Josh Sawyer talks about stealth mechanics

[Jobby]

Ah i see, well we can deal with circles, I'm sure they can be trimmed to solid "tilesets" or whatever devs call walls these days lol.

[Adhin]

Yeah a moving 'zone' for enemies when they walk would make some sense, not to far of course but... still. Would only really account for visual though at that point. To say someone on alert is only focusing on stuff 'ahead' of them with both their sight 'and' hearing is practically an impossibility. We can't really focus are ears to ignore stuff making noise behind us. And they are already simplifying it all to a single sneak skill check instead of movement + the act of physically hiding against spot and listen like in DnD.

 

I mean the circle is already two staged too. If your circle gets inside there outer zone of they're detection circle they become alerted to something. They heard something or think they saw some kind of movement. This causes them to move toward that point and if you don't get away in time and your circle enters they're inner circle zone the rouse is up and your detected.

 

Basically I'm saying they would have to split up and complicate the sneak mechanics before they have a real good reason to mess with this circle idea they got going. And for an IE-style game? This is a hell of a stealth system. Quite frankly I prefer it to even the base 3E style as it's more of a 'game' then just some non-stop dice rolls till the opposition gets lucky and rolls a 20. I know 20 isn't an instant win for stealth mechanics (least not in NWN) but untill you completely blew past their max check it kinda was more a game on 'when' you'd get spotted, not 'if' you'd get spotted.

 

Anyway outside of splitting up the skills for visual and auditory parts of detection 'and' sneaking I think the way they're handling the stationary circles will provide the best game play results. Anything else is adding a complexity to a base system that lacks the mechanics to support it. A Visual cone is pointless on the grounds there is no skill that governs your ability to 'spot' someone hiding, or the verse not being able to 'hear' someone. It's all 1 check. So that circles both your visual and auditory detection zone.

[Nonek]

Wonder if we'll get sneaking animations, or whether it's too late to implement. Personally I found the animations they added to NWN2 in some patch or other were quite pleasing, very atmospheric.

 

Edit: On further reflection I suppose one could merge a pose of readiness and wariness with a sneaking animations, if these are implemented. Shield raised, sword readied along with a slow purposeful gait etcetera.

Edited January 19, 2014 by Nonek

[Mr. Magniloquent]

I can see the utility of a circle that bulges out

 

 

Given what Sawyer said about how difficult line of sight would be to implement I doubt that will be present, Don't get me wrong it would be awesome but it just seems unlikely, the circles are a decent compromise, don't like the idea of scaling to level tho? (don't know if that has been misinterpreted on my part)

 

Instead of circle, use half-cirlces. That leaves a blind spot for vidual detection.

 

Audio detection is another matter.

 

 

I can see the utility of something that bulges out, perhaps like this:

perception.png

Red is noticed you, yellow investigating.  It's circular near the human since that's affected more by hearing, and vision extends conically from the entity.  You could even have a formula which made the yellow decrease faster with sneak ability than red.  

 

For ninja death attacks, you could have characters dash in with sneak attack ready before their target raises the alarm.  If they succeed and kill the mob, the dying character makes a ping, which hits the other detection shapes and interacts normally.  That way it isn't DX:HR levels of ridiculousness where you can muffle guys from their buddies five feet away, but it still allows the option in the gameplay.

 

The problem is that you would then have to make / test / debug several different types of perception cones.  While the first models human perception, this might model skulldr perception:

skulldrperception.png

 

Eventually that becomes a lot to implement, test, debug, etc.  They could do archetypes, human, beast, dragon, undead, etc.  But that still turns into a substantial amount of work for a less important subsystem.  That would be an amazing job for a modder though (although this is going to be pretty hard to mod from what I understand).  Or, it's work for obsidian to do in a game based on this technology.  The great thing about Project Eternity is that it's just the first, and other games can build substantially on its capabilities.

 

I was going to make this exact post, drawings included. I am in total agreement with what you've contributed here.

 

I don't think it would be too terribly difficult to implement over what (as we understand) is currently being created. Having a few templates for certain archetypes as you've said would make this sort of variance straight-forward.

