The first problem you will encounter when researching difficulty curves
is how they are used exchangeably with (plain) difficulty.
For example:
Youtube title "20 Recent Games With The HARSHEST DIFFICULTY CURVES"
Article title "Best Games With High Difficulty Curves"
Youtube title "3 Difficulty Curve Mistakes I Made In Game Development "
where the authors really just meant "hard games".
Then, the graphs come:
â² inc dec flat a^b wall
d â . â. â â : â
â .' â '. â...... â : â :
â.:' â ':. â â..' â..:
tⶠâ¼ââââââ â¼ââââââ â¼ââââââ â¼ââââââ â¼ââââââ
> Difficulty(t)
> where t is a snapshot of the game, commonly abstracted away as time for simplicity.
And even tho it looks like information from afar,
all of this is meaningless.
The minute you present this to anyone with critical thinking skills,
they will get confused as hell:
https://www.youtube.com/watch?v=vWM_X1pC3oE&t=36752s
One problem is that we lack a definition for difficulty, it could be either:
1. required resources
2. required skill
---
The first definition is rarely spelled out explicitly as such,
but is the meaning derived from the idea that games should contain power-scaling.
This is what you will find articles in part describing implicitly,
where having a curve is a good thing.
Lets plot it:
Dark Souls 3 Witcher 3
â² â² :
â â .'
â â :
â .. â :
â..--'' â.'
â¼âââââââⶠâ¼ââââââââ¶
Well, we got curves, but there is very little we can tell from this.
You would think, that this atleast indicates how a low level character
can kill high level enemies, however your brain is meta-gaming, actually.
Given that you do not know anything about the games,
you would not know that there is progression systems involved.
To hammer home the point further,
consider the difficulty curve of my imaginary game:
Curve Stomper
â²
â .----
â :
â :
â.--'
â¼ââââââââ¶
Seemingly, it contains, what would be referred to as a "brick wall",
however, the player actually gains the "Fuck you, I win gun" right beforehand,
so its a nothing burger.
If you noticed, I keep circling back to the notion of progression.
Fact of the matter is, the above definition *is* useful,
given that there is no character progression,
then the required skill level follows the challenge axes
and we can actually use our graph.
Ok, so, this thing is clearly not what we want, because its not generally expressive.
Still, it doesn't stop people from using it:
Anatomy of game design - Tom Smith
https://davetech.davetech.co.uk/difficultycurves
https://www.youtube.com/watch?v=GCQiktmq9T0
---
Using the second definition sounds more potent right off the bat,
but then we are faced with another issue:
How do we measure skill?
So, instead, lets use something in a similar spirit, but that we can measure.
"Imposed challenge" happens to be easily quantifiable.
For example -with a bit of telemetry- we could record each attempt at every level
and a boolean outcome.
While at it,
we should also grab where on the level the players fail (think MarioMaker),
and on which levels they end their sessions.
This later, I will call the retention data.
Lets replot:
Dark Souls 3 Witcher 3
â² â²
â Fail â
â rate â
â^^^^^^^ â
â â.......
â¼âââââââⶠâ¼ââââââââ¶
Bosses Quests
I mean, I can infer intel from that, that's good.
What if I look at my own game again?
Curve Stomper 2
â²
â A C
â.'. ....'. .
â 'B 'D
â¼âââââââââââââ¶
Not very smooth, lets review it:
A:
The difficulty spikes.
Apparently, my players are twice as likely to get caught here
by the CIA than elsewhere.
Consulting my retention data, I can tell that people are rage quitting.
I should lower the difficulty.
B:
The difficulty drops.
I think the few people that died here, did it intentionally even? Huh.
Again, my retention data shows that people are getting bored
and closing the game.
I must add stimuli.
C:
The difficulty spikes.
I will have to lower the difficulty, yadda, yadda.
Wait, no, people are not quitting!
They appear to like the challenge.
Thinking about it, it plays well with the story pacing.
I won't be touching this one.
D:
The difficulty drops.
Again, my retention is high.
Now, it could be that they are still high on C,
or perhaps they really need room to breath.
A common pacing issue is having constant high stakes that tire people out.
I think I will experiment with inserting more, somewhat easy sections,
to see whether that improves the over all retention.
I will also swap D with a normal difficulty level to see how that influences stuff.
Ok, but where does all this leave us?
I am looking at flat lines, and I'm trying to make them the right amount of scribbly.
What happened to the nice cool bell-curve subsections?
I mean, if I were to implement the increasing graph I was shown at the start,
that would mean that the start of my game is boringly easy
and the end is buttfuckingly-hard.
Yet, some insist that even if players dont enjoy it, its a good thing:
https://www.moddb.com/tutorials/reflections-on-difficulty-pacing-in-puzzledorf
---
Now, I would like to tackle the "time" component.
As stated before, its a simplification.
Our game could have levels, bosses, quests or other means of progressing.
We dont know, but generally they will be sequenced in some way,
therefor time suffices to what is some arbitrary game state we find significant.
For a concrete game, we *could* be better of specifying whatever is appropriate.
