The Zelda-Rogue Design Trap
Combining Zelda and Rogue-like elements isn’t a game design panacea
Procedural and random generation are the cornerstones of rogue-like design; we’ve seen numerous rogue-like games push these elements further than ever before, too. In just this past year, titles like Dead Cells, The Binding of Isaac, Spelunky, and Dwarf Fortress have all done great things with rogue-like design principles.
But there are design traps involved when building rogue-likes. One such trap I occasionally see is developers trying to build a “Zelda Rogue-Like” experience (that is, combining design elements of traditional Zelda games with the random or procedural elements of Rogue-likes). And although this might sound like a magical combination, it really doesn’t work from a game design perspective.
Procedural vs. randomised
Let’s start by defining the difference between procedural and randomly-generated content. Randomized content refers to set elements that are shuffled around in an environment or level; enemy positions in Dead Cells or the types of treasures that appear in The Binding of Isaac are examples of this.
Procedural design is when the game generates something itself from scratch either before starting the game, or while running it. Loot generation in an action-RPG, or the level layouts in Spelunky are relevant examples.
Rogue-likes typically feature both kinds of generation as part of the experience, and it’s important to understand just how much they impact gameplay.
Degrees of random
Not all random or procedurally-generated elements are equal in the eyes of the player and their impact on the gamespace.
In the past, I’ve discussed the idea that simply including random or procedural elements in a game doesn’t automatically make it replayable. Many titles that leverage either concept will still tend to adhere to a specific progression curve. Survival games are a great example of this; no matter how much the environment is procedurally-generated, the player still has to gather resources in a specific order — Minecraft is a prime example here.
In the absence of a more suitable term, I’d argue there are “hard” and “soft” elements that can be randomized or procedurally-generated in a game. A “soft” element is something that doesn’t impact gameplay or the overall progression loop of a title — examples might include enemy spawn locations, the basic structure of environments, or the position of different biomes.
A “hard” element that does impact the player’s ability to actually progress through the game is something that directly affects what said player can actually do (or not do) at a given time. This might include elements such as what type of upgrades spawn, new items introduced to the game, the types of biomes generated, enemy types, new resources, and so on.
The crucial factor is that what the player is doing in the game must be directly impacted in order for each play through to feel different.
This is why loot in action-RPG design works so well: every new piece of gear can change the player’s build (and thus, directly impact gameplay and progression).
Whereas changing the map around doesn’t usually lead to substantially different experiences, as the player is still doing the same things. The reason why Spelunky holds up so well is that even though the gameplay itself doesn’t change, entirely new levels are generated, which continually gives the player new ways to test their core abilities. This is what leads me to the idea of the “Zelda Rogue-Like” and the inherent problems with it.
The Legend of Rogue
The Zelda franchise is considered to be one of the best action-adventure series in the industry. Much like the Soulsborne franchise, every aspect of Zelda is hard-coded to push and guide the player through to the very end. Even Breath of the Wild — the most open-ended take on Zelda yet — still incorporated key progression points in the form of the Divine Beasts, the Master Sword, and Shrines.
On the surface, it may sound like taking that handcrafted design — including its progression elements — and dropping them into a procedurally-generated world is the holy grail in terms of gameplay. Unfortunately, though, there are several fundamental flaws with this concept.
When we look more closely at the Zelda gameplay loop, the player is essentially doing two things: exploring to find points of interest (dungeons, items, etc…) and performing a task at said point of interest. The latter is arguably more interesting (and important) than the former. The world of a Zelda game is always inherently designed to force progression — what I mean is, the environmental design will guide the player to their next objective in order to proceed.
Placing a randomized world on top of the Zelda formula ends up making things both less interesting and more frustrating at the same time. If the world design is always going to change based on procedural generation, for example, then this necessitates a less complex environmental design in order to accommodate for the various combinations that will emerge from a procedural system. Outside of any major points of interest, the overall world itself is likely to be less interesting to explore — especially when compared to a world that has been “hard-coded”.
Not only that, but it creates a situation where the player might find points of interest but have no idea if they’re actually supposed to be there or not. The beauty of Zelda’s formula was that each new dungeon and point of interest each built on top of the previous one. This makes the progression curve more engaging, and allows the designers to continually raise the stakes with each new dungeon.
Breath of the Wild went in a different direction by making every dungeon beatable with your basic load-out. On top of that, it balanced the dungeon locations around how each or difficult it was to physically reach them in the world — this idea requires “hard-coded” design, and is likely to be much more difficult to achieve with procedurally-generated worlds.
Another problem is that because the in-game upgrades are the only elements that actually change the game experience, simply shuffling or procedurally-generating the environment isn’t going to change how someone plays the game. A popular design that developers are using is one where they generate the overworld map procedurally while still making the individual points of interest hard-coded: see Darkwood for an example of this. In this example though, changing the environment around doesn’t work; it’s not impacting how a person progresses through the game.
If the player always needs items X, Y, and Z in order to win, then all you’re doing by shuffling the environment around is changing the amount of time needed to acquire those items. As well, you’re now unable to introduce unique or differing challenges in the overworked itself, because you’re unable to know what order the player is acquiring these upgrades. What’s worse is when each point of interest is designed in such a way that they require the use of items collected from previous points of interest (i.e. following the traditional Zelda dungeon concept) — but in this case the idea doesn’t execute well, because the player has no idea about the rule when they first arrive at a point of interest (and because, again, the developer doesn’t know when/in which order the player is tackling each one).
This is why the better rogue-likes focus on a core gameplay loop that then expands out from the center and changes based on the items given to the player. Both The Binding of Isaac and Dead Cells each change dramatically based on which items show up as opposed to which environments the player is moving through. While Spelunky sticks with a basic gameplay loop, it mixes up what the player is doing thanks to the changing level generation.
I think the key element of any successful rogue-like is based on how many ways you can provide the player with a different experience. Every “Zelda-rogue” I’ve played makes me wish that the environment was fixed and the points of interest were, well…more interesting.
You would think that creating a procedurally-generated take on The Legend of Zelda would be the perfect game, but I’ve yet to see anyone fully accomplish this task (and before anyone says it, The Legend of Zelda randomizers don’t count). The amount of effort it would take to get these gameplay loops to sync up in the correct way is likely beyond most teams’ modest budgets.
The point of a good rogue-like is not about getting to the end — it’s more about the many ways the journey can articulate. There are possible ways that this magical combination could work (perhaps related to building a gamespace around different pools of items/points of interest) — but again, this would require both an enormous amount of talent and effort.
I actually spoke to a developer on the Game-Wisdom Discord about this recently. In chatting about this design approach, we agreed that it’s definitely possible but would require much greater focus on the procedural algorithm in order to create areas that would be “solvable” based on what items or abilities would drop in each respective seed.
Before I end this article, I do want to throw a question out there to you, the reader: Can you think of any other “dream concepts” of game design that just don’t work — or are unlikely to get right — in reality?