Space Narratives: A Map Generator Denounces Map Generation

I got my start in game design doing traditional Rogue-likes. My first game was 100 Rogues(now sadly rendered unplayable due to a combination of iOS updates and a long story involving legal rights), very much a typical turn-based Rogue-like, somewhat inspired by Shiren the Wanderer but with character abilities more along the lines of Diablo or Warcraft III.

Since then, I've made 4X games, wargames, a weird experimental DotA-Rogue hybrid, and other things in this space. To that end, random map generation is something I've always thought of as an obvious starting point for my designs. Whatever I did next, the first thing that I knew it would have would be randomly generated maps.

To say that this is because of my background in Rogue-likes would be having it sort of backward. I still think I had good reasons for wanting randomly generated maps in the first place; the general "proc-gen-ness" of the games were a big part what brought me to Rogue-likes in the first place. For me, the fascination with Rogue-likes had something to do with witnessing a pattern of games being too consumable, too puzzle-like, for too long. I thought of games like Advance Wars, which I enjoyed, but which you could kind of only play through once. Once you know the strategy that works, you kind of need a new Advance Wars map to play on; you've kind of "solved" that level (at least, for single player).

Gem Wizards Tactics

In 2019 I began work on Gem Wizards Tactics, my own answer to the issues I had with games like Advance Wars. As such, one of the first things I decided about the game was that it was going to have randomly generated maps. This would contribute to the game being endlessly replayable, with billions of potential map configurations that challenge the player in different ways each time.

Gem Wizards Tactics was a one-person operation pretty much, with me doing 95%+ of everything, so the amount of time I could spend on any one aspect of the game was limited. But I got the map generation working well enough. Forests would spawn in clusters, hills and mountains would be dropped in one at a time in random spots, and roads and rivers would shoot through the map in various ways. The towers (which are the objective) spawn at roughly the same distance from the player start, and player also has a guaranteed starting location that's paved and guaranteed to be nice and clear of obstruction, to help them navigate themselves into place for the battle (another thing I wanted to do differently was to skip the 'placement phase' many tactics games have).

Here are a couple of examples:

The mountain-looking tiles you see are impassable (usually). Rivers are hard to walk through, forests give you more defense, and hills give you extra attack damage.

Overall, the maps look alright and I haven't found the generation to swing the randomness too much. In short, it works.

But I'll never make another randomly generated map again! And I'm actually going to add a bunch more non-random maps to Gem Wizards Tactics in the next patch (coming in a week or so).

Places: "Space Narratives"

I think with videogames, in particular videogames where you have a little character or characters that you move around through a space, it's important that that space feels not just like a space, but like a place.

This is something that's been kind of hard for me to really internalize, as someone who has historically been really fixated on abstract rulesets, and who spends a lot of their time making tabletop games. This isn't to say that this "sense of place" thing isn't important in tabletop, mind you. I think it actually is, but with videogames, when you lack that sense of place, it seems like a much bigger problem for engagement.

There are probably a lot of reasons for this, but for now it probably suffices to say that it's a big part of the language of videogames. We're very used to videogames looking, sounding, and behaving in certain ways, especially when it comes to certain genres. For turn-based tactical wargames, I think a lot of people associate either things like Advance Wars, or perhaps SRPGs like Final Fantasy Tactics or Triangle Strategy.

All of these games have rather heavily narrativized maps that are meant to be a particular place, like outside a church, or near a castle, all of which are relevant plot locations. Part of the value here is that it helps to join the "story" (conversations, cutscenes) with the tactical abstract gameplay. But that's definitely not the full explanation.

My belief with Gem Wizards Tactics was that, because it's a much less story-driven game, we wouldn't really get as much value out of hand-crafted maps as a game like Final Fantasy Tactics would. But I think I was surprised by how much... anti-value the randomized maps seemed to bring to the game!