[okey231]

will it be possible to create a fighter that has a very high deflection.... don't really like the idea of a tin can fighter who has to wear the heaviest armors and take damages with shield and raw toughness... wat I mean is like a skilled fighter who can make good evasion steps while fighting??

[TrashMan]

 

 

Instead of circle, use half-cirlces. That leaves a blind spot for visual detection.

 

Audio detection is another matter.

 

And in which direction would the half-circles face? That was the difficulty he was mentioning in regards to line of sight as far as i could tell. 

 

 

The front. DUH!.. unless you have eyes on the back fo your head

 

The circles being a sound+visual combo is problematic, as it makes sneaking from behind practicly impossible.

Also, sound and visual detection are vastly different.

I can easly see a man standing 500-1000 meters from me, but I sure as hell can't hear his footsteps at that range.

 

I confess, a perfect solution escapes me. Would you even need a visual cone, given that at the scale and given normal human sight, a person should see from one corner of the screen to the opposite? So a cone would technicly have no meaningfull "range".

 

Snekaing would be about avoiding the visual cone and being quiet. Kinda like commandos. So it's more about beign quiet than it's about being a master of camouflage.

The visual overlay for it should be toggable, since there is no point in having it on all the time anyway.

 

Edited January 20, 2014 by TrashMan

[PrimeJunta]

I just hope they'll make the sneak animation as cool as Arcanum's. Prowling FTW!

[Sabotin]

At some point it was said that rogues would be able to set up their own "sneak attacks". Maybe one of these will be some kind of distraction, that temporarily turns off a perception circle? Otherwise I too have trouble figuring out how you'd approach someone for a stab.

[PrimeJunta]

At some point it was said that rogues would be able to set up their own "sneak attacks". Maybe one of these will be some kind of distraction, that temporarily turns off a perception circle? Otherwise I too have trouble figuring out how you'd approach someone for a stab.

Engage them with your fighter to get them to stay put, outflank with your rogue, stab stab stab, mop up blood.

[Adhin]

Your attack range may actually exceed your detection if you have sneak up high enough. Though the 'Sneak Attack' is a 3E rename from backstab. For PoE it's a 'flanking' which just requires a friendly attacking the target 'from the other side'. Which I think means even if the enemy is attacking the rogue, just having an ally on the flank will give you the bonus. Hoping anyway.

 

-edit-

Oh just forgot, also Rogue's get an ability that turn them 'truly invisible' so... if there is a difference between a 'Sneak Attack' and a Flanking bonuses then that's probably how they can set it up, in the middle of combat or otherwise.

Edited January 20, 2014 by Adhin

[Lephys]

I understand that there are many a technical reason for not going all out on this, and that I can't really judge whether or not they SHOULD make things more complex, etc, in practice, based on the resources at hand and the difficulty of doing so. I'm just saying that, ideally, there would be a lot to be gained by, at the very least having audibility and visibility separately represented.

 

Honestly, I'd even be cool with no facings. With an enemy "seeing" me with his back turned (still just use circles, instead of cones). But, it would still be objectively more interesting stealth gameplay if they could figure out a way to have, say, a sentry with a tiny "vision" circle in a dark, cobblestoney courtyard, but still have a large audibility circle because your footsteps on the stone are louder/echo-y.

 

In other words, if it's just "sound-o-vision," all mashed into one thing, then we miss out on all the factors that just affect either sound or visibility.

 

The game will hardly be terrible if we just get generic "stealth" circles. I'm happy with that, and trust that they'll do it as well as they can, which is all we can ask, really.

 

You wouldn't even need to do anything extra for scent and/or Skuldr's soul-sense. You just go "Oh, it's dark? But this thing's smelling you, so it still gets a huge circle anyway, as opposed to a tiny vision circle because of the darkness." Even though the thing's not "seeing" you, it would still represent that none of the typical factors apply to its particular detection radius. I mean, unless you implement wind direction and stuff, which starts just getting unnecessarily complicated for a game that isn't all about stealth.

 

But, with separate vision/hearing representation, everything would have three circles:

 

1 for your ability to detect via hearing...

1 for your ability to detect via vision...

And 1 to represent your ability-to-be-detected area. (We'll call it your Presence circle.)

