Riffs by Raph: Sandbox versus Themepark
I just watched a couple of videos about sandbox vs themepark games (in particular one by NerdSlayer and another by Josh “Strife” Hayes)… One thing that struck me about the ways players often talk about this (because at this point the history is so old) is that people think of sandbox as the older version of MMOs, and themeparks as newer. But that’s not right – sandbox is not the older form.
Sandboxes are the evolution of themepark MMOs, not the antecedent.
Part of the reason why this isn’t clear is because most players today haven’t played what themeparks were originally, back on the text virtual worlds called MUDs that led directly to MMOs. Given that I suspect I am partly to blame for these two words having currency in the first place, I thought I’d put in my two cents.
MUDs had more “species” than MMOs do today. Caveat: I’m gonna really oversimplify this history here. If you really care, the MUD Wiki on Fandom, started in the wake of the great Wikipedia purge of MUD content, has you covered. At peak in the early 1990s, MUDs had mostly abandoned some of the older sorts (such as “talkers” and “scavenger hunts”) and solidified around a few types.
There were creative worlds such as MOOs where anyone could build anything. Today’s inheritor to this tradition is something like Second Life. Maaaybe Roblox could be considered this, but really Roblox is more oriented around making games than any MOO was.
Many many MOOs ended up less about creativity than they were about just hanging out – basically, the progenitors of social worlds ranging from Habbo Hotel to There.com, and eventually even the hellsites known as Facebook and Twitter.
There were pure roleplaying environments, usually on MUSHes -- an activity that these days happens more on Discord probably! Or in GTA Online, weirdly. These usually had LARP-style combat, with mutual-consent style systems for interaction.
There were a few roleplay combat MUDs like Armageddon where you had to apply to even be allowed to play, with a full roleplaying background. They had all the gameplay of the combat games, but you could be banned for breaking character!
Then there were pure hack n slash games, the Diku model but also most LPMuds. These were about starting out at level 1, going to zones for your level, killing mobs to lvl up, then off to the new zones that you could now handle with your increased power. While there were tough encounters (the “raids” of their day) the levelling was the principal gameplay. There wasn’t usually a major narrative element there, though – it wasn’t until World of Warcraft that we saw what I call the “quest-led game.”
Combat MUD zones were assembled more or less carefully to supply the right level of challenge for players or groups of a given level range, reward with gear of the appropriate power, and were assumed to be “run” over and over until you got diminishing returns. Each zone was often fictionally themed, and usually they made no sense sitting next to one another. If you want to analogize to a “land” in a theme park, you are really not far off at all. And hence some of the origins of the term…
Earlier in MUD design, these zones even “repopped” all at once. That is to say, players ran in, and killed stuff, and the zone respawned all the mobs at once, on a timer. It was like resetting a little stage-play; the NPC actors hit their marks and reappeared at their start location. It was actually seen as a notable innovation when individual mobs repopped on their own timers. Why did something like that come to be? Well, because many of us who were making MUDs wanted more… realism. More immersion.
Here’s where the history gets interesting. MOOs and LPMuds were highly scriptable – the former exposed it to users, the latter to builders. But either one was radically different in that way than the most popular game engine, which was the Diku style (we called them “codebases” back then).
Dikus were entirely hardcoded. All the content in the game was just data in fields. And that meant you had to be a programmer who knew how to code C on a Unix machine to add any new behavior to the system. On the flip side, adding data was really easy. On MOOs and LPMuds, you had a lower barrier for adding features, but you also had a higher barrier for “just making content.” Most Dikus could, with minor text editing, share their zone files, because they almost all played exactly the same!
(Ironically, the high barrier for LPMuds actually led to something similar, where because adding consistent functionality to games was hard, people released “mudlibs” of functionality, which resulted in lots of clone games there too).
As you might guess, pretty soon this led to more Dikus than any other type of MUD. But it also led to… envy. LPMuds usually pioneered the cool features in combat MUDs, because they could. They also had (gasp!) quests. Not a new idea, of course…Quests had existed in the earlier MUDs predating the whole LP/Diku/Tiny ecosystem… they were in fact implicit in the scavenger hunt model that went all the back to gathering items to put in a display case in Zork; and present in the design of MUD2 (it had eight of them by 1991, and you had to solve some to become a wizard)..
