Jan 3 Update: More Ai Smartness

[Domarius]

You all probably know where I stand on this by now. I agree with oDDity in principal but I think restricting saves in some manner is enough to cause the tension associated with knowing if you fuck up you've got a lot to lose. It really makes scarey things more scarey.

[ZylonBane]

So, unless you're prepared to be a man and accept some consequences for failing, there's no point swaggering around boasting about how you want the AI to be more realistic and the game more difficult. Best to keep it at the current childish level that all games maintain.

Let me guess... if you beat someone at a "childish" game like, say, Chess, you celebrate by punching them in the face? After all, losing isn't apparently a sufficient consequence, right?

 

Perhaps some day you'll come to accept that normal humans play games for FUN, not as affirmations of their sexual identity.

[oDDity]

It's not an equal comparison. Chess is a purely conceptual game with clearly defined and rigid rules. I know chess is often compared to life, but it's not an allegory for life, it's just takes some natural concepts and tenants from life and reduces them to their purest intellectual form.

The dark mod and other games seek to directly simulate both the unpredictable nature of reality, and the logical consequences of it. Therefore, it's pointless to clamour for more realism and more difficulty when all paths lead to the ultimate unreality.

We need a new paradigm. The current one is a contradiction of purpose.

The idea that you're pretending to be a real person in a real world populated with other real people in life and death situations, but with absolutely no risk involved at all, is ridiculous, the idea that you should be having 'fun' in these pretend and toothless life and death situations is even more ridiculous, and this whole charade of having the player 'die' is a joke, yet it pervades every game.

[sah]

We need a new paradigm. The current one is a contradiction of purpose.

The idea that you're pretending to be a real person in a real world populated with other real people in life and death situations, but with absolutely no risk involved at all, is ridiculous, the idea that you should be having 'fun' in these pretend and toothless life and death situations is even more ridiculous, and this whole charade of having the player 'die' is a joke, yet it pervades every game.

 

A little like a simulator, isn't it. I'm thinking about military-used sims that armies use to train recruits. You play a sim to get a feel of how things work in real life, and while RL cannot be modelled up to every little detail, most of its mechanics can. When you lose in a sim you sometimes can look at a recording of your actions and see what you did wrong.

 

I'm not saying TDM should aim at becoming a sim, only that simulators are somewhere between traditional Thief and your idea.

[Komag]

I like Thief and other games, I have fun playing them with the current paradigm, so I don't need a new one.

[Maximius]

Let me make a few clarifications, I am not calling for unbridled realism at the expense of every other factor. What my key example was intended to describe was how a very simple thing like the treatment of keys could inject just a dash of realism in a scene and my ultimate point was that lots of little details like that, made with lots of gradual little tweaks that aren't necessarily that time consuming when picked off one by one, can raise the bottom line of realism without massive overhauling of AI or textures or whatever the heck.

 

Its true that real realism is impossible, but then my goal was never realism for its own sake. Rather, I look to good video games for what I call "cinematic moments" when my actions, the AI and/or environment, and my sense of suspended belief all come together to produce that magic that makes gaming truly addictive to me. Realism adds to these moments, never to a sense that the whole thing is real. Realism requires a level of suspended belief and it collapses when too much is asked of it. The difference for me is whether I finish the game or not.

 

By addressing lots of "little realisms" I think that a lot of pressure can be taken off of the player to judge a game experience in terms of realism. Then, along with other things like action, strategy, using the environment, and especially storyline you get into the game easier, the unavoidable unrealities are mitigated a bit by the absence of a lot of smaller ones. Small stuff like food that tumbles off the plate when you steal it, plants that shake when you brush through them, are things I'm looking into. AI interactions, wonderfully adaptable in the Mod, can be tweaked to represent this as well. I'm sure there are limits due to practical matters of processor power and such but there has to be a lot of gray area too that can be "filled in" with details that aren't going to run down FPS or utterly fracture gameplay. Thats what I'm interested in.

Edited January 8, 2007 by Maximius

[PinkDot]

The dark mod and other games seek to directly simulate both the unpredictable nature of reality, and the logical consequences of it. Therefore, it's pointless to clamour for more realism and more difficulty when all paths lead to the ultimate unreality.

We need a new paradigm. The current one is a contradiction of purpose.

That's why there are thousands of computer games, one passing by after another, with only few of them still glowing like embers in doused campfire. The inconsistency which computer games can offer nowadays makes them so short-lived human's creations. Hopefully it's only because computer games are so new medium and didn't shaped their classic form yet.

My favourite game is board game GO - the oldest board game known (about 4000 years), which has only one main, very simple rule, which tells when stone (pawn) is alive or not. As a result of that rule with relation to the board (which is the environment for the stones) are hundreds more complex rules and strategies, for studing which you can dedicate your whole life.