 

Everything would always have the same-sized detection (first two) circles, depending on their Perception (and factors of their creature-type, determining the difference between their hearing and vision circles -- for the things that hear better than they see, or vice versa). You would only adjust the THIRD circle for environmental/skill-related factors. It's really dark? Shrink down your Presence circle. Is it really bright outside? Your Presence Circle gets enlarged.

 

Then, everything works as currently planned. If your presence circle overlaps either of their detection circles, they investigate and/or confront you. It's as simple as that. If your Stealth skill sucks, and/or the environment is noisy and/or well-lit, your Presence circle is larger, and you must stay that much farther away from the sentries to remain undetected. If your skill is amazing, and/or your environment is quiet and/or dark, your Presence circle is smaller, and you can get that much closer to patrolling things.

 

You don't ever even need to see any Presence circles other than your own (your party's), because you can either detect a foe or you can't. If you can't detect it, you're not going to know how much closer you need to get to be able to do so, and if you CAN detect it, then you already know that, however large its Presence circle is, it's now overlapping your detection circle.

 

Sure, it's 3 things for every enemy, but it doesn't use any facings or anything. And it adds 7,000,000% awesomeness to the specifics of stealth-pertinent situations. (I counted, o_o)

 

Of course, without facings, you could never actually sneak up to anyone/anything. But *shrug*, again, I can live with that.

 

And if we can't even do that, and we just stick with the current design, I can live with that, too. But, my hopes are something along the lines of the above bit.

[J.E. Sawyer]

Our general approach to the stealth system is to use the current implementation unless it proves to be too simple in the long run.  Other than time, there's nothing that prevents us from adding on elements like cones of sight and other tidbits, but we are, after all, replacing a system where failure and success came down to a die roll.  It is extremely easy to make a stealth system that is overwhelming in complexity, so we're starting with something that is basic but feels like a solid starting point.

 

Your attack range may actually exceed your detection if you have sneak up high enough. Though the 'Sneak Attack' is a 3E rename from backstab. For PoE it's a 'flanking' which just requires a friendly attacking the target 'from the other side'. Which I think means even if the enemy is attacking the rogue, just having an ally on the flank will give you the bonus. Hoping anyway.

 

Rogues in PoE qualify for Sneak Attacks in a large number of circumstances now, almost all of them being conditions on the target (e.g. Flanked, Prone, etc.).  In the case of Flanked, the rogue doesn't even have to be one of the characters doing the flanking.  Many of the other conditions are ones that rogues gain the ability to inflict.  At higher levels, they gain additional bonuses if their targets are suffering from two or more of the qualifying conditions.  So if there's a Flanked target that gets knocked Prone, the rogue gains even higher damage bonuses.

[Lephys]

Our general approach to the stealth system is to use the current implementation unless it proves to be too simple in the long run.  Other than time, there's nothing that prevents us from adding on elements like cones of sight and other tidbits, but we are, after all, replacing a system where failure and success came down to a die roll.  It is extremely easy to make a stealth system that is overwhelming in complexity, so we're starting with something that is basic but feels like a solid starting point.

I just want to re-iterate how in-support of this approach I am(and I'm sure many others who are talking about complexifying it a bit, are). I think we just can't help but analyze and juggle ideas around, if only ultimately for our own benefit. I don't think there's a huge "make it more complex or GTFO!" mob or anything, . Just for what that's worth.

 

Rogues in PoE qualify for Sneak Attacks in a large number of circumstances now, almost all of them being conditions on the target (e.g. Flanked, Prone, etc.).  In the case of Flanked, the rogue doesn't even have to be one of the characters doing the flanking.  Many of the other conditions are ones that rogues gain the ability to inflict.  At higher levels, they gain additional bonuses if their targets are suffering from two or more of the qualifying conditions.  So if there's a Flanked target that gets knocked Prone, the rogue gains even higher damage bonuses.

... YES! This sounds friggin' amazing!

 

So, apparently enemies lying flat on the ground are... "prone" to greater punishment from Rogues?

 

(The only purpose of that question was to make the joke. Sorry about that. I actually feed on a specific type of energy generated whenever faces and palms make contact.)

