A.i. Q & A

[oDDity]

There's no such thing as difficulty, only player incompetence.

Concentrate a bit harder, think it through, and you'll do it second time. No need for ten attempts.

We're not making this game for imbeciles, we're making it for hardcore stealth fans who know their way around a Thief level.

[Pyrian]

One possibility would be to combine the techniques, such that only a player-set-off chain reaction that went down in the amnesty period would go unnoticed - a relatively reasonable reaction. (Indeed, one of the things I liked about the "amnesty period" solution is that the guard wouldn't be able to pick out player actions from a chaotic situation - although, as you pointed out, in simple situations it can break down.)

 

Can you chain the flags such that a chain reaction caused by the player is picked out from chain reactions caused by NPC's?

Edited April 6, 2005 by Pyrian

[Ishtvan]

Yeah, a combination method would work well. As for chaining the flags so that an object with PLAYER_TOUCHED set to true or whatever that bumped into antoher object within a certain amount of time would also set PLAYER_TOUCHED on that object.. that sounds like a pretty good solution. You'd just have to figure out how to clear all the flags when the chain reaction was over.

 

[edit: I guess you could just clear the flag when the object comes to rest, assuming that D3 physics works right and it doesn't keep wobbling around in place or something]

 

Btw if you have any interest in helping out with the programming, we can always use more (altho this decision is up to our lead programmer Sparhawk, not me), so if you're interested, you might want to shoot an email to recruiting@thedarkmod.com

Edited April 6, 2005 by Ishtvan

[TazmanianDevil]

If you were able to play back the final game as a movie (with no loading interruptions) you'd have this Matrix type video of this guy dodging everything perfectly and using everything at the perfect time, not wasting a single bit of anything.

That would be cool to watch.

[TazmanianDevil]

AI. At this point, the AI believed that it had just been hit by an object out of the blue, so it went to a full alert

Being startled by being hit be something isn't necessarily a bad thing. If I slightly kick something and a box falls on me, I might freak out but I'd probably quickly realize that I was the cause after which I'd probably be more cautious. So maybe instead of a full alert, you could trigger just a heightened alert.

 

And perhaps, you can decide that you were the cause if the object that hit you was close to you. I don't know how easy it is to determine the original location of an object before it starts moving.

 

I would definitely prefer no reaction over the guard totally freaking out though and starting to look for me. That's like a dog chasing it's tail.

[Ishtvan]

Well, I'm not talking about a case where you have like 10 boxes stacked up and it bumps a box at the bottom and they all fall on them..

 

I'm talking about the case where a box is in their path, and they kick it out of the way (this kicking out of the way is already included in D3 btw). Due to physics, when they exert force on the box, the box exerts a force on them (equal & opposite reactions and all). So in this case, it really doesn't make sense for them to freak out. Of course they feel the box touching their foot, they intended to kick it, so it should not be surprising that they feel some physical contact when they kick something!

[sxotty]

That would be cool to watch.

Have you seen Q2DQ? That is quake 2 done quick, it is quite entertaining to watch what the people thought of...

@ Oddity I have seen plenty of times in games where it devolves into that. If the levels are intelligently designed that is good and all, but I would rather they were difficult enough I wanted to save often, and rather there were different ways to suceed at different times. It was boring in TDS when there was only one route, and it is easier actually when there is only one way to do something, then you don;t have to think about which is a better idea.

Edited April 8, 2005 by sxotty

[oDDity]

I'm just the sort of player who seeks out difficulty. I don't play games for fun, I play them to challenge my mind.

For example, I just finished soloing Baldur's Gate 2, soloing as a thief on insane difficulty. It was more frustratiion then fun, but I was set on doing it, and I did it no matter what the difficulty was - it required lots of lateral and creative thinking and improvisation, which should be a prerequiste for all games worthy of my time.

[encomium]

It was boring in TDS when there was only one route, and it is easier actually when there is only one way to do something, then you don;t have to think about which is a better idea.

HAhah Oddity, yer the greatest!

 

 

I agree with sxotty. If there were more than 1 way to accomplish missions, that'd be amazing. It would obviously require hella more work and planning, which I'm not sure is what this team needs at this stage.

 

 

Edited April 8, 2005 by encomium

[Springheel]

Well, we're not designing missions right now, we're designing the tools to let OTHER people design missions.

 

However, we are going to be including more ability to physically interact with the gameworld that should allow for some interesting possibilities.

[Ishtvan]

I'm just the sort of player who seeks out difficulty. I don't play games for fun, I play them to challenge my mind.