I wish computer games had such consistency and simplicity. Nowadays computer games are like narcotic visions, which take you to other reality but works imperfectly, cause still reminds you that it's "only a game" - f.e. you see the door, which you want to open, but you realizes that it's only flat texture imitating door. The direction that current games go (or try to go) is some Matrix-like reality probably, which is a questionable worth.

[Ishtvan]

@Maximus:

It's harder than you think to just build up realism by adding "little bits" of realism. I'm generally for more realism with customizable options, since different people obiously have different ideas about what's fun vs frustrating, but I disagree with your statement.

 

Working on the mod, I've found that when you increase the realism of one aspect of a model, it often necessitates increasing the realism of a different apect of the model, otherwise the whole model will break down and be less realistic than you started with.

 

An example we've seen on the public forums is a demand that the AI have different vision parameters in their peripheral vision than their central vision. Right now guards have equal detection probability across their entire field of view. The claim is that they should have different light/movement detection abilities in their peripheral vision. Sounds innocent enough, right? But if you were to just implement that by itself, calculating peripheral vision based on the direction their head is facing, you'd have a situation where their corneas are alway physically straight forward in their head. Within their FOV, you'd get predictable and abrupt changes in their vision reactions based on where you were standing relative to their head facing, which is not what happens in real life.

 

Real people on guard are constantly scanning their eyes over the field of view allowed by their head. By just adding the peripheral/central difference without adding eye scanning movement, we made the system less realistic than it was before. To un-break the model, we would have to do a lot more work and make an eye scanning algorithm independent of the head turning algorithm, and take into account psychological factors like how wary they are, fixing their eyes on something suspicious or moving across it, etc.

 

The net result after all that work is: An idle guard who's moving their eyes around randomly would have an equal time-averaged detection-probability across their entire field of view. That's the same result as what we started with, after a lot of hypothetical work. Making just one small thing more realistic did not improve things, instead it broke the model and necessitated making other things more realistic, and in the end it effectively didn't accomplish anything.

[Maximius]

hmm, i see what you are saying ishtvan, thats an interesting point about one upgrade demanding another. I could see something getting tweaked, say spilling food from a stolen plate, that then demands water must spill out of a drinking glass or a bucket because if one then why not the other? Or upgrading an AIs reaction to one event then cries for upgrading in all similar events or something.

[sophisticatedZombie]

I'm probably reitering what has already been said, but there are many ways to handle missing loot:

 

The map designer can adjust the amount of alert caused by something being spotted missing from a high level to 0. In most cases, the map designer would probably make most loot not trigger a noticeability alert. For example, a stack of coins on somebody's desk may have been moved by someone else, and a guard wouldn't necessarily assume that it was missing if it wasn't their own gold. So, it could realistically be marked as not causing an alert.

 

However, that priceless ancient statuette on the pedestal in the foyer might be a cause for concern if it is suddenly missing. But, what if the map designer gives the player a special note prepared in advance, with knowledge about the name and signature of a curios restorer who sometimes works on things int he owner's collection, saying "I'll have the statue repaired within 3 days, as I've already found a substitute for the cracked gemstone eye." Well, they could script things so that placing the note on the statue pedestal turns off the visual stimulus (or causes guard to remark on the cracked statue eye and how they didn't do it.).

 

In other words, we want to provide enough AI to ALLOW the map creator to have a lot of interesting emergent behavior. The scripting engine and the hooks we will provide should allow map designers to create interesting and varied solutions to problems. The map maker can tweak the senses and stimulation levels to provide whatever gameplay experience they want. More than likely, many will choose to have different types of planning and forethought required at different difficulty levels.

 

And, if you don't like the behavior, you can just override it in scripts and def files that come with your map (or in other people's maps for that matter.) In conclusion, it should be possible to make maps that appeal to various peoples tastes. Our job in building AI is just to provide the breadth of behavior possible so that the map makers can exploit it as they see fit.

[Mark]

We need a new paradigm. The current one is a contradiction of purpose.

The idea that you're pretending to be a real person in a real world populated with other real people in life and death situations, but with absolutely no risk involved at all, is ridiculous, the idea that you should be having 'fun' in these pretend and toothless life and death situations is even more ridiculous, and this whole charade of having the player 'die' is a joke, yet it pervades every game.

I think your problem is that you're just not willing to meet a game half-way in order to be immersed. You refuse to have your disbelief suspended, instead insisting that the game must be totally real with respect to the way you interact in its world. I call it a problem because I don't understand how you could possibly expect any game to ever meet your standards, even in the abstract. Ultimately, if a game has real consequences then it's not a game, right?

[kohan69]

There's no point making AI realistic, when it doesn't matter anyway, even if they do catch and kill you, you just start from your last save.