What we shouldn't do however, is whatever the fuck Tom Smith thinks he is doing
in Anatomy of Game Design page 263:
> Giving players more agency gives players the ability
> to adjust this curve on their own, which is a powerful way to make
> sure the difficulty curve adjusts to each playerâs needs.
= Grant Theft Auto 3 Difficulty curve =
â ooooooo ^^^^ oooo
âoooooo ^oooo^ ooooooo
â ^ ....
â^^^ .... ^^^ ....^^ ... ^ .
â.... ....... ...
â
ââââââââââââââââââââââââââââââââ
> This form of difficulty tracking is mostly intended for video games where
> there is a long sequence of single-player content for the player to consume
> in a linear fashion. GTA3 breaks this rule by not following a linear
> sequence.
He is perfectly correct on the first part.
However, he should have never drawn whatever eye cancer that is.
He took time literally, making his data into a sausage-blob.
Instead, one is meant to break it down into missions.
Then maybe sort it by average playtime of entering
and -with rigorous cross checking of retention data-
identify pacing issues.
---
What about multiplayer games?
Team Fortress 2 would have a curve.
What it most certainly wouldn't have is this thing:
â² skill
â ....
â :
â...: experience
â¼ââââââââ¶
Which is proof that r*dditors are incapable of reading graphs,
and would mean that things click suddenly and you ascend to godhood.
Anyways, the curve is fine:
yeah, I got backstabbed for 18th time,
but I keep oneshotting the enemy heavy so I'm still having fun.
That is to say, large teams and short kill times makes it possible
to focus on frags within my skill level, making the difficulty feel flat(ish).
Track Mania speed running would have an exponential curve,
but I only need to beat my own record to feel accomplished.
Counter Strike would have a curve, but the match making always places me
with players of similar skill level.
Wait, wait, wait? Similar skill level?
Does that mean we can measure skill?
---
We could actually, there are a number of rating/ranking systems.
For example, we could employ ELO -because its simple-
or Glicko-2 -because its not.
Just a note on ELO in specific:
its used commonly and indiscriminately,
even tho PvE contradicts the assumptions it was forged under;
this is fine however, ratings will still converge to values that are meaningful to us.
The thing these systems require -besides magick constants to choose-
are success data.
Meaning our recorded telemetry is perfect input.
Huh, how lucky that I so happened to write my sections in this specific order!
But now we are forgetting about resources.
Our first attempt might have been naive,
but that doesnt mean we should write off resources.
Remember: it helps no one to be reductive.
To underline my concern, take my experience with Dark Souls 3.
I farmed to level up my character so I would have an easier time with bosses.
Put otherwise, I made up for my lack of skill with resources.
Inâconclusionâresourcesâandâskillâshouldâbeâconsideredâtogether.
After we find some kind of estimate to convert between resources and skill,
we can replot yet again:
Imposed challenge
â²
â '
â . .....
â . : === :
â . ; o o :
â . --- ; /_\ :
â .' l '. ;: '_' :;
..... â..--'' '. '. ;:...:;
: === : â¼ââââââââââââââⶠ'. '._/ ^^^^^ \
:-@---@-; skill+resources '. |
: ; \ |
:../ \..; | |
;:.\ /.:; | |
;:...:; | |
^^^^^ | |
Its the graph from the beginning! ö
Ok, can we read this graph to balance our campaign?
Hell no, for one its 3 dimensional.
There is a time component hidden and this representation *assumes* that
the player gets better and richer over time.
But look at it, its majestic!
If some retard showed this in my face without context,
I would think it describes how someone in a better position is punished more and more
Especially if they had labeled the axis "difficulty" and "skills/resources":
> https://supersonic.com/learn/blog/difficulty-curves/
But what kind of sadistic monster would consider such game design?
Huh, I have an idea.
What if I made a rogue-like dungeon-crawler,
I would have every 3rd level be a filter, regardless if the player is doing well.
Or I could create a colony builder where the size of raids is dependent on your wealth.
I'm not all evil either, imagine a coop survival horror and some poor noob fella
lost in a high level area (because that's where his buddies invited him)
who is down bad on resources.
I would throw him a -loot- bone, maybe even adjust the enemy AI slightly,
to make him feel like its his lucky day.
We just rediscovered "Dynamic Difficulty Adjustment" aka DDA.
---
Some closing thoughts.
I believe that passive aggressiveness is low T, therefor allow me to state this plainly:
**The available mainstream resources were trash.**
The jeet-articles, slop-videos and amazon-books contain poorly thought out,
conflicting and hand-waving zodiac-sign-salesman speech,
stretched out to the 10 minute mark,
only to tell you: "too much is precisely the amount that's bad for you".
Regarding the graphs in particular, I sincerely believe that the journo scum
is at the end of a terribly long game of telephone where they started with DDA.
I will also note, at no point did I see realistic imposed-difficulty/time graphs
and their subsequent analysis.
The gamedev-gurus do not care about numbers
and no one serious appears to have published their experiences.
I imagine its because it would require excessive playtesting
to record the required information, such as an open-beta,
but AAA doesnt do open-betas.
However, even with very few playtests,
its significantly more reliable than Steam Reviews.
The closest I found was "PMID: 34355222", which is great,
but deals with different questions.