Edited January 21, 2014 by Lephys

[anameforobsidian]

The game will hardly be terrible if we just get generic "stealth" circles. I'm happy with that, and trust that they'll do it as well as they can, which is all we can ask, really.

 

This can't be stressed enough.  Even simple stealth circles are still a big improvement from the infinity engine games.

 

Honestly, I'd even be cool with no facings. With an enemy "seeing" me with his back turned (still just use circles, instead of cones). But, it would still be objectively more interesting stealth gameplay if they could figure out a way to have, say, a sentry with a tiny "vision" circle in a dark, cobblestoney courtyard, but still have a large audibility circle because your footsteps on the stone are louder/echo-y.

 

You're thinking about it from the wrong perspective.  Making sound is a property of the individual making the sound, hearing is the property of the individual doing the hearing.  It's generally hacky and bad design to attribute properties to the wrong agent.

 

But, with separate vision/hearing representation, everything would have three circles:

 

1 for your ability to detect via hearing...

1 for your ability to detect via vision...

And 1 to represent your ability-to-be-detected area. (We'll call it your Presence circle.)

 

The problem is that you're not creating 3 circles, you're creating five.  Agents have the following circles under your design:

Investigating a noise,

Alarmed by a noise,

Investigating a vision,

Alarmed by a vision.

 

That means that in a 6v6 adventurer fight setup you have 30 separate circles to keep track of.  That takes even more area than just a cone and is pretty visually confusing.  You could do something like a light yellow that fades all the way to dark red, but it's still not great.

 

You don't ever even need to see any Presence circles other than your own (your party's), because you can either detect a foe or you can't. If you can't detect it, you're not going to know how much closer you need to get to be able to do so, and if you CAN detect it, then you already know that, however large its Presence circle is, it's now overlapping your detection circle.

 

Sure, it's 3 things for every enemy, but it doesn't use any facings or anything. And it adds 7,000,000% awesomeness to the specifics of stealth-pertinent situations. (I counted, o_o)

 

I do agree with the idea that the stealth system should apply to all agents, but I do think that they could do an investigating circle from the player's end very well.  Enemies who enter the investigating circle are very transparent.  They fade in as they get closer.  Furthermore, what does this scenario add that the current system doesn't?

 

Finally, reactions based on position are probably easy to do, especially considering they have to track orientation already.  The sound / light qualities could just be a variable they add to every scene to change the sizes of the stealth circles, but that wouldn't really offer the stealth game feel since touching grass and cobblestone would have the same circle in the same scene.  So the only other way they could implement it would be by "painting" sound/light values over the scene itself (maybe with global modifiers for ambient effects).  That's a significant amount of extra time / money spent on an incredibly tedious and creatively unfulfilling task.

 

I would argue that if they wanted to make an isometric stealth game in this engine they should prioritize the following features:

- Perception mechanics where character orientation matters.  The core mechanic of a sneaking game is reacting to a guard's actions; guards are who you sneak around.   Not too difficult technically, but could get visually confusing with a bunch of mobs / PCs.

- Line of sight mechanics.  Being able to hide is a very important part of sneaking.  Tedious and moderately hard to implement in current engine.

- Light mechanics.  Staying out of the light is another typical genre behavior.  Since the players are humans, we prioritize lights over sound.  Very hard to do in current engine.

- Reactive materials.  Stepping on metal and making a noise is another basic genre behavior.  Very tedious and moderately difficult to do in current engine.

- Sneak attacks.  Players expect them.  Trivial to implement by themselves, but require an orientation based perception system.

- A tailored skillset.  A game about a thief should have more thiefly skills than a game about a full party.

- Guard AI.  Very hard to do in any engine.

- Real time stealth minigames with enemies still active while it's going on.  Adds a nice bit of pressure that a skill check lacks.  Moderately low difficulty to implement.

- Verticality.  There's a lot to be said for going over and under enemies.  Moderately high difficulty to implement and very tedious.

- Stealth kill animations.  Looks cool, some players will enjoy, but takes a lot of time to do extra work.  Easy to implement, but expensive.