 

Best to keep it at the current childish level that all games maintain.

I completely disagree, or I probably missed the sarcasm.

 

Saving system could be restrictive. Best example is Hitman1 (no saves during mission) and Hitman2 (limited saves per mission per difficulty level)

One idea for a mapmaker of a 5 missions in a town could be: if caught, arrest and put to jail, if die, erase all saves.

 

I understand that balancing gameplay and realism a challenge, but one thing that could always be better is the AI.

 

 

 

I've found that when you increase the realism of one aspect of a model, it often necessitates increasing the realism of a different apect of the model, otherwise the whole model will break down and be less realistic than you started with.

 

For example: Overuse of the wine-bottle gun?

Adding realism form one direction, yet creating unrealism from another point at the same time?

[ZylonBane]

Saving system could be restrictive.

But it won't be, because 99% of PC gamers hate the mother-loving shit out of restricted save-game systems. And with good reason.

[Ishtvan]

@SZ: Well said. That's a good explanation that re-emphasizes how we're increasing the breadth of the AI capabilities as you say, but it's up to the FM authors to decide when it makes sense to use what capabilities.

 

For example: Overuse of the wine-bottle gun?

Adding realism form one direction, yet creating unrealism from another point at the same time?

Yes, that's another good example. AI standing beside a waist-high table and shaking their first at you impotently: Not realistic. AI doing something about it and throwing stuff at you: more realistic, but then you have to consider what they would throw at you, make models for it, talk about more advanced systems where they have a limited inventory or determine stuff to throw from the type of room they're in or pick up stuff from around them to throw, etc.

 

I think in this case it's manageable though. The fact that the AI have an unlimited amount of random crap to throw at me won't keep me up at night. Ideally the AI would be able to mantle up on the table or chop off the legs with their sword, but that's not practical at the moment. Also, they'd probably be very vulnerable while mantling, if you're standing right there with a sword, so that might not be a realistic strategy anyway.

[Ishtvan]

But it won't be, because 99% of PC gamers hate the mother-loving shit out of restricted save-game systems. And with good reason.

 

Yes, reasons like the random crashes or mission-ending bugs in Hitman1, 2, etc that inevitably happened when you were 30 minutes into a mission. Glitches are inevitable. When glitches occur, saving and restoring can be more immersive than not. For example: suppose you step on some piece of geometry and get permanently stuck or die for no reason. If you've saved recently, you can go back a few minutes in time and pretend that never happened. If you have to repeat a whole 30 minute section because your character got permanently stuck on a crack in the sidewalk, believe me, that is not more immersive.

[Gildoran]

I still don't think that's a sufficient argument against save-restricted systems, since a good system would probably be able to take that into account and design around it...

[OrbWeaver]

Save restriction is the gaming equivalent of DRM. It fails to achieve what it is supposed to (stopping piracy/increasing realism) and just gets in the way of legitimate uses (backing up CDs for personal use/saving your progress to resume later).

 

If such a feature were implemented in the Dark Mod, it is practically guaranteed that the first mod released would be a tool to disable this annoying behaviour -- essentially wasting the time of the developers who implemented it AND the developers of the crack who have to un-implement it.

[Gildoran]

If I were to implement it, I'd just add a cheat cvar to disable it. Players who don't like it could easily disable it, but it would send a clear message that turning off save restrictions is not the standard way to play the game.

[oDDity]

Exactly, what Ishtvan describes is a worst case scenario and is too rare an occurrence to take into account when designing a system.

Next thing you'll come up with is that a bolt of lightning might hit your house and knock your power out, so it wouldn't be fair to restrict saves.

You know my point of view, I want no saves in a mission at all. You get one chance, one life, and that's it. Idont want to hear any crap about' Ohh, but I'm so busy, I only have 10 minutes every decade to play games so this would'nt work for me'

IF you haven't got a few hours to set aside to play a game properly, like you spend a few hours with a movie, then don't bother at all, and don't whine about it.

 

But it won't be, because 99% of PC gamers hate the mother-loving shit out of restricted save-game systems. And with good reason.

Well, thats an easy solution then.

What we need are a new set of gamers to replace the current bunch of wankers that the game industry panders to. Spineless geeks who want all the thrills and spills of dangerous life and death situations, but without any risk or inconvenience at all. You, for example, are a perfect specimen of generic gamer that will hopefully soon be extinct.

People like you grow to like whatever you're fed, we'll change your diet, and soon enough you'll be happy again.

[OrbWeaver]

Spineless geeks who want all the thrills and spills of dangerous life and death situations, but without any risk or inconvenience at all.

 

What is the point of a computer game, other than to simulate the thrills of real life without the risk or inconvenience? If you wanted the risk and inconvenience you might as well do the real thing.