In videogames, regardless of how rulesy and abstract the particular game is, there is a basic level of surface-level aestheticized-interaction that's super important. How does it feel to move a character from one place in the world to another place? How does it feel to jump? How does it feel to dash? A lot of times, developers think of these things in terms of "polish" or "juice" or "feedback": how quick or responsive the controls are, how many/which visual/audio effects are triggered,

But a really big part of that feel is about the context for the move. Is it a move up a big giant hill? Is it a move over a river? Is it a move around a corner? All of these things are like "space-narratives" that require a certain level of narrative coherence for the map.

For example, if you have a giant, two tile thick river rushing through the middle of the map, that's a big, bold and clear "space narrative": there's a giant river in the middle and so getting through it safely is going to be a big deal. You can trap the enemy in there, or you can try to rush through it somehow, but in any case it's a very distinctive concept of a map.

Another (very simple) example would be a map that has a bunch of impassable tiles in the middle, creating an O-shaped level. Now we have like a "fork in the woods" type of space-narrative - do you go left, or go right? Or bring some each way?

On an aesthetic, feel level, these kinds of space narratives are super-duper important for giving player actions meaning and making them feel good.

The obvious concern that any designer making an "infinitely replayable" game with maps that have strong space narratives is that it's very likely that dominant strategies will begin to form in such maps. For example, "oh, on the river level, the answer is always to let them come through the river" or "oh, on the donut level, you always want to send all your troops to the right." But is there anything we can do about that? Is it possible to avoid those dominant strategies and have a strong sense of "place" for a map?

Ocarina of Time Randomizer

I had already been thinking for awhile that you don't really have to randomize the whole map in order to avoid a lot of that dominant strategy business. I've suspected for awhile that you only need to randomize a certain amount of stuff, and if you do, it's basically as good as a new map each time, while still maintaining a good "space narrative".

My recent experience with Ocarina of Time Randomizer made me a lot more confident about that. If you're not aware of it, it's a community made mod-patching tool that lets you randomize the location of items in the game. It actually has a ton of different options for what to randomize (you can even randomize sound effects!), but by default, it basically just swaps locations of major items. So where there used to be a piece of heart in some random hole behind a waterfall, might now be the hookshot. It might take awhile before you find a sword, or you might get a really powerful tool early on, way before you're supposed to.

Remember this chest, right near the beginning of the game? With OOTR, it could have the Deku Sword in it. Or it could have the Eye of Truth, or the Zora Scale. Or 5 Rupees!

Just that simple act of changing the locations of items radically freshens Ocarina of Time, and lets players who have played the game dozens of times play it again and sort of feel like they're playing it for the first time. It's really cool, and I recommend all Zelda fans check it out.

But it's also a very strong signal that you don't have to randomize all that much to get good benefits of randomization. OOTR doesn't change the geometry of the world at all - it's that same world you already have printed permanently in your brain like it's the house you grew up in, but the order at which things get unlocked (or sort of unlocked) is now reorganized and all weird.

I've done some testing on non-random maps so far in Gem Wizards Tactics, and I think it's going to work. We still randomize a lot of stuff:

That is a TON of randomization still! So I have every confidence that the game just simply doesn't need the random maps, and will benefit greatly from having hand-made maps.

Conclusion

I think, looking back on it, if I'm being fully honest... part of the reason I didn't want to do hand-made maps in any of my games is simply that creating a map editor is a huge pain. Or alternatively, doing one of those things where you create the map in a text editor or something like that, is also a huge pain but in a different way. For Gem Wizards Tactics, a nice thing is, we actually have a fully functional map editor now! So even though the game came out back in February of 2021, and even though the Business Demon inside me tells me to just move onto the next game, I'm implementing non-random maps, coming soon to a patch near you.

Going forward, I don't know if I'll ever make another videogame that has randomized geometry. I'll be spending a lot more time thinking creating more "space narratives", and as a result - I think - making games that are a lot more fun to be in.

~

If you enjoyed this article, consider supporting my work on Patreon.