Sword of Mana: World Design

Recently, I was talking about how bad Sword of Mana is because it foreshadows scripted events. This time I would focus on another curious and fatal difference to predecessor Mystic Quest:

Click here to see the world of Mystic Quest in full resolution.

Mystic Quest is like the NES Zelda: it has no scrolling. The world is divided into separate screens. Sword of Mana uses scrolling to some extent. The “camera” follows the avatar and the world scrolls underneath the avatar as he is moving. Curiously enough Mystic Quests creates a more consistent and believable world. Let me explain the trick:

One important hint is that in Mystic Quest the screens do fit together like puzzle pieces. The player can go ahead and actually draw a map of the game. Strategy guides often even feature a fancy map created out of screenshots. When it all comes together, it is a piece of beauty (see above).

This is a part of the world of Secret of Mana. Note that this map also includes dungeons as they are difficult to separate from the overworld. Click here for full resolution (3MB)

Sword of Mana lacks this consistency. Although the world scrolls beneath the player, it is still divided into manageable chunks, presumably because of technical limitations or to simplify scenery changes. These chunks, or levels do fit together in some way but the result is not a complete map. There are gaps and holes. The whole task is pointless anyway. If you look closely you will realize that the levels of Sword of Mana are linear. Your path trough the world has barely any forks. Most places have only two connections to other places: an entrance and an exit. Navigation through such a world becomes trivial so the exact spatial layout is no longer important.

This made me think. It seems like Mystic Quest and Sword of Mana follow fundamentally different strategies also when the overall scheme of level design is concerned. This inspired me to create some kind of typology of meta level design, of game world design.

While writing the typology I’ve come up with a concept to measure the economy of game design. The idea is that everything you design in a game will be used or seen by a player at some point. A game is more economical if the game elements you design are used and seen more frequently. Thus, I introduce the UPP factor (Use Per Play). For example, a typical boss in a Final Fantasy game has a UPP factor of 1. It is encountered only once during a play session. After it has been conquered, it’s gone and you will never fight it again unless you restart the game. Of course, a normal enemy in a Final Fantasy game would have a much higher UPP factor because it is being re-used by the game designers many times in random encounters. Because we are talking about level design, the UPP factor will only refer to how often a player visits a place in a game, so re-usable elements play no role.

Linear World

A linear world is a world like the world of Contra or R-Type. The player progresses along a linear path. Navigation is trivial and therefore plays no important role in the gameplay. To help to motivate the player, the game presents a series of intermediate goals along this path. The overall feeling is rather like a theme park ride – a game on rails.
The advantage of this design is easy game balancing, especially in a game where the player’s avatar becomes stronger as the game progresses. Generally, the level design seems to be easier this way. It is also easier to apply a story on the game.

The first level of the NES Contra. A classic example of linear world design. Click here for full resolution.

Games with this worlds rarely involve any kind of backtracking so one big disadvantage is that every place in the world has an UPP factor of 1 which is a bit uneconomical. Then again, this kind of world is used in short games where there is no save function so player will restart the game often and visit the places more frequently.
The major, obvious disadvantage is that this kind of world is boring so the game has to make up for it elsewhere. Contra offers a tight action-packed gameplay so players don’t really mind.

Linear World with Branches

A world similar to a linear world where at some point, the player is allowed to stray from his path to peruse an additional goal. However ultimately players run into dead ends and come back to the main path. This world design is used in many modern games as it offers the advantages of the linear world but provides some freedom for the players. It is also the world of Sword of Mana.
Linear worlds are really great in short games where players frequently restart the game and therefore experience parts of the game multiple times to try different things. This is especially true if you add branching. If replaying is part of the idea, players can easily check out the branches they missed and experiment with different choices.
However, if the game is long and features save points, replaying is not really an option. Then, the UPP of content in the optional branches drops to something between and average of 0,5 and 1, depending on how many players will decide to backtrack if they find out they missed a branch. This is bad for players because they get less for their money and bad for the game designers because they throw money out of the window.

