[MUD-Dev] DGN: Do levels suck?

Koster, Raph rkoster at soe.sony.com
Fri Dec 23 07:14:35 New Zealand Daylight Time 2005

Reposted from my blog (http://www.raphkoster.com/?p=225)

I've said in the past
<http://www.raphkoster.com/gaming/postmortem.shtml> that levels

A few things that have been written about lately, however, prompt me
to dig a little bit more at that long-held tenet of mine, because
while constant self-doubt is debilitating (trust me), it also often
opens up surprising new doors.

My objection to levels in the past has been based around the

  - The way in which they pull people apart

  - The psychological impact of constantly pushing a lever for
  another pellet

  - The huge content multiplier they impose

  - The mudflation arms race they create

On the other hand, it cannot be doubted that levels provide a
powerful incentive. Why do we have them? What good are they? And do
they indeed suck?


A brief history of levels

Levels were pretty much ripped wholesale from Dungeons and Dragons.
Richard Bartle has commented that he put them into MUD1 because they
provided clear regular feedback on advancement. And that they do.

It's worth looking at some of the things that changed as they came
into MUDs, though. I don't know how many levels there were in MUD1,
but in D&D there weren't very many. The very notion of having 70
levels (and in the case of many text muds, "remorting," and in the
case of EQ, "Alternate Advancement") is silly when looked at through
a D&D-circa-late-70s lens. Now, I got to D&D late, with the Dungeons
and Dragons Basic Set in the red box, and with the stack of AD&D
books I still have, a mix of yellow spined and original editions of
the 1st edition of the game rules. But a level 20 was nothing to
sneeze at in those games. It was extraordinary, in fact.

D&D was notable in its advancement model for a couple of reasons:

  - You got experience for anything the Dungeon Master thought
  deserved recognition

  - You never, ever played with a widely disparate group of

These two elements were fundamental to how the game worked. In the
translation to computers, however, the first was lost; going from
analog to digital, from human to mechanical, XP became nothing more
than a weighted count of creatures killed. This was a dramatic
change in the nature of what we call "roleplaying games," and the
computer gaming industry has been fighting it ever since. A D&D game
which was purely focused on advancement was derisively termed a
"Monty Haul" game, and its players "rollplayers" (a pun which has
surely been reinvented thousands of times by bored geeks).

I wish I could say unambiguously that the center of the pen and
paper roleplay session was its narrative, not its numbers. But that
Gamist/Narrativist tension (not to mention Simulationist, which has
affected the design of a lot of pen and paper games) led to some
classic D&D adventures on both sides of the fence: tournament play
which was clearly Gamist in nature, pushing to beat a specific
challenge <http://www.acaeum.com/DDIndexes/ModPages/C1.html> (the
referenced module actually ran on a time limit, and if you didn't
get your party out in time, they died from poisoned air); versus
heavily Narrativist stuff like Ravenloft
<http://www.acaeum.com/DDIndexes/ModPages/I.html> , which was such a
popular story that it led to the creation of an entire
setting. Either way, however, one thing is clear: even in the Gamist
settings, you were not rewarded solely for killing.

In the early CRPGs, you often controlled a party, rather than a
single character, and your party was tightly coupled in level as you
progressed through the story. This commonality between pen and paper
and CRPGs, however, was lost with the transition to muds and then
massively multiplayer games: pen and paper games were geared towards
narrow level ranges. You generally did not have an adventure with a
level 1 and a level 10 character running around in the same world at
the same time.  You might have campaigns that were set up that way,
but not individual sessions. Modules that were sold or designed by
players were set up for tight narrow ranges of content drawn from a
larger pool of available content described in the reference books.

Higher level did not imply higher difficulty either. There was a
real effort to have the nature of game change as you advanced; the
rulesets added stuff like followers, baronetcies, and so on at the
higher levels of the game. But you could have a murderously
difficult adventure with newbie characters, and a cakewalk one with
high-level characters. Even though the amount of experience points
needed for a level advanced as you played, the experience grants
were scaled by storyline and play session, not by arbitrary number
of kills.

In addition, the gap in damage-dealing power between a low-level
character and a high-level one was not all that dramatic; a mage got
1d4 extra hit points per level, and even a fighter only got 1d8. I
call this the power differential between a newbie and a maxed out
character, as it scales the content that is required in the game.

Much of this changed with muds, and the changes have carried through
to the MMORPGs we play today.


