Metcalf's law holds true for mmogs. The number one reason why people stay with a game is because of the other people playing it. The reasoning behind Metcalf's law is that for every new customer there is another potential relation for everybody else. By this logic a game with more subscribers is a superior product (not necessarily a superior game) to one with less.
However, is a large all encompassing community on the lowest level better than multiple smaller sub-networks? Metcalf's law doesn't say anything about this, so lets look at it a little. First of all, sub-networks will develop anyway. You don't spend an equal amount of time on the phone with everybody in the world. Some you talk to a lot - most of them you never call. But the phone network wouldn't be as valuable if the possibility to call anyone was removed.
Communities will often develop in a layered model. Should the developers be able to define the parameters of one or more of the layers? E.g.: how many people they want playing on the same server? By playing the game you've already entered the sub-network of those who play that particular game - a subgroup of those who play mmogs (a subgroup of those who have an Internet connection). I'm not thinking of the "put up a new server when the existing ones can't handle any more" load balancing we currently see in most games. I'm thinking of reasons other than software and hardware requirements.
There are many ways to calculate how many people you want playing on the same server; you have to look at the issue from multiple angles. What looks good from one point of view looks ridiculous from a different one. I'll look at it from two angles: the needs of the player and the needs of the game.
How many relations can you keep up at the same time?
The obvious thing to do would be to look at real life. How many people do you interact with daily? Most of us would have a reasonably short list: The people we work with, friends, and family. These and the people at the grocery store and the gas station are probably the only ones most of us interact with more than once or twice a week. And even if we expand the timeframe to weekly the number of people we interact with doesn't grow by much. Looking at monthly or yearly the numbers will grow though, most of us will meet people we interact with only once or a couple of times a year (or lifetime). But all in all, most of us have a reasonably small circle of people we interact with on a regular basis.
Let's pick a number; let's say you need about 250 people to satisfy your interaction needs for a year. You'll probably need a bit more than that to create immersion, but not that many. The faceless masses can partly be represented by NPC's.
However, we interact more readily on the web right? Well, not really. How many unique web sites do you think people visit in a week? According to one study we only visit 20 different sites pr week on average. Yes, I know some of you visit 50+ sites a day - but that's because you're browsing pr0n ;-P
In one game I was an active member of a guild that peaked at more than 2000 members. I left it for a guild that had about 500 members - and then left that guild for one with 75 members. What is interesting is that the number of people I interacted with on a daily or even weekly basis hardly went down. However I noticed a huge decline in the number of people I interacted with rarely or only once. There is a cap on how many people you can interact with in a single day - the number of people you want to interact with is probably even lower.
How many people do the game need?
Then why do we have all these games that are supposed to have 50k, or even 500k, players on the same server? What's the point? One reason is that you want to create a heterogeneous society in your game, if for no better reason than to attract more customers. So, if we assume my previous group of 250 is homogeneous, how many such groups do you need on a server? Obviously you can't tell the players how to play your game, so there will be groups of 25 and groups of 1000 - but 250 is probably above the mean.
The answer could be wildly different for different genres. It will fluctuate somewhat within a genre as well - but if you look at a few examples, certain numbers manifest themselves as more sensible than others. You'll notice that at 500 players there's already a 50% redundancy from a single players point of view, but you have only 2 homogeneous groups.
What with 2500 players? You'll have 10 groups of 250; this is pretty good - maybe not a heterogeneous society, but you probably cover all the basics. However, from a single player point of view 90% of the players are redundant.
If you double the number of players to 5k, you'll also double the number of groups to 20 - in most games this will be the equivalent of a heterogeneous society. You might not want to have more roles than that, even if you can. You can have only so many different roles before it becomes confusing. But from the view of a single player 95% of the population is redundant.
If you look at the goal of some of the games currently in production:
50k players on the same server: 99,5% of the people on the server is redundant to any single player. You can have 200 homogeneous groups of 250 people - chances are that many of these will be fighting over the same niche in the game. This might or might not be a good thing, depending on the goal of your game. However, the main problem with this many people is that you (currently) can't scale an event for more than a few hundred people - giving people the feeling that everything happens when they're offline (at most 1% of the players will be involved in any given event, leaving 99% wanting more).