[Ishtvan]

Exactly, what Ishtvan describes is a worst case scenario and is too rare an occurrence to take into account when designing a system.

 

It's not that rare in my experience. It happened a number of times playing the Hitman series, and every time I had to fight the urge to yank out the CD and bite it in two. I soldiered on because I enjoyed the gameplay, but each time a glitch made me lose progress it was very non-immersive.

[Domarius]

The proof is in console games (yes, non-linear ones too, like pretty much every RPG) which have specific save points, or save locations, and this works great. Fortunately these platforms have little choice but to use save points, and implementing them well has become an art.

 

Carefully placed, it doesn't even matter if the game stuffs up somehow - well placed apart save points give a good balance between the risk of losing things due to artificial interruptions, vs the tension and fear caused by having a lot to lose if you fail.

 

Now that I have a PS2, I'm sucessfully experiencing this all the time on many games. (yes, non-linear ones too - I keep repeating this because I had the argument "Save points are for linear games only!" thrown at me many times, which is so untrue it's not funny)

 

And I applaud Quake 4 and Half Life 2, even though they are PC games, if you let the auto-save do the work for you, you realise that they are actually well placed save points. Though these games are linear so it's not a great example, but I think it shows the trend is catching on. The most important thing to note there is it's a completely optional system, so the average gamer isn't alienated. You can still save anytime you want if you want to (you wuss ) which is why when we (hopefully) implement a save restriction system in DarkMod, it would be purely optional.

 

What is the point of a computer game, other than to simulate the thrills of real life without the risk or inconvenience? If you wanted the risk and inconvenience you might as well do the real thing.

You need the risk to create fear. It might seem like a simple thing until you've experienced it. You feel more vulnerable and all the emotions are intensified. You want to plan ahead more. You devise plans based on risk priority rather than trial and error. If you know respawning is only a click away, you'll try something, die, try something else, die. If you're forced into having to restore from a point further back, you start to think like "Well, I could go that way, but I think it's actually really dangerous and I might try this other way even I might fail what I set out to do, at least I'll still be alive - and if I go this way, and I get into trouble here, then I have a plan B, I can retreat to here... " etc. And your hearts beating faster and faster, and then you get a huge adrenalin rush when you jump in and do it. Sometimes it pulls off first go, A-Team style, and the feeling is euphoric. Sometimes it fails, Jacky Chan style, and you're stuck in an intense heart-stopping chase escaping for your life, trying to think on your feet.

I know you can argue you can still do this while still saving anywhere you want, but the FEELING is different, and you do think differently.

[oDDity]

What is the point of a computer game, other than to simulate the thrills of real life without the risk or inconvenience?

 

That was my earlier point. The current paradigm is set that way, but who says that's the way it has to be?

We need a new one. I'm not saying you should risk your actual life of course, but there should be some element of risk and inconvenience involved, it's the only way to effectively simulate life an death situations, where you actually do have something to lose in that life and death situation you're pretending to be in. Otherwise it's just meaningless, childish nonsense.

The basic principal is, that when you're risking something real, the whole experience becomes more realistic as a result.

 

If you wanted the risk and inconvenience you might as well do the real thing.

 

What, really be a water arrow-wielding thief in a steampunk universe?

 

It's not that rare in my experience. It happened a number of times playing the Hitman series, and every time I had to fight the urge to yank out the CD and bite it in two. I soldiered on because I enjoyed the gameplay, but each time a glitch made me lose progress it was very non-immersive.

Ok, so you're basing this entire thing on your experience with one buggy 5 year old game then?

I'd call that quite rare.

[ZylonBane]

The proof is in console games (yes, non-linear ones too, like pretty much every RPG) which have specific save points, or save locations, and this works great. Fortunately these platforms have little choice but to use save points, and implementing them well has become an art.

Save-point systems are only ideal for people who have no problem spending large blocks of time playing a game. For people who actually have jobs and lives, save-anywhere is essential. A video game is, ultimately, an entertainment experience-- just like non-video games, books, movies, and TV shows. And what all those have in common is that you can stop and start them AS YOU WISH.

 

Vocal detractors of save-anywhere are, without fail, prima donna control freaks.

 

(Note also that consoles are starting to use PC-style saving now that they have to storage space to allow it. Save points were never a feature, they were just something console game makers had to deal with, so they made the best of a bad situation.)

[OrbWeaver]

You need the risk to create fear. It might seem like a simple thing until you've experienced it. You feel more vulnerable and all the emotions are intensified.

 

I don't doubt it, but you are assuming (or dictating) that the player should want fear from a computer game. Many people, myself included, play computer games to relieve stress, not to increase it.

 

Besides, the Cradle and the Haunted Cathedral managed to achieve fear without using save restriction, so it cannot be a prerequisite.