The common characteristics of muds' use of levels are these:

  - Levels are earned with experience

  - The world holds characters of all available levels

  - More levels were added

  - The power differential grew dramatically even between the lower

  - Each level is harder to get than the previous one

The wrinkles here are many. For example, the ways to get experience
have changed over the years. A focus on killing things developed,
even though there had been in earlier games a collecting game (in
early AberMUDs, for example, you had to gather stuff and drop it in
the well to advance). In the early 90s, there was much emphasis on
the notion of "exploration XP" - giving XP for entering particular
rooms on the MUD, or tracking how many rooms a player had been
to. This sort of mechanic never fully supplemented the XP for
killing, but is sadly lacking in modern MMORPGs. "Quest XP" is also
a very common mechanic that has been minimally used overall,
although World of Warcraft makes heavy use of linearly directed
quests and you can advance quite well through questing.

Because of the codebases used in many of the most popular muds,
quests were difficult to implement, and having them was actually a
major selling point in the first few years of the 90s. But puzzles
and quests have a long tradition in the mud world, and they're
sometimes far more intricate than the stuff in the MMORPGs today,
generally thanks to the immense flexibility that text can bring to
the table. (If Legend wouldn't freak out, I'd post the walkthrough
of the Beowulf quest just to illustrate the point).

Despite the presence of quests, however, killing things became the
primary mode of interacting with a virtual world despite the wide
variety of possible interactions. The "XP run" was born... lacking
the play-session scale of pen and paper gaming, levels were defined
instead by using a baseline of "number of creatures to kill to get
the next level." Typically, like in D&D, the levels were given a
larger and larger required experience point cost, but without the
saving grace that rewards were scaled by the necessities of
storytelling. Instead, what was rewarded was repetition: "I need 20
rats to level up."

The second fascinating wrinkle is the way in which this had
distorting effects on the reward scale. Increasingly, games came to
treat level as implying a level of difficulty - or at least,
tedium. While the first few levels of a mud were notorious for being
difficult, the general design trend was towards offering bigger and
bigger rewards as you rose through the levels, and thus requiring
bigger and bigger enemies, often requiring bigger and bigger
groups. While this trend did not receive its apotheosis until the
days of "raids" in EverQuest, the seeds were clearly sown earlier:
you leveled because it got you better stuff so you could fight
bigger things that gave you better stuff that...

The result of this cycle is that more levels needed to be added to
the typical D&D style progression, because that retained players
more, offered more regular rewards (if you think about it, the
reward feedback given by a pen and paper game was actually pretty
sparse), and provided a direct point of comparison between
games. Because the power differential was being increased, more
levels needed to be added.

The long-term result is mudflation. Players reach the top of your
level ladder, but you need to keep them occupied, so you add more
levels, and with them, more powerful items to serve as rewards. But
then you have these powerful items trickling around the game
economy, so everything in the game gets a little bit easier. This
makes people level faster to the top, which then results in your
adding more levels...

Even aside from the classic mudflation effect, you also have what I
call database deflation
<http://www.raphkoster.com/gaming/databasedeflation.shtml> , which
is the devaluation and redundancy of your statically created data,
occurring simply by the fact that you added more levels, regardless
of whether there are players present or not. Any given monster or
obstacle can literally be evaluated as a % of the path needed to
reach the maximum level; by adding levels, you are adjusting the
percentage the monster is worth.

The upshot of all of this was three-fold:

  - a common practice of engaging in regular character wipes or
  equipment wipes. (!)

  - the invention of numerous systems to push players through the
  same content repeatedly without increasing levels. The best known
  of these was probably "remorting," which allowed players to take a
  maxxed out character and start it over as a different class, but
  with the same identity and gear. I've constantly been surprised
  that this hasn't been applied to the graphical games.

  - the invention of a whole host of systems to prevent players of
  disparate power from playing together.

The latter is important, because it gives rise to twinking, to level
limits on gear, to soulbinding items, to sidekicking and mentoring,
to PK level limits, to PK zones and safe areas, to group level
restrictions, and to the concept of level-limited geography.

That's quite a host of side effects. Power differentials between
levels are at the root of countless systems in modern
MMORPGs. Twinking exists because godlike characters can help
mouselike characters. Same for grouping level limits. Level limits
on gear exist so that swords from Valhalla don't fall into mouselike
hands. Soulbinding exists for the same reason. And PK zones and
level-limited geography exist so that godlike characters do not
crush mouselike ones.

Lastly, sidekicking and mentoring, which I believe were first seen
in City of Heroes - wow, what a brilliant hack! We'll allow people
to temporarily change level to get past all the barriers we just put
up because we included power differentiating levels in the first
place! It seriously is a genius solution (I mean that quite
honestly) but it also points out exactly how many undesirable side
effects have come about from levels over the years.