The Semite Falls level of Final Fantasy II (from the “Dawn of Souls” GBA remake). The player will stray (orange line) from the direct path to the exit (yellow line) to get the treasures in the branches. Note that one branch contains no treasure (red line). The player has no choice but to swallow this random punishment. This is the best-case scenario. If a player accidentally chooses the direct way to the exit, he might want to backtrack to check the missed branches.

Sid Meier said “A game,is a series of interesting choices”. The worst thing about linear world with branches is that it is a series of annoying choices. I often find myself in dungeons, which were designed with this philosophy. I encounter a fork where clearly – one path will be the “right” one and the other will turn out to be the dead end. Even if I find the “right” path, I will backtrack to the fork just to make sure there was no hidden treasure in the dead end. This sucks. The choice is between bad or really bad: At each fork I randomly decide between EITHER the right path but where there is the potential of lost content OR the wrong path with maybe additional content but certain backtracking. So navigation is in some way a part of the gameplay but it is very scarce and doesn’t provide satisfaction.
It is a phony world design where game designers are trying to hide the linearity by giving the player choices they don’t really want to make anyway.

Linear World with Forks

A linear world with forks is a world where at some point, the player can decide either to go one way or the other way. After the choice has been made, going back to see the other option is not possible. The few examples of this world design include R-Type Final, Starwing and New Super Mario Brothers.

Starwing for the SNES offers 3 different routes. Each consists of different levels with different difficulties. However, the last and first level are the same.

Forks is something rarely seen in games for an important reason. While branches introduced content with UPP between 0,5 and 1, forks introduce content with UPP of exactly 0,5 and it drops even lower if there are forks within forks. So in a game with branches, some players might never visit a place but in a game with forks no player will be able to visit every place in one session.
Again, like with branches, this only makes sense if replaying the game is part of the idea but don’t fool yourself: rarely any player will ever replay the huge 30h+ behemoth most modern games mutated into.
Forks have other disadvantages. For example, applying a story might be a bit more difficult. Nothing important may happen during a fork unless you want to create two different stories for the game part after the fork. That’s why forks are often introduced at the very end of the game, where there is not much story left.
However, in some way the decision in a fork scenario is a bit more humane and interesting then in a branching world because there is no “right” choice, at least from the point of view of overall UPP.
The funny thing is that game designers implement this strategy to give the players more freedom – to let them decide where they want to go. However, in reality they take away freedom for them. The players don’t decide where they go, instead they decide where they will never be able to go.
Of course, this doesn’t apply to short games. In short games forks do make sense and even enhance the gameplay. In subsequent playing sessions, navigation becomes slightly more important as the players need to keep track of which places they already visited. R-Type final clearly profits from this advantage.
To sum it up, in branching worlds the player has to decide between “bad” and “really bad”. In worlds with forks the player has to decide between two different possibilities, both equally “bad”.


A network is a word design where a number of places are interconnected with each other. The player can move freely from place to place. Goals are spread across the world to encourage exploration, rather then drive a linear progress. Examples of such games include the old NES Zelda game and both Logic Factory games: The Tone Rebellion and Ascendancy.
Networks have major advantages over linear worlds. They provide real freedom as they offer player lots of choices. They can create rich and convincing environments for the players. The UPP of most places in a Network is far above 1 which makes them more cost efficient as well.
The flip side is that they are a pain to balance. Players might quickly stumble into areas where they aren’t supposed to be yet or get bored because they quickly obtained some powerup which removes the challenge from large portions of the game. Players also might get stuck or even get into a position where they won’t be able to complete the game. Networks are also hard to tie into a story because it is difficult to predict the order in which a player will visit the various places.
Of course a network world is also much more difficult to design. The different places need to be interesting enough so the players won’t get tired of visiting them over and over again. Also it isn’t really suitable for quick and short games because exploration takes a lot of time. The effort to create a network world only pays off if players have the opportunity to explore it.