For example, I just finished soloing Baldur's Gate 2, soloing as a thief on insane difficulty. It was more frustratiion then fun, but I was set on doing it, and I did it no matter what the difficulty was - it required lots of lateral and creative thinking and improvisation, which should be a prerequiste for all games worthy of my time.

Did you just set an insane number of traps around dragons before waking them up?

[Ishtvan]

I agree with sxotty. If there were more than 1 way to accomplish missions, that'd be amazing. It would obviously require hella more work and planning, which I'm not sure is what this team needs at this stage.

I'd just like to point out that in T1-T2, there were MANY ways to accomplish a mission. So it's been done before, it can be done again. It does not require more work or planning, it just requires good FM design and mapping skills. It takes more work to make maps with rooms that are not all cubes too.

[oDDity]

Did you just set an insane number of traps around dragons before waking them up?

THere's only one 'must kill' dragon in SoA anyway. Trying to lure it to the snares is a pain in the ass in itself. You can only set a 7 traps in one area with ToB installed. There's more later in ToB though.

 

Mages are a real pain in the ass, they seem to have some anti-backstab spell, and of course, true sight.

[Vadrosaul]

I'd just like to point out that in T1-T2, there were MANY ways to accomplish a mission. So it's been done before, it can be done again. It does not require more work or planning, it just requires good FM design and mapping skills. It takes more work to make maps with rooms that are not all cubes too.

I love you Ishtvan

[Dram]

There's no such thing as difficulty, only player incompetence.

Concentrate a bit harder, think it through, and you'll do it second time. No need for ten attempts.

We're not making this game for imbeciles, we're making it for hardcore stealth fans who know their way around a Thief level.

That's true oDD. But as an example, my mod for Thievery has the option of failing the mission if you get seen (a thing all iron men love) and let me tell you that i have a pass rate of 1 pass to 20 fails and i'm not exactly a newbie either. I've played T1,2,3 and Thievery, etc. I know how to sneak but sometimes you just make a mistake (like accidentally press jump instead of crouch...). I do too search out very hard games and enjoy a good challenge, but sometimes i do feel like a little bit of an easier ride.

[Domarius]

Have you seen Q2DQ? That is quake 2 done quick, it is quite entertaining to watch what the people thought of...

Thanks for the heads up, that should be really cool to watch

[FishFace]

On reactions to physical things:

 

I think that setting a player_touched flag is too simple, certainly. For example - imagine if two people walk down a corridor, and one knocks something into the other. The injured party should get annoyed with the first person.

I guess you have to think about what you'd do in the situation. If you walk past a table, and the candle falls off, you'll assume that it was your own fault. On the other hand, if it just fell over, then you'd get spooked. If there was nothing indicating what had caused its movement, then that's what'd happen - you'd be freaked out. Otherwise, you'd start looking for signs of an intrusion. Obviously, if something was lobbed across the room, you'd go over to where it came from.

 

Likewise, if a box falls on your head and you were walking past a large stack, you'd probably assume you'd somehow nudged the stack - it was sitting awkwardly, or something. If it flew across the warehouse, then you'd run over to where it was, and start yelling for whichever young taffer was back there to get out here to be given a good hidin'.

[Ishtvan]

If you want to code in deductive reasoning and a human psychology model for the AI, be my guest.

[FishFace]

I only gave suggestions for good AI, not easy-to-code AI

 

But seriously, having proper deduction for events, even rudimentary, would be great. It's similar to how groovy the 'blame' AI was in TDS. Remember all the pagan/hammer battles? They made Stonemarket lethal, because everyone was looking fer someone ta smash!

[Ishtvan]

I think an attribute for who touched it works pretty well, that can be propagated along in physical chain reactions.

 

If the AI knocked a table and knocked a candle over, they would know that they did it, and not freak out.

 

If an object hits the AI, or falls and makes a sound or whatnot, the AI checks to see who the "instigator" was. If it was an enemy, then they get alerted.

 

Yes, in reality, friendly AI could knock something over and then alert another AI. But in reality, if two guards did this, the guard who caused the ruckus would explain verbally that it was his/her fault. Otherwise there is no way for the guard that was alerted by a friendly to get de-alerted.

 

So yes, it's conceivable that we could code someting where a friendly AI can knock something over and alert another AI, and then realise that that's what happened (had part), and explain to the other AI that it was just a false alarm, because they knocked X over.

 

Not only would this take a lot more coding to make a friendly AI realise that they alert another AI, then talk to that other AI and de-alert them, but it would also take more voice sets work.