So… DikuMUDs started to add scripting systems, which were generally much more limited than the ones on LPMuds, but fit better with the whole “fill out some rows in a form to add content” development model. Some of this was used to just add nicer NPCs, with conversations, reactions to the environment, etc. But on a few Dikus, it led to making quests too, and it was actually a lot easier to build quests in a “quest system” than to one-off hand code them in an LP mudlib that didn’t implement a “system” for it.
I played on Worlds of Carnage, the first DikuMUD to have scripting like this, and then worked on LegendMUD, which was one of the very next to have scripting – and Legend’s was quite a bit more powerful, apparently.
Carnage is the first Diku game where you really see stuff comparable to quests from World of Warcraft. But it was still very much a Diku! We would sit at the fountain waiting for the spider queen or whatever to repop, which we would check by typing the HUNT command over and over. Then we’d zoom over and kill it in order to maximize our XP per hour. But you could also have a newbie sword that talked to you and gave you advice, or visit a zone based on Romeo and Juliet full of quotes from the play… and kill ‘em all. The scripting on WoC made for better combat too, because you weren’t stuck with just the built-in combat AI for your enemies. AI had more sophisticated combat tactics, because designers could add fresh behaviors to the monsters.
On LegendMUD, though, we chased after something else. We pushed the Diku style to add things like… a tavern where you could play blackjack. An entire instanced Last Man Standing minigame. (Pretty sure we lifted that from an LPMud). An out of character lounge where you could attend a lecture series. When a famous player character retired, we could create a bot with their name AND give it all the catchphrases and RP the player had developed, and set it loose.
And above all… quests. Real ones. Stuff that in many ways was quite a bit fancier than what WoW eventually did. This, in fact, was what my signature design style was. Highly narrative, intricate puzzles, and immersive storytelling. It would take a looong time before any MMORPG had anything comparable. If you follow that link, you’ll see open world events occurring as the quest proceeds, done without zone or quest locking, etc. Heck, there was one quest that redid the entire zone spawn.
These were absolutely “rides” the player went on. And I loved them, and loved making them. But they were also a ton of work. And we and other MUD designers were wondering how to get more of this stuff to happen without having to hand-code it all.
Enter “simulationism.” We used the term actively on MUD-Dev back in the day, making the distinction between simulation and “stagecraft.” Simulationism was the idea that you could get the MUD’s basic behavior to the point where cool stuff happened organically. MUD2 had far more simulation than MUDs did in the 90s… or any MMOs today!
A key inspiration in the mid-90s was a game called DartMUD, which featured farming! And crafting! Crafting in other games was mostly one-off stuff tied to quests, or really simple magic item creation. DartMUD tried for a fuller player economy.
The state of the art has moved on so far that it’s hard to convey how important the idea of “let’s make these behaviors generic” was. An example from a bit later, from EverQuest 2: they had dogs that chased cats. Simulationism says “dogs dislike cats, and attack them. Cats flee when attacked by dogs.” In fact, it says “mobs have hates, and aversions.” It’s generic. But the old scripted way was hand coded, so it only worked with that cat and that dog.
In fact, EQ2 did in fact just hardcode some cats running a loop around the city, with some dogs running the same loop a few feet behind them, and called that “virtual ecology.” Contrast that with Ultima Online, a simulationist game where wolves actually did hunt rabbits.. and at first, the rabbits learned from the experience, and got stronger!
When people today say sandbox versus themepark, they mostly mean simulationism versus stagecraft. WoW is a giant amazing piece of stagecraft. So is FFXIV. But Eve is basically simulationist. The tradeoffs were evident early on.
So simulationism was born as a way to make fantasy worlds richer, more immersive… in a sense, to “make the ride better.” Despite the knock against it today, it was not seen as a way to offload the effort onto players – back then, that’s what MOOs were for. In fact, UO was fully intended to have the same sorts of quest content that LegendMUD did. You can even find the bones of it – a hedge maze here, some inscriptions in a dungeon there. But we simply did not have time to make any of it, and so you got a freeform “sim world.”