The upshot is that whereas in D&D levels were used to bring people
together, in MMOs today they are used to keep people apart. In a pen
and paper campaign, it was considered mere politeness to allow a
newcomer to skip to the level of the current adventure; this is
inconceivable in today's distortion of the system.

If anything, this little history just illustrates the ways in which
levels have changed over the years. It's important to realize that
most of these side effects didn't exist in the original D&D model
because it proceeded from different assumptions. Rather, they are
all adaptations caused by the use of the model in a very different

Typical level distribution:

In the end, one thing tends to remain constant: a graph of
population of characters at each level in your game database will
generally show something that looks a lot like what you see when you
hold up your left hand and try to make a V. If you have an uneven
rate of advancement from level to level, you will get a slightly
jaggier graph as players accumulate at the "hell levels," but
broadly speaking this graph always holds true. The start of the
graph is your influx of newbies, the downslope in the lower levels
is because of abandoned characters, and the spike at the end is
where everyone ends up. The middle levels tend to have a far far
lower population.

Seeing this graph, it's clear why adding content at the top and
causing mudflation is the typical path: it's what would satisfy your
customers.  But it has huge implications on content creation costs
and on the notion of user segmentation into "cozy worlds"
<http://www.zenofdesign.com/?pP0> .


Levels and feedback

The usual case for having levels is made on the basis of
feedback. Now, the very first thing we need to get out of the way
here is to clarify that we are discussing here what the MUD-Dev
mailing list terms "goal-oriented play." This is what Bettelheim
classifed as "games" rather than "play," and it's effectively the
dominant mode for most games designed today. It's not, however, the
only mode. Games such as Animal Crossing and online games such as
There demonstrate that free-form or low-pressure environments can
succeed and attract an audience.

But since it's the dominant mode, let's consider what it means. One
of the main things it means is that levels push towards cooperative
rather than competitive play. The reasons really require another
essay, but suffice it to say that disparate power levels are
incompatible with fun competitive play. In just about every
competitive game throughout history, players are given equal
footing. The very few asymmetric games are ones where the metrics
are not standard defeat, such as Fox and Geese, where the geese win
by entrapping the fox, but the fox wins by attrition of geese.

In the world of software-based games, of course, asymmetric games
are not only common, but are by far the most typical sort of
game. Just about every videogame you play is likely asymmetric in
its core mechanics. The player has different capabilities and
different statistical traits than the challenges they seek to

Asymmetry is what really opened the door to levels. In symmetric
games, it's simply not a likely design to choose. But in a
cooperative game (or a parallel game of competition, rather than one
of direct competition) where one wants to track relative progress of
multiple participants, it makes perfect sense, and hence the origin
of levels in pen and paper games.

The notion of levels as feedback is important here. Many have
claimed that contemporary MMO designers are consciously creating
Skinner Boxes: that in effect designers are making conscious use of
operant conditioning, most specifically with random reinforcement
schedules.  Frankly, I haven't even been in a design meeting where
that was discussed as a tactic to use, though I have frequently had
conversations after the fact where designers have evaluated their
designs and concluded that what they were doing had that effect.

Whether it's intentional or not, there's a host of powerful
psychology effects that levels as currently implemented give, and
it's not all about Skinner Boxes:

    - The aforementioned random reinforcement: you don't know
    exactly when you'll skill up, so you keep doing whatever gave
    you a little bit of reward

    - What Robert Cialdini might call "the commitment fallacy" -
    once you have a few, you figure you're in for the ride and may
    as well finish off the ladder. People don't tend to like leaving
    things half-done.

    - Another powerful tool of influence: social validation. Levels
    are publicly displayed, and serve as a significant social marker
    of status.  And humans are hardwired to seek status and

    - The "gated community" effect. It's been observed many many
    times that people want what they haven't got. Just as clubs will
    intentionally create lines outside a door to drive traffic, and
    just as it's a time-honored technique of retail and carnies to
    hire a claque of folks to make the business seem popular,
    exclusivity in online games is a powerful motivator. Levels
    effectively put content behind a velvet rope, which just makes
    us want to get inside.

    - Finally, one of the most compelling aspects of levels is the
    lure of power. Levels promise increments to a player's health,
    their damage per second, and so on. People like feeling more
    powerful - it's not social validation, it's the game system
    itself giving them validation.

Ironically, it's this last one that causes all the other systemic
problems in the game. You could have random reinforcement without
upping hit points. You could have exclusive clubs, gated content,
publicly displayed status, and a treadmill of ranks to climb,
without changing the power differential between levels.

