Hammerite Undead patrols incident logs on the vampire or something. For an FM to be included in the official history, it would have to adhere to some guidelines, cant kill off main characters, cant change the storyline too radically w/o informing someone.

 

Developing a rich storyline and world history opens up new ways of creating realism that dont necessarily require a ton of re-coding or scripting. Just seeing that vampire in the middle of an other wise unrelated FM will build a stronger sense of immersion, cause you saw him doing something evil a month ago in the same area.

[obscurus]

I remember when I used to say this too

Like I said I am perhaps a bit mad. i think about it like this: I like realism because one, I have freedom approximating what i would expect to have if i was really there, and two, things behave similarly to the way I expect they would. That to me is fun. you might find it not fun, but I think a lot of badly designed games have convinced developers that realism must be thrown out the window to make a game playable or fun and sell lots of copies, which I dispute quite vigorously...

 

I will use an analogy. I like analogies, even slightly dubious ones Simple, semi-realistic games like Doom are like driving a car with automatic transmission. As simple as a car can get. More realistic games like Rainbow Six type combat simulators are like driving a car with manual transmision - a bit more complicated but with finer control. Flight simulators or my desired hypothetical reality simulator would be like flying a plane - considerably more complicated, sometimes frustrating, even occasionally boring maybe, but in the long run vastly more rewarding than driving an automatic car. Ask any pilot what they would rather drive - a fighter jet, Apache Gunship or a family sedan with cruise control...

 

Since I don't often find reality to be in the slightest bit boring, why would a realistic game suddenly become boring?

I like the original Thief games more than any others I have played, but I am a progressive, I am not content to recreate Thief with better graphics, I want to enhance everything. Thief played the way it did - graphics, sound AI, physics - because of technological limits at the time, and it's peculiar setting. Lets make something slightly different, even perhaps, better?

 

 

Each to their own I guess

[Domarius]

...I have freedom approximating what i would expect to have if i was really there, and two, things behave similarly to the way I expect they would. That to me is fun. ...

 

That is true for me and a lot of other people, but what you can't yet appreciate is that there is MORE to gameplay than just adding more realisim.

 

All I'm saying is, just plonking more realism into a game doesn't naturally equal more fun.

 

Each feature has to be considered in the context of the game and how it plays.

 

For example, adding sillhouettes detection would make hiding even harder, and not offer any of the other details of real life that would help you to hide, like maybe throwing a cloth over yourself in a dark corner - little things like that that woudl only work once in a specific situation, you can't possibly program every possibilty into a game.

 

You have to simplify things, both for making the game play smoothly, and also so you don't fall into the trap of NEVER releasing yoru game by trying to make a full on virtual world simulator.

Edited May 19, 2005 by Domarius

[Ishtvan]

The best way to reduce your silhouette is going prone, but that opens up a whole 'nother can of worms.

 

[Edit: Or just do a little more thinking and don't step in front of brightly lit stuff. That's not much of a drawback for the believability that it adds, IMO.]

Edited May 19, 2005 by Ishtvan

[sparhawk]

My question were more rethoric one, because this is only a small part of what a gamecode needs to perform.

 

1) Yes, 2) Yes, and 3) AI is pretty smart most of the time, so lots.  And as far as I know dynamic shadows are almost entirely handled by the GPU anyway, so they should leave the CPU with plenty of clock cycles for the other stuff (that is the whole point of having a separate graphics chip).

 

That's wrong. Dynamic lights are rather expensive, because they need a renderpass for each light. IMO the GPU is involved but it also takes a lot of time to set it up and execute it. That's one reason why you should limit your number of lights per surface to two in D3 for a normal map. Of course you can use more, but you should take care about the general complexity about such a part of the map to avoid FPS dropping. Calculating a shadow essentialy requires you to make a renderpass from a given light in order to determine where the shadow obscures the geometry.

 

SC has a lot of scripted bits so I don't know how much of that is scripting and how much is AI, but the AI in a lot of games is getting better and better these days,

 

Of ocurse they are getting better. Machines are getting faster and I guess the next step is to have multiple CPUs. GFX cards already start to do this. You can by dual cards already.