Drawing maps of dungeons… those were the days…

It is important to understand that in a network world, navigation automatically becomes a major part of the gameplay. Players will need tools and hints to successfully travel between places. Spatial continuity like in Mystic Quest in one such tool which simplifies navigation because it allows players to create a mental (or even paper) model of the network. Other tools to simplify navigation include in-game maps, in-game compasses and signs.

Controlled Networks

I lied. Except the mentioned The Logic Factory titles, almost no game world based on a pure network. Even the old NES Zelda is a mildly controlled network. A controlled network is a network, where access to certain parts of the game is restricted. The players need to obtain certain items or fulfill certain requirements to be able to access more parts of the network. However, at some point during the game, they are finally able to access the whole network. Also, unlike a linear branching world, players are always to some extent free to move around between different places. They can also select different routes to reach a destination. There are many examples of such games. The Metroid series, The Zelda series, The GTA series and of course Mystic Quest.
The advantage of a controlled network over a simple network is that game designers have some means to balance the game and even add some story. However the advantage of a network world is based on the amount of freedom in spite of control and the quality of control.
The border between a controlled network world and a linear world with branches is rather blurry. One can imagine adding branches and interconnections to a linear world until it becomes a network. On the other hand, you can imagine a network so heavily controlled that it becomes a linear world. It is important not to destroy a network world through too much restrictions.

The world of GTA3 consists of 3 islands connected with bridges. The bridges are all broken at the beginning of the game and open only after the player finishes certain, unrelated quests.

The way the game controls access is also important. For example, an NPC telling the player “You can’t go in there right now” is quite stupid. Also very popular are broken bridges which magically open up as soon as the player completes some unrelated quest (GTA 3 for example). Borderline stupid are “doors” – obstacles which can only be removed if the player finds a special item, which isn’t useful elsewhere. I call those solutions stupid because they control the accessibility of an area in a very obvious way.
One concept which seems to work well is giving the player some new ability enabling him to overcome some kind obstacle which isn’t unique but rather occurs frequently in the game. On top of that, the new ability should have some other uses so overcoming the obstacle is not its sole purpose. This way, the obstacle doesn’t seem artificial but natural. In Mystic Quest, the player obtains a whip, which can be used to cross a river. The whip itself is useful as a weapon and can be used to cross many other such obstacles of the game. The same game mechanic lies at the heart of the Metroid series and the Zelda series.

The Mystic Quest whip at work in a dungeon. Like the Zelda grappling hook, it can be used to pull the avatar over obstacles.

Funny enough Sword of Mana manages to include this game mechanic but gets its wrong at the same time. Some parts of dungeons are sealed off with rocks which can be destroyed with certain magic. The rocks always looks the same, different colors only indicate which spell is supposed to be used. The different magic spells have various elements assigned to them so it would have been extremely simple to come up with a more creative kind of obstacle for each spell – for example an ice block for the fire spell. Instead, the player is treated with a generic magic rock and quickly learns “Aha, now I’m SUPPOSED to use some kind of magic spell, *yawn*”.

The dumb stone seal from Sword of Mana. Time to use magic.

To close up this long post, it is interesting how by not being able to apply scrolling technology, old games almost automatically create network worlds. Every screen becomes a place connected to 4 different places. Those networks are also most peculiar because the navigation decisions occur at a very high frequency. On every screen, the player needs to decide where to go next. The resulting gameplay is one where navigation is even more important. At the same time, dividing the world into a rectangular grid simplifies the process of drawing custom maps of dungeons and levels. There seems to be a gravitation towards a navigation-centered game design.
As a result the world design of Mystic Quest and the resulting gameplay is simply completely incomparable to the its so-called remake.

Krystian Majewski

Krystian Majewski was born in Warsaw and studied design at Köln International School of Design. Before, he was working on a mid-size console project for NEON Studios in Frankfurt. He helped establish a Master course in Game Design and Research at the Cologne Game Lab. Today he teaches Game Design at various institutions and develops independent games.

