Stealth and monster sleep

[Nick]

In 4.0.5, stealth works essentially as follows:

1. With stealth 0, a monster will always "wake up a bit" (ie have its sleep counter reduced) if the player is within it's scanning range. The chance of waking up a bit is reduced by 1/2 for every 3 points in stealth, so a character with +6 stealth has a 25% chance of reducing the monster's sleep counter, etc.
2. The amount the sleep counter is reduced is 100/d, where d is the straight line distance between the monster and the player.

I have a vague feeling that it might be better to have the amount of sleep reduction affected by the stealth as well, but it's not a strong opinion. I am currently redoing monster pathfinding, so this seems like a good time to examine stealth.

So, what do people think? Is stealth working well currently? Is there an obvious improvement to make?

[Philip]

Well, if you're already reworking noise and monster pathfinding, you could make the monster sleep counter be reduced by noise, and noise reduced by stealth. That way stealth affects monster pathfinding, and if there's a simple way to see the noise being produced, there's also a relatively simple way of seeing how much you're waking up monsters. This would mean somehow changing how monster scanning range works, but that's already part of the monster pathfinding project, right? It would also mean figuring out how stealth should affect monsters who track by telepathy, and how their sleep should be interrupted.
I feel like it would be rather dangerous to double-dip on the power of stealth, for fear of making it even more powerful than it already is, but it all depends on game balance.

[Derakon]

I have a vague feeling that it might be better to have the amount of sleep reduction affected by the stealth as well, but it's not a strong opinion.

Stealth is already on an exponential curve; doing this would further exacerbate that curve, making unstealthy characters really, really unstealthy and stealthy characters...even more impossible for monsters to detect than they already are? That seems like an odd choice to me; the exponential curve itself is kind of odd to begin with.

If you want to revamp stealth and make it meaningful, I feel like that means that you need to have a system where the player can make meaningful choices in the current moment to impact their stealthiness, and then you need to expose enough information to the player that they can make those choices. That sounds like a hard problem. It also almost certainly means starting by scrapping the existing awakeness system.

[Pete Mack]

The exponential curve is appropriate. The effects of each point of stealth is exactly as valuable as the previous (until you reach perfect stealth.) Every other function would have undesirable behavior at one end or the other of the power curve.

[Derakon]

The exponential curve is appropriate. The effects of each point of stealth is exactly as valuable as the previous (until you reach perfect stealth.) Every other function would have undesirable behavior at one end or the other of the power curve.

Is it really appropriate? When you say that each point is exactly as valuable as the previous point, what do you really mean? In practice stealth acts as a poorly-defined timer for each monster -- you have a certain number of turns (modulated by distance to the monster) before the monster wakes up. If you go from 5 turns to 10 turns with your first 3 points, then from 10 turns to 20 turns with your second 3 points, that's not exactly as valuable. Exactly as valuable implies a linear relationship, i.e. the second 3 points should have taken you to 15 turns.

Put another way, why is the utility of stealth plotted (and reasoned about) on a log-scale graph?

[Nomad]

If you want to revamp stealth and make it meaningful, I feel like that means that you need to have a system where the player can make meaningful choices in the current moment to impact their stealthiness, and then you need to expose enough information to the player that they can make those choices. That sounds like a hard problem. It also almost certainly means starting by scrapping the existing awakeness system.

Maybe this is too simplistic, but how about making stealth a straightforward percentage, and rolling against that each time the player takes an action to see if it wakes the monster? With appropriate modifiers for distance and "noisier" actions.

So moving into a square at default distance x from the monster would be the standard action that gets the basic roll of Stealth against a d100. If you're further away or doing something quieter like resting, you get a bonus to the roll. If you're closer or doing something noisy like attacking or tunnelling, you get a penalty. Different monster types could also have their own 'alertness' modifier representing how easy/hard they are to wake.

ETA: I guess, for logical consistency, rather than rolling for every surrounding monster, there should actually just be a single roll per action to determine how much noise you make doing it (d100-Stealth for walking, modifiers as applicable for other actions) and then that would be tested against the wake-up threshold of all monsters in range (based on their distance and alertness).

[Egavactip]

What is wrong with Stealth as it is? I have zero complaints about Stealth. This seems like change for the sake of change and I don't like it.

[Pete Mack]

Nomad: a linear function makes early stealth useless--a hobbit and half troll become almost indistinguishable if you use a scale with 16 gradations, and a defender(+1) is indistinguishable from (+4). You'd need to cut to range over which stealth varies.
It's not an accident that noise is measured in decibels in real life, just as in the game.

[Derakon]

Nomad: a linear function makes early stealth useless--a hobbit and half troll become almost indistinguishable if you use a scale with 16 gradations, and a defender(+1) is indistinguishable from (+4). You'd need to cut to range over which stealth varies.
It's not an accident that noise is measured in decibels in real life, just as in the game.

Your logic isn't sound. Let's say that every +1 equates to an extra turn spent immediately adjacent to the monster before it wakes up. A half-troll with a stealth of 0 gets 0 turns next to the monster before it wakes up, while a hobbit with 8 stealth gets 8 turns, and a level-50 hobbit rogue (or that same half-troll with a ton of extra stealth gear) gets 16 turns. I wouldn't call that indistinguishable. And even if you disagreed with me, it'd be trivial to add a multiplier so that each point of stealth buys you 2 turns, or 5 turns, etc.

What's the real utility that the exponential system is buying us?

[Philip]

I agree that stealth is pretty much fine already, but that isn't to say it couldn't be better. Anyway, when monster pathfinding (and as such, noise) is being changed already, monster awareness seems like it should be changed too. For myself, I feel like it would be nice if sleep and pathfinding were related in some way.

[Pete Mack]

@Derakon-
Nomad's proposal was linear in the other direction, meaning that early points made little difference. Your proposal causes later points to be of lesser value: the difference between 12 and 16 turns is essentially immaterial.

[Derakon]

@Derakon-
Nomad's proposal was linear in the other direction, meaning that early points made little difference. Your proposal causes later points to be of lesser value: the difference between 12 and 16 turns is essentially immaterial.

That depends entirely on how much you have to do in the vicinity of the monster. If you're sneaking through a vault, those extra four turns could be very important indeed.

Current stealth feels effectively trimodal -- there are characters that can count on stealth always working (so that no monsters wake up unless they take damage), there are characters that can count on stealth not working because they're so noisy, and there are characters that generally have some monsters wake up while they're in the middle of doing things, while others stay asleep. I'm fine with the gameplay impact of this, but it's kind of hard to measure at anything more than this high-level, super-qualitative level. That makes it hard to reason about in gameplay terms, and thus hard to make decisions about.

[Pete Mack]

I'm not sure why you wish to reason about it. As a player, or a game designer? As a player, it's easy: adjust you game play to your current stealth mode. Watch for monsters that have high alertness like uniques and Hydras. I don't like the idea that it should be completely deterministic.

[Nick]

OK, my working plan then is to keep the system pretty similar (and in particular maintain the "mean time to wake up" of monsters), but make the code read a bit more intuitively. The main change I would think of is to have monster sleep-reduction based on the stealth-affected measure of distance I have monsters using for tracking, rather than distance as the (PASS_WALL) crow flies.

[Pete Mack]

This sounds a bit like change for the sake of change. Is there a list of issues and desiderata on this somewhere?