 

and most engines are using dynamic shadows etc now (D3 seems to do this better than some engines, but game engines aren't that easy to tell apart any more).... 

 

Most engines? Which ones? I'm not aware of much engines which are doing real dynamic lights. I hope you are not thinking of HL2 because it can not do this. I can show you a lot of examples which show you that the shadwos in HL2 are either precalculated or wrong.

 

I know you probably need a team of full time programmers getting payed big bucks to expect fully optimised AI,

 

I wouldn't say that, but of course we have only a limited number of coders, so we have to select what we can do at any given time in order to finish something in a reasonable time. One advantage of community projects is that they don't need a budget and therfefore can implement stuff which would be commercialy stupid for a professional project.

 

So when graphics, physics and sound are taken care of by separate processors,  the CPU will have a lot more cycles free for gobsmackingly smart AI (or so is the industry plan).

 

Of course such support can only be supported in general if it is widely spread.

[obscurus]

Damn you Sparhawk, stop shattering my illusions, they are all I have! :lol:

 

Seriously though, I know that dynamic lights are expensive, but I am still pretty sure it is the GPU that bears the burden of that expense (ie, the GPU does the renderpasses required, not the CPU). But maybe I am horribly mistaken about how this stuff works

 

In terms of engines with dynamic shadow support, I suppose I am thinking of engines like Thief Deadly Shadows, Splinter Cell (basiclally modified Unreal engines), Doom 3 obviously, The Cry Engine supports them nicely (although Far Cry doesn't really show them off, being largely outdoors and all), Codecreatures, the upcoming Unreal 3 engine will blow your socks off, and there's3D gamestudio A6, the Torc engine (available in the near future) and I think Deep Shadows (not entirely sure about this one)... OK maybe that isn't most engines

 

And I simply must have sillhouettes! It is beacause they make hiding harder, I want hiding to be harder..

 

And crawling prone too, yessss I musst have it...

[sparhawk]

Prone = no

 

Silhouetting = Maybe. This depends mainly if we can get enough FPS to make this work. The problem here is that ou have to render each silhouette as a seperate pass for each AI that can see you. This is rather expensive to do, so it might not work.

[obscurus]

Oh well. I guess I'll have to code some stuff myself I s'pose if sillhouettes are too hard, that's it. But at least coding movement modes should be fairly simple for those of us who want't it to do. Like I said, if you add it, you can just choose not to use it and unbind the key for your missions if you don't think you can make a mission that would be balanced with it in.

[oDDity]

Prone = no

 

Silhouetting = Maybe. This depends mainly if we can get enough FPS to make this work. The problem here is that ou have to render each silhouette as a seperate pass for each AI that can see you. This is rather expensive to do, so it might not work.

I don't think that's necessary. Only the guards looking in the players direction need to be able to see his silhouette. THat isn't going to amount to more that one or to guards in the vast majority of cases.

In the case where you're being chased by a bunch of guards only the guard closest to you needs to see you silhouette.

[sparhawk]

You don't know how many guards are looking in your direction at any given time. This means if your are standing in a doorway to the sleeping room of the guards and they are awakend, suddenly they all stare at you and you would have to do the calculation for all of them.

 

Of course there are some optimizations and AI not looking in your direction would be excluded. It also doesn't make sense to do this for an AI which is chasing you and can already see you, because it wouldn't matter. Same for AI that is already fighting you. But there can still be a sufficient number of AI requiring the test and this is rather expensive.

 

One way to optimize it, could be to calculate an angle from which this can be seen. This way you only have to do it once and then check which AI is within that angle and looking in your direction. Might be possible to ommit the calculation for all the other AIs in such a case.

[FishFace]

In the case of many AI waking up, then you could lump all of them together, since they're all looking from similar angles, only one needs to see you and he'd shout to the others. At this point, you may get away with dropping the sillhouetting in anything but the deepest, darkest shadow because you'll be moving quickly (i.e. very easy to see) and they'll know exactly what they're looking for.

 

Also, you could get away with an extremely low-resolution pass (128x128 or so) for sillhouetting, I'd think.

[Pyrian]

Some examples of when realism sucks:

 

1) Long travel times. I love to go hiking, but doing it in a computer game which can't possibly render nature in all its glory just doesn't do it for me after a few minutes. Realism would involve hours or even days.

