[Request]Reduce RNG influence

General feedback and discussion of the game.

Moderators: Forum Moderators, Developers

Yogibear
Retired Developer
Posts: 1086
Joined: September 16th, 2005, 5:44 am
Location: Hamburg, Germany

[Request]Reduce RNG influence

Post by Yogibear »

I really wonder why nobody has brought this up yet. Are you afraid? :P :wink:

I'd normally have someone else set this up, but for this special feature there are already a couple of (working) solutions. So instead of having to start from the beginning, we already have something to work with.

Nevertheless, i would like to collect some opinions on this. I prepared a wiki page for it which is still under construction which does a short summary. I am still collecting stuff so don't wonder if your idea / implementation is not listed yet. Also for what is listed, there might be information missing (any help is welcome :wink: ).

http://wiki.wesnoth.org/WesnothExperimental_Feature_RNG

Please let me know what you think about it.
Smart persons learn out of their mistakes, wise persons learn out of others mistakes!

User avatar
Zerovirus
Art Contributor
Posts: 1693
Joined: July 8th, 2009, 4:51 pm

Re: [Request]Reduce RNG influence

Post by Zerovirus »

To answer your question, at least on my part, it's not so much as a lack of drive on such an infamous proposal as an honest dislike of this idea. I've imagined Wesnoth without the luck, and frankly, I'd find it much more boring.

User avatar
Zarel
Posts: 700
Joined: July 15th, 2009, 8:24 am
Location: Minnesota, USA
Contact:

Re: [Request]Reduce RNG influence

Post by Zarel »

I have an idea! What we do is, we distribute damage in a Gaussian distribution!

Wait, we already do that? Okay, I'm out of ideas. :P

Perhaps we could just make it so every unit will hit at least once.
Proud creator of the :whistle: smiley | I prefer the CC-0 license.

User avatar
PeterPorty
Translator
Posts: 310
Joined: January 12th, 2010, 2:25 am
Location: Chair, In-Front-Of-Computer

Re: [Request]Reduce RNG influence

Post by PeterPorty »

OR... we can make every unit have 1 attack that always hit, all units having the same attack power, all units having the same move points and hitpoints, remove traits, remove abilitties... we (by "we" I mean the developers) could actually remove every bit of strategy in the game... and... will that make it a better game, will it be more fun to play? nope, it won't.
"The real world is for people who can't imagine anything better."

User avatar
Gambit
Loose Screw
Posts: 3266
Joined: August 13th, 2008, 3:00 pm
Location: Dynamica
Contact:

Re: [Request]Reduce RNG influence

Post by Gambit »

Let's keep this on task gentlemen.
This is the one corner of the forums where you are expected to [try to] discuss luck with a straight face. Try really hard.

Yogibear
Retired Developer
Posts: 1086
Joined: September 16th, 2005, 5:44 am
Location: Hamburg, Germany

Re: [Request]Reduce RNG influence

Post by Yogibear »

Zarel wrote:I have an idea! What we do is, we distribute damage in a Gaussian distribution!

Wait, we already do that? Okay, I'm out of ideas. :P
I know that you are trying to be sarcastic, but i will try to take this seriously because it illustrates a common misunderstanding of wesnoth's rng game mechanics.

1. There is no (statistical) flaw in the RNG
From a statistical point of view (Gaussian distribution and stuff), the RNG is perfectly alright. Several suggestions have been made to switch to other RNG-algorithms, but they won't change what is being felt wrong. Because statistics is not relevant here, the influence on gameplay is what matters.

2. Hits is not the same as damage
If i have a wose and a shaman and the wose never hits and the shaman always hits, i can balance hits perfectly even for every turn. Because what is influenced by the RNG is hit or no hit. But again, for gameplay it is very different. Because what matters much more for gameplay is damage, not hits. That is essentially why EV counts damage and not number of hits.

So what we are looking for is a distribution of damage, that prevents too large deviations.


And btw, i totally agree with Gambit. This is for serious working on stuff. I have made this fork to be able to work in peace and not to argue with people who don't want to tolerate other opinions. If you can't contribute to that please stay off.
Smart persons learn out of their mistakes, wise persons learn out of others mistakes!

