The multiverse in which we play

If the Many Worlds hypothesis is true then we are living in a multiverse of parallel realities and alternate histories. Everything that could have happened did happen. At least somewhere. There are different views of how these parallel universes might differ from one another, forming complete taxonomies of universe types.

It’s interesting to consider what kind of experiments could be conducted in order to prove that these realities exist. But it overlooks the fact that we interact with parallel realities all the time. Worlds that obey different physical and logical laws. Worlds that have their own unique landscapes. And worlds that share a geography but act out alternate histories. Worlds that many of us visit on at least a daily basis through readily available portals.

At the time I’m writing this blog post there are currently over 500,000 people playing DOTA 2. That’s more than the population of Manchester. The number of people currently playing the survival games Rust and DayZ is roughly the population of Bath. There are live stats available from Steam. The peak concurrent users for Steam today was 7.2m people. The fact that games are popular is not news to anyone, but that’s a lot of people visiting a variety of virtual realities. And its fun to think about them as more concrete spaces and consider the different ways in which we access them.

What follows is some follow-my-nose research on different types of game worlds, largely biased towards games that I’ve played or are familiar with in some way.

The Lifetime of Pocket Universes

What should we consider to be a separate game universe?

A game server, which might be multi-player or single player, is a portal for accessing at least one virtual environment. Some game servers host a single persistent game world — or pocket universe — that will stick around for the lifetime of the server (barring system admin interventions). Other game servers will provide access to many, short-lived game worlds. Some may persist for only a few minutes, others for longer.

For example most first person shooters cycle through game worlds that last for around 10-15 minutes. But some offer a more consistent environment: all DayZ Standalone servers host the exact same map (Chernarus) but the game clock and state varies between servers. Its possible to jump between servers and appear in the exact same location but at different times.

This is something that has been limited in recent updates to DayZ because players were travelling within the Chernarus multiverse to get unfair advantages on other players. E.g. looting the same location across different servers, or getting the upper hand in a fight by flanking someone by jumping between servers. It’s been restricted by placing increasingly longer wait times for people hopping between servers. Although I found it to be an interesting game mechanic and I’d love to play a game in which it was a central motif.

While at any one time there may be multiple servers hosting a copy of the same game world, there are often differences. Time zones are one, but enemies and loot may also spawn randomly, meaning that while the physical layout of the worlds are the same, their histories are different. Player actions are obviously significant and some game worlds offer more opportunities for permanently affecting the environment, e.g. by building or destroying objects.

At the extreme end of the scale are game worlds that are based entirely on procedural generation: no two pocket universes will be exactly the same, but they will obey the same physical laws.

Number of Pocket Universes

It’s hard to get decent stats on the number of game servers and their distribution, some of the details are likely to be commercially sensitive. This is one area where I’d like to see more open data. Its not world changing, but its interesting to a lot of people.

The best resource I could find, apart from the high level Steam Statistics, was Game Tracker. This is a service that monitors game servers running across the net. Registered users can add servers to share them with friends and team mates. Currently there are over 130,000 different game servers being tracked by their system, spanning 91 different games. This will be a gross under-estimate for the size of the gaming multiverse, but is a useful data point.

There’s some interesting analysis that could be done on the distribution of those servers, across both games and countries, but unfortunately the terms of use for GameTracker do not allow harvesting of their data. Being able to locate game servers in the real world tells us where those universes intersect with ours.

Of course there are also some games in which there is only a single universe, although its state and geography is split across multiple game servers. Most MMORPGs operate on this basis, with Eve Online probably being one of the more interesting. If only because it has a time dilation mechanic that kicks in when lots of players are co-located on a single server: the passage of time slows down in a local area to allow all necessary computation to take place. The Eve game world actually spans more than one game. The Dust 514 FPS exists in the same universe and there are ways to interact between the games.

MMORPGs also use “instancing” to spawn off smaller (fractal?!) pocket universes to allow groups of players to simultaneously access the same content in a sand boxed environment. This blending of public multi-player and private single-player spaces within game worlds is part of what is known as “mingleplayer” (which is an awful term!)

Demon Souls and Dark Souls (both 1 & 2) offer another interesting variation and, in my opinion, one of the earliest implementations of mingleplayer. In Dark Souls every player exists in their own copy of the game world, but those worlds are loosely connected to those of all other players. In Dark Souls 1 (and, I think Demon’s Souls) this was via a peer-to-peer network, but in Dark Souls 2 it’s a classic client-server set-up. In all of these games it’s possible to invade or be invited into other worlds to help or hinder them. There are also a number of mechanics to allow players to communicate in a limited and in some cases automatic way between worlds. Typical of the series, there are also some unique and opaque systems that allow items, creatures and player actions to spread between worlds.

Game World Sizes

So how big are these pocket universes? How to they compare to one another and with our own universe? There’s a few interesting facts and comparisons which I’ve dug up:

A number of people have also collected together game maps that show the relative sizes of different game environments:

In Game Statistics

Game publishers collect statistics on how players move through their worlds. Sometimes this is just done during testing and level design in order to balance a map, in others the data is made available to players in real-time to help them improve their game, etc. There was an interesting article on statistics collection in the Halo games in Wired a few years ago. These kinds of statistics collection tools are a fundamental part of many design tools these days.

There’s been some interesting visualisation work around these statistics too. I wonder whether any of this could be applied to real-world data? For example balance and flow maps provide different perspectives on events. And here is a visualisation of every player death in Just Cause 2.