The worst thing is that in many cases, it's a lie. Until the fairly
recent advent of flat level curves, the common practice was for each
level to require a bit more kills than the last. A typical way of
balancing level systems is to say, "Level 2 will involve killing 20
even matches, Level 3 will involve killing 21 even matches," and so
on. The XP value is then set for each level based on formula that
provides more XP for higher level mobs, but keeps to this
boundary. When done correctly, it then provides a fixed and
straightforward scalar factor you can use to reduce the amount of XP
granted for kills of creatures of lower level. The result is no
"hell levels," a very gradual increase in "grind" with fairly rapid
feedback at low levels, and (ironically) a net reduction in actual
player power against even matches. In terms of the levelling game,
an even match is worth less - to keep at the same pace of
advancement, a player is put at an increasing disadvantage.

Another common way in which players get weaker as they go up in
level is the difficulty of what is considered an even match. It is
not uncommon for the targets intended as an even match to be groups
of enemies (which means a force multiplier), to require groups to
tackle; to have absurdly out of scale hit points; to have special
attacks beyond the norm of the equivalent level player; and so
on. Of such things are raids born, forced grouping bred, and guilds

The "grindier" games are ones where this level curve and
accompanying power differential is more extreme; rather than a
linear increase in number of kills required, they may actually
involve an exponential rise.  Even where the level curve is fairly
flat, as in World of Warcraft, the use of increasing difficulty

The reason, of course, is not feedback - it's content.


Levels and content

Flatly, levels are a content multiplier.

Look at the dilemma faced by the level-based games which try to
minimize the grind by providing a flat advancement curve. A player
must engage in 20 kills of an even match to advance a level. The
next level, they must engage in 20 kills of an even match. The next
level, they must engage in 20 kills of an even match.

What the designer needs to change to make this fun is the definition
of an even match. Because games are about learning, the player must
be given an increasingly complex situation to handle. In a
level-based situation, the increased variables are generally the

    - The abilities the player can bring to bear on the problem:
    skills, spells, weapons, etc.

    - The tactical situation surrounding the problem: other enemies
    who might assist, the landscape, coordinating friends in your
    party, etc.

    - The abilities that the enemy brings to the challenge: special
    attacks, increased damage per second, etc.

    - The amount of correct choices the player must make to win the
    challenge (which is expressed by increasing the enemy's hit
    points by a factor larger than the increase in player's damage
    per second). You can think of this as directly analogous to
    making you have to remember more and more colors when playing

Compare to Tetris, where only one variable increases: speed. Or
compare to typical symmetric competitive games, where the sole
variable is the skill of the opponent in using the abilities they
have, and where tactical situations are emergent. This is almost an
embarrassment of changes to make. The fact that so many variables
are required speaks to the poverty of the basic combat model to
serve as an entertaining game in its own right.

If you think I am saying that typical RPG combat sucks as a game,
you're right.

So now the designer is obliged to create scenario after scenario
with all these variables. This is the process of creating
content. What exacerbates it, however, is those pesky power
differentials that levels generally imply.

In order to prevent players from doing the sensible thing and
maximizing their return on time invested by minimizing deaths and
maximizing predictable advancement (also known as "bottomfeeding"),
developers of level-based systems must create their content in
bands. It's typical to see that for a given player level, the
available dataset of challenges is +/- a few levels from the level
they are on. A level 10 character may be able to fight a level 6 for
minimal XP gain, and may be able to tackle as high as a level 15 and
get lucky. Everything else is out of reach either from a reward or a
feasibility standpoint.

This would be the point at which I suggest you read another old
thing I wrote. Go on, this will still be here when you get back.

OK, now, what you just read contained a number of points about how
these games are played:

    - Players will be playing content in parallel. You've got
    multiple users, so you need to provide available content for
    each of them.

    - You as a given player will have competition for content
    resources only within your rough level range.

    - You will need to provide content bands proportional to the
    amount of power differential between your highest and your
    lowest player.

    - Each content band must contain sufficient content to keep the
    player entertained, or they will term that level "boring."

    - Since a level band cannot offer significant statistical
    variation (by definition, since otherwise the content would be
    in another band), this means that the content variaitons within
    a band will likely have to be in the form of narrative, tactical
    situations, or something else - something that is not merely
    power differential.

    - Lots of games fail at providing this, which is why each
    content band consists of killing 5000 orcs or crafting 7000
    blaster barrels.