 

So overall it would require a lot of work, and the only payoff would be maybe 1 or 2 players saying "That's sort've cool.." or for people who enjoy watching slapstick AI comedy play out before their eyes.

 

Stepping back in realism, the next most realistic option is to just assume that AI are always in communication (but don't actually show it), so that when a friendly AI knocks something over and alerts another friendly, the friendly will not go to alert because it is assumed they know what's up. Otherwise, if they get alerted, we would need a way to communicate and de-alert.

 

So overall, I would rather spend the time working on something that would lead to much improved gameplay rather than putting a lot of work into making the AI respond realistically to this specific situation that doesn't really impact gameplay at all. (If AI just don't go to alert when a friend causes the physics reaction, it is not significantly different for gameplay than if they go alert for a second and then the friend tells them it was their fault, don't worry)

[Pyrian]

Personally, I found the senseless battles in T:DS to be rather weird.

 

Anyway, in terms of deductive reasoning, I think that people can reasonably figure out what they caused and didn't cause, making a simple flag sufficient. The "amnesty" concept would allow a guard who nudged something to get pelted - in a few circumstances that could be realistic, but in practice it just doesn't add value.

 

Programming in actual deductive reasoning is not remotely plausible.

 

EDIT: I do like the idea of "oops, I knocked it over"->"careful, you knuckleheaded taffer!" barks, though I don't think it would be necessary to have them involve any sort of alert-state modification.

Edited April 11, 2005 by Pyrian

[FishFace]

In the realms of pseudocode (I don't know the style of Doom 3 coding, sorry)

 

if ( isFriendly( offender->touched ) ) {
 after( offender->touched::bark( imclumsy ) );
 self::bark( clumsy );
}

 

Not too tricky?

 

I mean, you could extend it to something like:

 

if ( isFriendly( offender->touched ) ) {
 self->offendedbyfriend = offender->touched;
 self::alert( 2 );
}

...

void spotObj ( object *spotted, float distance, float light ) {
 if ( spotted->offendedbyfriend == self ) {
   self::bark( imclumsy );
   spotted::bark( clumsy );
   spotted::alert( -2 );
   self->offendedbyfriend = NULL;
 }
}

 

I hope you can understand that! But basically, it doesn't look too difficult to do it either completely scripted, or with an element of the offender having to actually see his clumsiness.

Obviously this could be made more complex by having a timer, to prevent the guard wandering all the way around the mansion and returning to find his mate blundering around, then simply calming him down - more likely he'd forget, perhaps.

Also, you can figure in an anger idea, to make the offended thwack the other after a while... Although this particular method would result in you not being able to pretend to be the guard, and lob stuff at the other guy from the shadows. That, I suppose, would require more deductive reasoning. Although perhaps it's not entirely infeasible - if you were to make the offended hunt for someone to lay the blame on, then the first person he sees near where the attack came from - boom!

This would be great fun, if you could sneak behind an enemy archer, then shoot another guard in the foot. He turns around, spots the archer and chaaarges toward him!

 

Maybe I'm getting carried away? hehehe...

[Ishtvan]

Thanks for the suggestions, and we'll think about it. Like I said though, we still have a LOT to worry about in terms of making an AI a challenging opponent before we start spending a lot of time on specific situations like this.

Edited April 12, 2005 by Ishtvan

[Drakon]

I agree with oddity - game, to me, are much more fun if they are challenging. I played through T3 before they came out with the difficulty fix. I have no idea how much it would have helped, but it seemed rather easy the first time through. Kind of ruined the experience...

Anyhow, winning is much more rewarding if it was enjoyable AND difficult. More strategy, less button clicking. Throw in some rewarding gibbing/exploding/etc to make it even better.

 

Also, another probably useless fact about Drakon, is that I don't care too much about visual quality. I care much more about atmosphere and setting (as long as the graphics aren't bad enough that they take away from the atmosphere) Atmosphere is a great gameply helper... Baldur's Gate and Thief: TDP had great atmosphere. Half-Life had a pretty cool setting, too.

[FishFace]

I'm afraid I'm a bit of a sucker for purrdy graphics, which Doom3, HL2 and TDS have in ample quantities. With graphics as good as they are, immersion from graphics is improving all the time. Although this was broken in TDS at regular intervals due to bugs, there were times when it was absolutely fantastic.

Nonetheless, you can be immersed without good graphics - but I think it's much more difficult.

 

AI that react appropriately (for me) brings atmosphere, because it means you can act as you would in real life, and get real-life feedback (a la getting guards to fight by pitching them against each other) But in the end, it's just as important to get the music, the graphics and the other elements of gameplay right.