User avatar
krotop
2009 Map Contest Winner
Posts: 433
Joined: June 8th, 2006, 3:05 pm
Location: Bordeaux, France

Re: [Request]Reduce RNG influence

Post by krotop »

I suggested few years ago a different alternative in one of those long, exhausting-for-devs luck thread.

In principle, it was supposed to be a parapet for statistically-unlikely long series of misses or hits to stop.

In practice, each unit is granted a 'karma' value, which would be an additional characteristic of the unit (like the HP, XP, or MP), with a starting amount even for every unit type. This karma increases each time the unit misses, and decreases each time the unit hits. When the karma is reaching the upper limit value, the next strike is automatically granted to compensate for all the misses the unit had before. Inversely, the unit automatically fails the next strike when the karma reaches the down limit.
Similarly, I thought that a 2nd karma value could be added for the defending unit, granting an auto-dodge if the unit is getting hit more than statistically appropriate, or gets an auto-hit if it goes too deep into the ninja path.

[slightly off-topic]
I don't know for sure if this forum is meant to gather different alternatives or if we should instead concentrate on discussing the ideas that got experienced already.

At the time I proposed this, I was completely lacking the skill to make the idea come into flesh. But I now believe to be able to code that with WML, it's just that I'm already into more projects than I can bear, so it's prolly out of question for me to launch a playable thing before this summer, in the case this idea is actually seductive.
[/slightly off-topic]



EDIT : I did not insist much on why I'm promoting this idea.
As a player of wesnoth I do enjoy a lot the way the damage is distributed, the fact we have to plan taking account when things go wrong, or the contrary let ourselves an additional option if the most favorable outcome of a fight occurs. I have no real problem to accept that a 50/50 or even more favorable situations goes wrong several times in a row. It's the game and when I do gambles I quite accept the price.

But at a certain point, when things go worst than what my mind can anticipate, for me or the opponent, I feel the game to be unfair with one of us and fun gives room to frustration. This situation of frustration occurs for me when me or my opponent gets wild series of hits or misses, series that have below 5-10% chance to happen, like a 0/3 with a mage when igniting a push, 2 grunts missing all hits on a unit with a low 40% defense, or a spearman receiving 5/6 hits in a village. That parapet system avoids this to occur while letting a large room to the luck factor.

Max
Posts: 1449
Joined: April 13th, 2008, 12:41 am

Re: [Request]Reduce RNG influence

Post by Max »

Yogibear wrote:So what we are looking for is a distribution of damage, that prevents too large deviations
here's something that might fit: http://gna.org/patch/?func=detailitem&item_id=1086
shouldn't be too hard to add as an additional option.

i actually like this idea - it reduces streaks of bad/good luck without changing game mechanics.

tsr
Posts: 790
Joined: May 24th, 2006, 1:05 pm

Re: [Request]Reduce RNG influence

Post by tsr »

OT? (feel free to split if deemed necessary):
Spoiler:

User avatar
Zarel
Posts: 700
Joined: July 15th, 2009, 8:24 am
Location: Minnesota, USA
Contact:

Re: [Request]Reduce RNG influence

Post by Zarel »

krotop wrote:...each unit is granted a 'karma' value...
A "karma" system (krotop's idea, Max's patch 1086) has been thought of before. It's not a good idea, because it means each attack isn't statistically independent. Which is terrible, because 1. it makes probability calculations ridiculously complex, which runs counter to Wesnoth's philosophy that they should be simple. It also means that probability exploitation becomes even worse - wait for a unit to accrue high karma, then reserve it for situations where you really need a hit to land.

Plus, it doesn't even deal with the core problem: Localized luck surges are the source of the biggest complaint. All a karma system does is ensure you have as much good luck as bad luck. On "average" - every game ends up with everyone having roughly equal amounts of good luck and bad luck - a karma system won't change that much. The problem is that you got the bad luck when it mattered, and the good luck where it didn't - and a karma system won't change that much, either.

A karma system would make luck screw worse - now you're intentionally spending high karma where luck matters and low karma where it doesn't - now you still have luck, but it's the luck of "will I have a unit with high karma when I need one?" - in other words, now, not only do you have luck, but you have luck that affects your luck - which doesn't do much to decrease the influence of luck, incidentally.

