The stealth meta

Started by Greve, June 13, 2020, 06:08:01 AM

Reply to Harmless's tangent at the end.

I was bitch slapped for talking about code and math, but the issue here is clearly about the code and math as described by OP.

If the code today is even 10% like the code I posted and was shamed into removing, the problem still exists as described.

If you aren't afraid of math and code -- what this discussion is about -- here is some analysis, x axis is your character's current scan skill level and the Y axis is your character's skill as modified by penalties and bonuses.

In the old code, your scan score gets hit by pretty brutal penalties, the worst I saw was `skill /= 4`, these is also a case where it is halved. Similarly, there is this huge bonus if you have the same stealth skill as the opponent, up to a maximum of 20 points or 20% maximum. To model this in a crude way, we see a spread of penalties and bonus like this (green is bonus, blue is no bonus, red is penalty):

In the way the old code worked as modeled here, if you had master scan at 100 and were suffering from just one of the possible penalties, you can only see people who have a hide skill of 30 or lower. That is a huge #fail for someone with master scan. Similarly, if you scan caps out at 60, you can only see people with hide 15 or lower under just one of multiple possible penalties. The max bonus of +12 in this scenario can't really outweigh these huge -45 penalties, etc. This fully explains the "Scan either works or it doesn't, and I've only seen it work at master" phenomenon others anecdotally reported.

Let's imagine they've made changes to make these penalties less than half as extreme than they are now:

Here we see the same problem can manifest. If you have middling to high scan skill, the penalties are still huge and would lead to the situation we all know and describe. It would take pretty significant drift from the old code to get even to this spread, and if I had to guess as to which of the models we're closest to today it's something like this.

Lastly, when most of us play RPGs, we're used to seeing bonuses and penalties like "+2% to hide" or "-3% to scan". To model a system like this with a max bonus of 10% and a max penalty of 20%, this is the distribution:

This is the spread we're most used to seeing because it is the most fair, and logically a distribution of bonuses and penalities like this would likely address most concerns. Example, if you have a scan skill of 80, but are suffering a 20% penalty, you still have a chance of seeing someone with a stealth skill of only 60. This feels fair, in contrast to our system today, where even at skill 80 you face penalties that prevent you from seeing anyone.

While recent numbers and data are hidden, and hidden for good reason, old data points indicates how egregious this problem used to be. Because the previous state of these bonuses were "totally fucking extreme," I am willing to bet the state of these bonuses today are "less fucking extreme but still extreme." Given that my data is old, if my data is even 50% like or 25% like the code we have today, we're seeing some out of proportion bonuses and penalties. Modeling these bonuses and penalties leads to situations people describe where scan is broken at low levels before it will suddenly seem to work.

Anyway, this is just food for thought for staff and players. We're talking about an issue of a "skill being broken," which is fundamentally a question about code and math. I did some research and ran some models, my recommendation is if bonuses and penalties greater than +/- 20% still exist [and in the old code they were everywhere], they should probs be tweaked.

Quote from: Harmless on July 11, 2020, 09:41:00 AM
I am in fact asking for a reduction in how much one sees if they succeed in looking at a hidden person. Instead of getting the full mdesc, maybe just a shape, the cloak, the mask, the build. One piece of information only. It would be a nerf to the LOOK command.

It would be best counterbalanced by a boost to scan, but then the thief would be less likely to lose their life because all the detecting/scanning party got is a vague sense of a creeper about at that time wearing this or that. The thief could, with adequate perception themselves, realize they are spotted, escape, change costume, then return for more. There would be more interaction because near misses would be a more frequent occurence instead of all or nothing detection as it is now. Thieves would be free to take more risks, and so would the non thieves.

I am sorry I didn't spell it all out clearly, because it is a years old idea. It was more thoroughly discussed on a mechanical level in prior threads, such as one I started recently.

In favor of something like this if we buff scan per my proposal. Scan is already abused in annoying ways, like players having MUD client GUIs that spam scan, or players that do `scan;l n; l n;l n...` and `l n` 2000 more times until they manage to see something. While Harmless's proposal wouldn't fix that abuse, anything that we can conceive of to [1] fix the issue of scan not working at lower levels ALONG WITH[2] fix the abuse of scan with new features like this or the `point/reveal` feature, that would be ideal.
ARMAGEDDON SKILL PICKER THING: https://tristearmageddon.github.io/arma-guild-picker/
message me if something there needs an update.

eh, the green line should start with the formula `x + 20` instead of `x * 1.20`, but my argument is mostly concerned with the penalties so it doesn't need to be edited.
ARMAGEDDON SKILL PICKER THING: https://tristearmageddon.github.io/arma-guild-picker/
message me if something there needs an update.

Yup. In the thread I linked I suggested that we make scan entirely passive, such that it continually scans for you instead of relying on you manually typing LOOK to activate it. By the end of the 12 pages of that thread I think there were several good suggestions to help fix the pvp stealth environment. Difficult to code maybe but lots of options within to fix a kind of dead system.
Useful tips: Commands |  |Storytelling:  1  2

Quote from: Alesan on July 10, 2020, 09:18:12 PM
Nobody sits in taverns anymore because if you're seen sitting in a tavern for any longer than absolutely necessary, you become a "tavern-sitter" which is an inferior type of player than others who are actually doing "meaningful" things in the game.

+ 1

Thank you so much for this post. It really made my day.  In fact, it made my week!   ;D   I loved it for the truth it expresses but even more for the wonderfully sardonic tone.

A scale of depth I feel is with in reason, is in simple factors such ass room size, level of capacity in the room, number of people in the room. Would like to see a coded depth added to each door in a room.

-Your PC is in a three person tent alone, no fly goes unnoticed if they are sitting at the 'door'.
   = Resting or sleeping would be done in the middle of the tent. Can still sit on mats/rugs at the door.

-Standing at any 'door' would give your Apprentice Joe Watchmen a solid view of passers by.
   =Make anyone standing at the door a easy shot for ranged weapons. Unless shield in hand.

- Huge room capacity, room is %90 at capacity, a armored half-giant may slip-in if they wanted to.
  =Will always be easy to watch a door way for a HG. PC might learn to shut doors more often.

Where are all these numbers coming from?