Now, the amount of content required is driven, in the end, by your
player population and distribution across levels. Time to pull out
that graph again...

Typical level distribution across levels:

What we see here is that in order to alleviate competition, you'll
need to provide a huge amount of content at the highest level band
in your game. The effort you went to in order to provide a lack of
competition to account for the initial surge of players moving
through the middle levels will become obsolete, as the simultaneous
population in the midlevels will drop over time. The single largest
wave of mid-level players you will ever have, most likely, is in the
first few months after launch. After that, you'll have something
like 50 times the "bandwidth" for mid-level players as you will
actually need.

This is a massive overspend. You can think of it this way: When the
initial population of players came into the game, it was a little
higher than the level of the red box. There was some attrition and
some slow levelers and some reaslly fast ones, but these distribute
along a bell curve. Then the bell curve moves through the levels
just like a wave.  The red box is the "high water mark" of this wave
of players moving through the levels. In order to provide a lack of
competition for resources throughout the leveling process, the
developer will have had to provide content that fills the volume
shown in the red box, so that the peak population of a level band
was always accomodated. But the mature playerbase's need is only the
area under the curve. Compare the area of the two spaces.

This is why there are vast echoingly empty adventuring spaces in
most mature MMOs. It's also why the pressure to add solely at the
top is so overwhelming: to reduce contention, you have to keep
adding variations.  Pretty soon, the only way to do that is to add
more levels, because you've exhausted all the other ways to provide
ongoing learning and therefore ongoing fun. Hello, mudflation.

Now you see why I call it "database deflation". Not only do you end
up rendering the expended time players have invested thus far worth
less as a proportion of their total advancement, you are also
letting the air out of your accumulated content. You are pushing
players through a learning process which renders each level band
less challenging for them. You are likely introducing new ranges to
the power differential in the game, often attached to items which
trickle down and effectively shrink the lower content bands, often
to nothing.

You also see why remorting into a different class, and "altoholism,"
are so common. They allow re-use of this content in a manner that
presents different puzzles by giving different abilities to the
player, thus rendering every problem somewhat fresh (not totally,
mind you). The more diverse the tactics available to classes, the
better this will work.

There is one very nice side effect of this, though. Player


Levels and cozy worlds

One of the things that works the best in level-based games is the
sense of camaraderie with players who are levelling at the same rate
as you are and started at around the same time. World of Warcraft
makes excellent use of this, as did EverQuest in the early days. In
a nutshell, the segmentation in player power caused by levels also
meant that a given zone was a cozy world: a welcoming, reasonably
sized community where most everyone knew each other. By isolating
players both geographically as well as with power differentials, a
"movable feast" community was created.

At the higher levels, this breaks down, of course, as the population
at max exceeds the capacity of any one place. But if you're not in a
guild at that point, you're unable to enjoy the content anyway, and
the guilds become the cozy worlds instead.

So it is that the greatest weakness of levels - the fact that they
prevent people from playing with one another - can also be their
greatest strength; arguably more powerful than any of the Skinner
Box sort of bits of psychology. Group identity is routinely cited by
players as the most powerful retention factor in online games.

The question is whether one needs levels to accomplish this. Let's
consider the factors that seem to go into creating a
success. Leaving aside the basic question of whether you have fun
gameplay at a core systems level, the things that have been listed
throughout this article are:

    - feedback for achievements

    - public status based on achievements

    - gated communities that require special status to enter

    - the lure of power based on significant achievements

    - regular changes or variation in the challenges undertaken
    within a given playstyle

    - cozy worlds created with players segmented based on when they
    entered the game and the rate at which they leveled; or
    self-selected by players

Bottom line: none of these need hit points to go up. None of these
need the traditional notion of levels as we know it, actually. Nor
do they need any of the other sorts of "levels-in-disguise" things
like skill trees, actually. Power can be satisfied with a number of
things, including collection mechanics, customization, and yes, even
actually increasing player power relative to challenges on a
separate axis from their comparison to other players. (A game where
as you rose through level, you levelled faster? Horrors.)

How cozier can our worlds get if we remove the artificial barriers
that the legacy of levels from a 30 year old game system has given
us? Can we satisfy those players who want the ding? I suspect the
answer is yes, but I'll leave the actual systems design to you. Most
systems people tend to propose leave out hitting the full set of
bullet items above.

So, my answer in the end? Levels don't suck in every way. There's
plenty of good stuff they bring to the table. But if we're smart, I
think we can have all that stuff without levels themselves.

MUD-Dev mailing list
MUD-Dev at kanga.nu

More information about the MUD-Dev mailing list