These are three reasons why nearly every luck proposal has been rejected. I'd prefer to search for a different solution. To that end, I'd prefer "luck should not affect luck" to be a ground rule. KISS and all that.
PeterPorty wrote:OR... we can make every unit have 1 attack that always hit, all units having the same attack power, all units having the same move points and hitpoints, remove traits, remove abilitties... we (by "we" I mean the developers) could actually remove every bit of strategy in the game... and... will that make it a better game, will it be more fun to play? nope, it won't.
You do realize that you could say that about any proposal with the intended goal of "[Request]Reduce RNG influence", right? Don't construct straw men.

My proposal was serious.

It would screw with the randomness system, though. Lower number of attacks would mean more reliability rather than less. Would mean HI would do a lot of damage to an elusivefoot unit, which doesn't seem right.

Or how about... grazing? When an attack misses, it does 1 damage instead of 0. Maybe 2. Would keep the same balance of "more attacks = more reliable"
Proud creator of the :whistle: smiley | I prefer the CC-0 license.

User avatar
krotop
2009 Map Contest Winner
Posts: 433
Joined: June 8th, 2006, 3:05 pm
Location: Bordeaux, France

Re: [Request]Reduce RNG influence

Post by krotop »

Zarel wrote:A "karma" system (krotop's idea, Max's patch 1086) has been thought of before
In case this isn't quite clear, the patch from akuu is also a 'parapet' system as in it avoids streaks of the RNG, but still different from the idea I proposed. akku is making a global modification, while I'd like to modify locally and independently for each unit on the battlefield. I prefer to insist because I don't find the difference to be obvious, while it is actually matterful.
Zarel wrote:it makes probability calculations ridiculously complex, which runs counter to Wesnoth's philosophy that they should be simple
That complexity is subjectively evaluated, in my eyes you are quite exagerating it, not that I would pretend to be objective either. The difference with the current system is only that casually an outcome is determined regardless of RNG. Appart from this special dice roll, which should not occur more than once every 2 or 3 turns in a duel from what I can tell of my experience of the game (except in the case of berzerker special fights), the anticipation of the outcomes remains the same.
Zarel wrote:It also means that probability exploitation becomes even worse - wait for a unit to accrue high karma, then reserve it for situations where you really need a hit to land.
You can't easily decide to wait for getting karma, it is not something you have control over, RNG decides that for you. Moreover, a unit having high karma means that it failed a lot previously, it is not an unfair abuse to tab on 1 guaranteed hit for all the times you missed before. If I had to choose, as a player, I would want poor karma and no guaranteed hit, meaning that my unit got pretty lucky before, than the opposite.
Zarel wrote:it doesn't even deal with the core problem: Localized luck surges are the source of the biggest complaint.
I don't know if I interprete the localized luck surges correctly. From what I saw on the forum, the complaints are generally that people see 'crazy' things such as having one's unit take many hits despite on good defense terrain, or miss many times in a row with a good cth. I don't think I have ever seen someone ranting on the forum that they miss with the high damage dealer, and succeeds the unimportant hits, although this is a recuring argument when people analyse what should be perfectly even luck. So I'm tabbing on the fact localized luck surges means a local streak of good/bad RNG rolls while it is statistically even if you take into account the next and previous RNG rolls. And if I'm interpreting correctly, then that 'unit parapet' system exactly aims at solving this problem (I'm changing a bit the name, because I recall 'karma' to be used for other systems already, and fear a confusion)
Zarel wrote:The problem is that you got the bad luck when it mattered, and the good luck where it didn't - and a karma system won't change that much, either.
That is quite true, that if you attack a unit and made 0/2 with your wose but 2/2 with the shaman it won't change. This is not the sort of things that frustrates me, or what I try to correct, because I accept that locally the wose may totally miss and either plan a failure case, or accept the consequences of the failure. What annoys me, and what I try to correct, is to do 0/2 with the wose and 0/2 with the shaman and miss again and again with the next units, which the parapet system can avoid.
Zarel wrote:now you're intentionally spending high karma where luck matters and low karma where it doesn't - now you still have luck, but it's the luck of "will I have a unit with high karma when I need one?"
My point of view is rather that you don't want to have high karma, it is only a small compensation for the bad luck you had. You can take advantage of it for guaranting an important hit, if you have the opportunity to at the right time, but I consider this advantage to be fair considering the precedent counter-performances of the unit. Inversely, having low karma gives the unit a temporary disadvantage, and maybe not at the best time for you, but it is a punishment for it to have been over-lucky.