8 responses to “Sword of Mana: World Design”

  1. K. Thor Jensen

    This is a really excellent blog, I’m so glad I found it. Bookmarked! You might want to check out the game company I work for, Gamelab (www.gamelab.com) – we have designers and programmers here trying to do similar things on our blog in terms of design analysis. Plus we make good games.

  2. Krystian Majewski

    Well, if you really are THIS k. thor jensen then frankly, I don’t know what to say.. Except maybe that YES of course we know gamelab and that we think you guys are fantastic and that we love your work and that you are a great inspiration for us! It is a great honor for us seeing you visit our blog. :-D

  3. Yu-Chung Chen

    Nice article!

    Good points and well written. Also UPP is a nice term, reminds me of the the concept of data-ink introduced by Edward Tufte.

    Speaking of whom, his chapter on Layering (in Envisioning Information) delivers a good explanation of how the Zelda and Metroid series manage to keep high UPP places interesting – they are layered with gameplay elements for different stages of progression. By tying major rewards to new ways of seeing and “using” the world, the revisits of those high UPP places are refreshed. The layering also makes the balancing manageable. I guess (never actually did that kind of level design).

    But this core formular not only “addresses the problem” of burnout of high UPP content, but rather in turn induces a kind of the 1+1=3 effect, in a positive way: the world feels much richer because things aren’t there only to be solved in a quest, but feel more like a natural part in the world. Check out this
    Gamasutra feature
    , too.

    You said that new abilities should have multi-purposes. I totally agree. Layering and enhancing UPP also applies to game mechanics, which was exactly my fundamental idea to deepen my game Gravity (currently called Lovely Dilemma). By tying tokens and behaviors to different dimensions of a verb, these can be balanced separately (layering), ideally benefitting from the concept of 1+1=3.

    About replayability in huge games, I expect to rant about it in my FFXII diary. But your analysis says it all already.

    Oh, and it’s great to have visitors – and even a top-class one at that! A great motivator to post.

  4. Krystian Majewski

    Thanks for the valueable feebback!

    I didn’t think about data-ink when I wrote about UPP but it fits perfectly! He certainly had some influence on the way I think. However, UPP is also related to LostGarden’s HPD

    I totally agree about your observations on layering. Again, I completely missed that Tufte connection. I’m planning on writing something about layering in a post about Zelda.

    Applying UPP on Verbs is brilliant! Of course! Why didn’t I think about it! BTW, after your running project you simply have to get Gravity online somehow.

    Thanks for the Gamasutra feature. It is a perfect follow-up to this post. Also, Kevin Carter gets extra kudos from me for mentioning Another World. Yet another unknown masterpiece.

  5. Evil Dan

    Wow. That was beautifully articulated. Thank you so much for putting down in writing your thoughts about UPP. This is a wonderully usefull concept for developers.

    You might be interested in this article I wrote as a critique of the Legend of Zelda for the NES:

  6. Pablo

    You’ve got some good shit going on here

  7. Chris

    The link to The Tone Rebellion is broken, it should be:

  8. Gonzalo

    very good blog.

    Clever points in design… this remember me RPGMaker’s good old days. I write ’cause i wanna make you a request. Do you have another bigscale maps of these games, Final Fantasy Adventure and its remake Sword of Mana? I am nearly obsessed with the first Seiken Densetsu (finishing first time in 1992), and I’d like to get all these maps. If you don’t have it, what program u used to open the source file?

    Thank you for your time.



Game Design Reviews is a Blog used by a group of game designers from Germany to publish and discuss their thoughts on various games. The blog consists entirely of reviews of games. Each review focuses on the important game design ideas and concepts of that particular game. We also run a second, more informal Blog called Game Design Scrapbook.


follow Krystian on Twitter
follow Yu-Chung on Twitter
follow Daniel on Twitter