- Perception beyond hearing and seeing.  Would be cool for animals/monsters/monstrous humanoids.  Anywhere from easy to hard to implement.  You could do anything from an extra circle to a set of properties affected differently by different materials.

Edited January 21, 2014 by anameforobsidian

[Lephys]

You're thinking about it from the wrong perspective.  Making sound is a property of the individual making the sound, hearing is the property of the individual doing the hearing.  It's generally hacky and bad design to attribute properties to the wrong agent.

You're right. I think it was that thought that THEN led me to the "only your 'Presence' circle gets adjusted by factors" idea that came after, in the same post. My bad. It's redundant to have darkness/acoustics affect both the detector's circles AND the detectee's "presence" circle.

 

The problem is that you're not creating 3 circles, you're creating five.  Agents have the following circles under your design:

Investigating a noise,

Alarmed by a noise,

Investigating a vision,

Alarmed by a vision.

You are correct again, sort of. It would make that many actual areas, but I was using "circles" to refer to the number of general representations for each person. Well... while it's possible to actually just have the investigating/alarmed states be a time-sensitive distinction (you're always only investigated until you've been inside the single circle for longer than X seconds), that wouldn't work as well as having the distinct area in which you're automatically fully detected (for walking within 5 feet of a person, for example). But, yeah. I realize my description didn't clearly match my meaning, but, for what it's worth, that's what I meant.

 

I do agree with the idea that the stealth system should apply to all agents, but I do think that they could do an investigating circle from the player's end very well.  Enemies who enter the investigating circle are very transparent.  They fade in as they get closer.  Furthermore, what does this scenario add that the current system doesn't?

True, but you still wouldn't need to see the enemy's "if this touches/overlaps any of that player's character's detection circles, you get investigated at the very least!" circle.

 

What does this add that the current system doesn't? Independent factors. You can have situations in which a single foe can't hear very well, but can see very well. Thus, you'd have to stay in darkness and/or blend into surroundings (or just break line of sight), but could get very close to this sentry as long as you weren't seen and weren't extremely loud (running around a stone courtyard in plate boots). The faster you move, the more noise you would make, etc. Whereas, if the sentry could hear super well but couldn't see very well, you can walk around in broad daylight, so long as you stay very quiet. This is all affected by which sentry you encounter in which environment. If you encounter the good-vision sentry in broad daylight, well... you're going to have a tough time of things. If you encounter the good-hearing sentry in an echo-y courtyard or corridor, well... the same.

 

So, that's what it provides. Currently, the system can adjust the circles accordingly, but there's only one representative circle, so nothing can see you when it couldn't hear you, or vice versa. There's no distinct method of sneakery by which one will have an advantage over any other method in any particular situation.

 

Also, for what it's worth, I don't know enough about this, but I believe they said that the process of creating the 2D environments from the 3D ones actually stores a bunch of 3D information (such as height differences of terrain, lighting values, etc.). But, again, I'm not sure what all it stores, specifically, or how useful that is in designing a more complex stealth system.

 

Obviously, Josh knows, and he says it's not exactly so easy that it's a no-brainer or anything, so it's clearly significant enough in its trickiness that they don't want to add complexity to the system unless they discover the current one to be too simple in practice.

[milczyciel]

I'm for one am glad about it. Even if less sophisticated and dynamic than composition of cones of view and range of sound from the "Commandos" series, it's still more interesting than old ways of "stealth" back from IE. Plus it's simple to implement and yet efficient enough to give an appropriate illusion of "something more".

 

In fact this is quite near to my former hopes for active "perception-wise" space around creatures, npcs and player controlled characters (with all it's implications), so I simply can't be unhappy about it.

[Jajo]

I don't have many personal requirements/expectations for the stealth systems, but one of them always is: Will a stealthy character be able to sneak into melee range and make a surprise/sneak attack against an enemy?

Considering how the circles mechanics has been described, I'm not sure this is possible anymore.

If it isn't, what is the reason behind removing this combat feature?

 

EDIT: I haven't read the whole thread, so this has most likely been answered before.

Edited January 22, 2014 by Jajo

[PrimeJunta]

I hope there's a time lag associated with detection. I.e., you'd have to stay within someone's detection circle for a moment before detection triggers. This would also make sneak backstabs feasible -- a stealthy character would be able to get close enough to run and stab before detection is tripped.