Some of the points we are discussing on are quite subjective Zarel, so I believe it'll be as hard for me to convince as for you to convince me. Those points are : (feel free to correct me)
- such 'unit parapet' solution is easy/complex
- the main problem is 'the long series of luck streak'/'the fact you get hits when you don't need, and miss when it's important'. For this point, from a perspective where we analyse how we can make the luck system perfectly fair, I agree with you. My perspective is to try to correct my own frustrations, so that's very subjective indeed, and in that regard I want to correct only the luck streaks.


@tsr: since you seem motivated enough to do the hard work, the least I can do is to thank you and give you the informations you're asking for :) Probably pms would be the most appropriate to discuss those technical things.
Don't trust me, I'm just average player.
***
Game feedback for the Nightmares of Meloen
Art feedback by mystic x the unknown

User avatar
artisticdude
Moderator Emeritus
Posts: 2424
Joined: December 15th, 2009, 12:37 pm
Location: Somewhere in the middle of everything

Re: [Request]Reduce RNG influence

Post by artisticdude »

Zerovirus wrote:I've imagined Wesnoth without the luck, and frankly, I'd find it much more boring.
Same here. There would be little or no randomness to the strategy at all.
"I'm never wrong. One time I thought I was wrong, but I was mistaken."

Sangel
Moderator Emeritus
Posts: 2232
Joined: March 26th, 2004, 10:58 pm
Location: New York, New York

Re: [Request]Reduce RNG influence

Post by Sangel »

artisticdude wrote:
Zerovirus wrote:I've imagined Wesnoth without the luck, and frankly, I'd find it much more boring.
Same here. There would be little or no randomness to the strategy at all.
Be that as it may, one driving force behind the creation of the experimental fork is to provide a place to try out different approaches to luck in Wesnoth. So let's not going raining on this parade. Instead, let's consider all the different approaches which can be taken. The two major approaches would of course be:
"Luckless" Wesnoth, where everything is strictly deterministic,
and
"Reduced Luck" Wesnoth, where luck is present, but handled differently. But there are many different ways to implement "Reduced Luck" Wesnoth, which is what people are discussing in this thread.

Anyway, to get back to the current conversation - how do people feel about multiplying Unit HP by 10 across the board? [Along with accompanying changes to Damage/Hit, etc]. That would give more room to implement reduced-luck effects such as "grazing" (reduced damage hits).
"Pure logic is the ruin of the spirit." - Antoine de Saint-Exupéry

User avatar
Gambit
Loose Screw
Posts: 3266
Joined: August 13th, 2008, 3:00 pm
Location: Dynamica
Contact:

Re: [Request]Reduce RNG influence

Post by Gambit »

Also; remember that fabi has already put a slider bar (kind of like the exp slider) in the fork that multiplies HP and strikes by values 1-3. This turns defense into "number of strikes to hit" instead of "chance to hit" with only a little variance.

Sangel
Moderator Emeritus
Posts: 2232
Joined: March 26th, 2004, 10:58 pm
Location: New York, New York

Re: [Request]Reduce RNG influence

Post by Sangel »

Gambit wrote:Also; remember that fabi has already put a slider bar (kind of like the exp slider) in the fork that multiplies HP and strikes by values 1-3. This turns defense into "number of strikes to hit" instead of "chance to hit" with only a little variance.
Yes, and that's one approach. Increasing the number of strikes does reduce the influence of luck, but it also increases the length of combat, as well as impacting other game mechanics (such as poison and slow). So, it's one potential solution, but may or may not be as various other proposals. Experimentation will tell. :geek:
"Pure logic is the ruin of the spirit." - Antoine de Saint-Exupéry

Post Reply