 

2) Death. Realism would mean you install the game, play from the start for a while, and then when you die, you're done. No re-loading a past save-game, no more playing as Garrett (he's dead), no more playing that game at all.

 

3) Effectively impossible problems. This mission can't actually BE completed, for whatever reason.

 

4) Too easy problems AKA drudge work. Haggling over every item in a bag full of loot; carving elemental crystals into arrowheads; eating, sleeping, bathroom, and so forth. Heck, the thief universe is relatively forgiving in this aspect.

[Ishtvan]

That's a good list of when realism sucks. I would add in:

 

Trying to represent things that are done unconsciously or through tactile feedback with some awkward control system.

 

For example: Walking. You shoudln't have to press "flex left ankle tendon. Bend left knee. Pick up left foot. Extend left leg forward. Put left leg down. Bend left knee..." Instead, we push forward and walk. Because walking is learned at a young age and done pretty much unconsciously in RL, it is annoying to have detailed and awkward controls for it in a game.

 

Another example: Parallax errors in a gun or bow. When you first learn to shoot a gun, you have to think about keeping the sights aligned. After a certain point though, it just comes naturally. Unconscious jitter aside, you keep the sights aligned without thinking about it. To actually do this in a game, you would need at least 3 controls for the pitch, yaw and roll of the gun, and it would take forever to just align the sights, let alone use them to aim. This would be ridiculous.

 

Similarly, when just learning how to fire a bow you might have a problem with drawing the string straight back, so when released, you torque the arrow and it shoots off to one side or the other. Again, once you learn it, you don't really have to think about it, but this would require very awkward controls ingame to control how straight you draw the string back.

 

Also, grabbing things. We don't think about this at all IRL, we just put our hand out, and when our nerves tell us that our palm/fingers are touching something, we close our hand and grab it. Completely tactile and unconscious. If you tried to put in controls for that in a game, you'd need a representation for tactile feedback in all the various parts of your hand, and some control for how open/closed your hand is, and you could grab things too hard or too light and make them slip out of your grasp, oh no! Again, that would be ridiculous with a keyboard/mouse interface.

[Springheel]

[Edit: Or just do a little more thinking and don't step in front of brightly lit stuff. That's not much of a drawback for the believability that it adds, IMO.]

 

It's a little more difficult than that...in RL, you can do a quick over-the-shoulder glance to see if there is anything bright behind you. In the game you would actually have to stop and turn around. I'm generally in favour of adding silouehette detection, but I think we would have to really simplify the concept, just like hiding in shadows has been simplified.

[obscurus]

Some examples of when realism sucks:

 

1) Long travel times. I love to go hiking, but doing it in a computer game which can't possibly render nature in all its glory just doesn't do it for me after a few minutes. Realism would involve hours or even days.

 

2) Death. Realism would mean you install the game, play from the start for a while, and then when you die, you're done. No re-loading a past save-game, no more playing as Garrett (he's dead), no more playing that game at all.

 

3) Effectively impossible problems. This mission can't actually BE completed, for whatever reason.

 

4) Too easy problems AKA drudge work. Haggling over every item in a bag full of loot; carving elemental crystals into arrowheads; eating, sleeping, bathroom, and so forth. Heck, the thief universe is relatively forgiving in this aspect.

I really can't disagree with you on any of these points...

 

That is one of the reasons I am not a big fan of overly large outdoor levels, the computer just can't handle the detail required to make it interesting.

 

There is a certainly a point where if you make the game too realistic, it will begin to suck, I guess I just think that point is a bit further down the realism road than everyone else

 

I think though that there should be limits on how frequently you save games that gets smaller on the higher difficulty settings, I actually do like to see how far I can get through a level without dying or saving....

[sparhawk]

For example: Walking.  You shoudln't have to press "flex left ankle tendon.  Bend left knee.  Pick up left foot.  Extend left leg forward.  Put left leg down.  Bend left knee..." Instead, we push forward and walk.

 

You are real pussies. If we would implement this, it would seperate the man from the boys. It would add a big deal of immersion to the game if we could do just this.