[Silent Winter]

I hope there's a time lag associated with detection. I.e., you'd have to stay within someone's detection circle for a moment before detection triggers. This would also make sneak backstabs feasible -- a stealthy character would be able to get close enough to run and stab before detection is tripped.

^This would be cool

 

I'm fine with the 'one-circle to rule them all' approach.  It's abstract, but from a gameplay perspective, it's both simple to understand and sounds fun to play.

I wouldn't object to more complexity if it can still fall into the 'intuitive to use' category, but I don't think it's necessary to spend a lot of time on it.

[Lephys]

Yeah, I mean, I'm not so much worried about how it's done as I am worried about what it provides. I don't need facing-specific cones because that's how real life works, but, at the same time, how will circles prevent sneaky literal back-stabbings, as Jajo asked. It's a good question, really, but more importantly, it illustrates what we want a stealth system to do, regardless of however it accomplishes it (or what other things it doesn't do).

 

I get that, as Junta said, you can use time delays. However, if someone's circle is 7 paces out, and the delay is long enough for you to sneak all the way up to their back and stab them before they even do anything about it, then what's stopping it from simply making it super-easy to sneak through people's circles in the first place, even when you're NOT trying to get up to them and kill them before the "buzzer"? And, I'm not suggesting it's impossible to do that, but, I just think it's a good question to ask. The answer to that question is important.

 

I mean, the BG/IE games' system allowed you to sneak up to and shank folks, but it also didn't have any tactical basis for this. You were just either invisible, or you weren't. Detection circles (as adjusted by relevant stats/factors) are immensely less primitive, but, now they sort of allow you to be detected via pure proximity, no matter how quiet a path you take, or how slowly you're moving, or how away-from-you the target is facing. Will that work the way it is? Sure. But, I will say it would be disappointing if there wasn't something that governs your ability to actually sneak to melee range.

 

Anywho, I'm just saying that it's not just some nebulous matter of "how accurate is it?," but rather, "what factors does it need to represent to allow for more than just stay-out-of-circles gameplay?" I'm sure they're addressing the matter of stealth backstabs, btw, and I'm simply curious to know their plan, when they're ready to reveal it.

[Tuckey]

These stealth mechanics sound awesome! So much potential for an all rogue party to cause mayhem. Much better than what they had in bg2 (it was still fun though)

 

All this talk of commando's, being one of my favourite games, makes me wonder what else they could add from that venerable series.

 

Some things off the top of my head that would be cool.

 

-Please let us use our pet as a distraction.

-I do hope for the ability to utilise disguises.

-Seduction as distraction

-Stuffing bodies in barrels/hiding in barrels

-Climbing up walls

-swimming in rivers to access better entry points etc

 

Probably not for this game but maybe something to consider for the sequel (rogue high-level abilities). Think I'm going to go play commando's now.

[teknoman2]

These stealth mechanics sound awesome! So much potential for an all rogue party to cause mayhem. Much better than what they had in bg2 (it was still fun though)

 

All this talk of commando's, being one of my favourite games, makes me wonder what else they could add from that venerable series.

 

Some things off the top of my head that would be cool.

 

-Please let us use our pet as a distraction.

-I do hope for the ability to utilise disguises.

-Seduction as distraction

-Stuffing bodies in barrels/hiding in barrels

-Climbing up walls

-swimming in rivers to access better entry points etc

 

Probably not for this game but maybe something to consider for the sequel (rogue high-level abilities). Think I'm going to go play commando's now.

Climbing up walls and swimming in rivers to access better entry points etc are going to be in as scripted interactions in certain places. they already showed some screenshots of it

[Tuckey]

@teknoman2

 

I was hoping for more in-game than scripted but still cool.

[Lephys]

We'll probably see both, to be honest (for more than just specifically climbing and such). But, I am quite excited about what the scripted interactions will allow.

[Nonek]

I wonder if one can draw away stragglers from a group through rubbing your circle (ooh err) against theirs, is that tactic possible or is the whole group automatically alerted I wonder?