You might enjoy this MUD-Dev post from 1998, which contains boggling moments like “nobody’s really tried a good storytelling MMO” and astonishment that “there's now an EXCHANGE RATE between UO gold and real world money…”
Contrast this to EverQuest. EverQuest was based on Diku gameplay, end to end. At launch, it had no quests, really, despite the name. Zones had level ranges (though infamously, lots of “zone sweeper” mobs). It famously only added any crafting because the dev team saw it in UO during beta. Today some folks like to say it is “sandboxy” but I’m here to tell you that it was absolutely a Diku-style themepark, of the pre-scripting period.
In fact, during EQ’s beta is where we can see the basic metaphor coming together, in this post from 1998.
Hence why I say that to some degree, the terms are my fault. Not that I coined them, of course; the concepts were very much in the air.
The result: when we put linear narrative quest content into Star Wars Galaxies, we called those portions of the map where those narratives lived “themeparks.” We didn’t have anywhere near enough of them, because ironically, we didn’t pick up the Diku lesson and have nice templatized quest systems.
But then along came World of Warcraft, and the whole game was linear narrative quest content. How? By spending around five times more money than any other virtual world developer ever had in the history of mankind. Before WoW, you played a Diku-style game to kill ten rats. After World of Warcraft, you had a quest to kill ten rats. It seriously changed everything, because both sandboxy and themeparky games had a problem with guiding players.
It’s hard to get across how big a revolution that was. We had chased simulation in part because of cost and scalability, which was a horrendous problem for us all, a problem that both Josh Hayes and Nerdslayer specifically called out in modern themeparks. And a problem EQ blithely inherited.
Now, WoW made MMOs accessible by fundamentally making them linear. Oh, not totally. But classic WoW actually punished you for leaving the quest line. WoW was less explorable at low levels than vanilla EQ, for example, precisely because of how carefully and well done the level scaling of content was.
Which brings me to the other terms I used to use: “worldy” versus “gamey.” The MUDs that wanted to seem more immersive, more like an alternate world, they didn’t color-code your opponents (there was a “consider” command but you had to invoke it manually). They definitely didn’t color code your gear. You didn’t march through “progression sets” of gear at all, really. A lot of gear was easy come easy go, actually. There was a general design consensus that even if a MUD was telling a story in its quests, the world was definitely not “your story.”
It was a world. Worlds do tend to be larger and more confusing and less approachable than games, for sure.
One last thing – these days, because UO set the template for “sandboxes,” and then Eve reinforced it, a lot of people identify sandbox gameplay with ganking, full player vs player environments. But they aren’t equivalent. UO had that for simulation reasons. Galaxies did not. Any given game can decide where to draw that simulation line. UO was trying to solve governance problems using simulation, out of a belief that to do it by controlling players would be too expensive. We were mostly wrong.
Bottom line: sandbox does not equal PvP combat.
These days, if you are a hardcore enough aficionado of online worlds, you might see reference to the term “sandpark,” often applied to games like Runescape, Black Desert Online, or ArcheAge. That definition is actually what sandbox meant originally. (Runescape at launch was basically very much like UO btw.)
You can build linear narrative themeparky content on top of a rich simulation. And sandbox does not have to be devoid of content. But it is impossible to do the reverse. Linear narrative themepark stuff is by definition breaking the rules of the sim.
Today, that simulationist impulse is visible in Minecraft. In Roblox, which is built on a physics sim, at heart. In No Man’s Sky and in Fallout 76. But it’s also visible in the Holodeck, or in anime like Sword Art Online (which was directly inspired by Ultima Online!).
In fact, it’s visible in pretty much every other fictional depiction of online worlds. When we dream of alternate virtual realities, they are always simulationist, because the whole point of dreaming is to dream of richer worlds and richer experiences, ones with more internal consistency.
This doesn’t mean that the guided onramp, the clear progression, and the crisp goal-setting of themeparks is bad. Hardly. It’s necessary for broader audiences. Even Minecraft has a much clearer starter set of goals than UO did. But!
Sandboxy stuff – worldy stuff – simulated stuff – is how your themeparks get better. And the biggest reason why the themepark line has dominated is because it tackles mostly solved problems, compared to building a true alternate world. It’s stuff that single-player game designers know how to do. Breadcrumbs, dialogue trees, cutscenes, progression paths. Expensive, but at heart predictable for the developer. Alas, also for the player, after a few run-throughs.
Me personally… I’ve been visiting these worlds for thirty years. I was done with predictable a long time ago.