500k players on the same server: 99,95% of the people on the server is redundant to any single player. You can have 2000 homogeneous groups of 250 people - a lot of these will not have an identity within the game. This is probably not a good idea, one of the reasons people play online games is because they want the illusion of being able to influence the world, how will you provide this illusion for 500k people?
See where I'm going? More players in a game add value to the game. More players on a server add value only to a point, after that point they can either be ignored or they will subtract value. The positive interaction with the additional people is equal to or less than what can be provided by a message board. There is no reason to have everybody in the same group on the lowest level; you don't break with Metcalf's law by having multiple subgroups.
Do the people who are not part of your relations add value?
When nine out of ten people on your server are not part of your relations, what purpose do they fulfill? What is it about them that make it preferable to have them on the same server as you rather than on a server of their own? Well, for one, they add background noise. If you removed all the people you didn't know from the main street in your city you'd get a slightly different feeling walking down it. The faceless masses add depth. But, depending on the game, they can add or subtract from your game experience in other ways as well. If a game doesn't have enough content to support the number of players it has - they will frequently end up being between you and an entertaining experience. In many games they have developed artificial means of solving this problem, e.g. you have to sign up on a list and then wait your turn to access content, making playing the game about as fun as going to the doctor. You might even have to schedule and reserve content weeks ahead. Unless there's an abundance of content, there is very little room for improvisation.
But if there is an abundance of content, what then? Then you have a different problem. Economics. Economics is the study of the allocation of scarce resources. Content is a very important resource for any game, so you want to have an abundance of it. However, whatever resource you want to build an economy around will have to be derived from that content - then how are you supposed to keep it scarce? Especially when content as a function of time is linear - and if there's an abundance of content, it will be linear as a function of players as well. There are many ways (most of which require more work and testing than most developers are willing / allowed to spend) to do this, but that's a different discussion.
Do all players have a possible role in the bigger picture?
Currently the answer to this question is easily no. Experienced players have very little to gain from interacting with newbies. New players don't really have much of a role at all; often they don't even get the illusion that they matter. This will probably change a bit in the new games, getting the feeling you matter a little in the game is almost as good a way to keep retention up as a good community. It is also something the developers can work directly toward, and not just facilitate for and hope it turns out for the best.
If you increase the size of the world to allow for more players, what else do you also have to change?
Different amount of players require different types of gameplay, and vice versa. Companies are becoming more aware of this, but they have no idea where the breakpoints are. E.g. if you want a full pvp game (like Shadowbane) - how many players do you require? It depends on the features of the game and your vision of it. Currently a number of servers in Sb have one guild or alliance that dominates it more or less completely. Why? The number of people on the server doesn't reach the point where it becomes next to impossible to organize a stable overwhelming force. A guild or alliance with more than 2000 members will usually implode because it is hard to organize, you can't keep everybody happy, and you will get internal conflicts. If they had managed 5000 players pr server there probably would not be one guild that dominated the entire server for an extended period of time, each guild would have to settle for a piece of it. Maybe it wouldn't have made any practical difference, but I think it would.
On the other hand, what if they had managed 50k players, wouldn't that be even better? It could be, but most likely it wouldn't. With this many players only a very limited number will be able to find any kind of safety. The result will be the same as if there are too few players - the powerful players will be able to cull the weaker ones easily. But now it is not because they can't be touched - but because there will be too many weaker players for them to organize and protect themselves properly (the weak players are more likely to prey on other weak players than to stand up against the more powerful ones). To put it differently: there will always be enough weak players around to keep the powerful ones happy (assuming they actually have 50k players).
Not to mention it's a lot easier and faster (and cheaper) to generate enough gameplay and content for 5k players and duplicate it on 10 servers than it is make something that'll satisfy 50k players.
With a pve game like EQ you only need enough people to cover all the roles in the world. 3k people could be more than enough, since what you want to do is to encourage teamwork. In addition less people makes it easier to balance high-end encounters.
There are currently games in production that require 500k or more players for their "endgame" to be viable. This seems a bit stupid. What if they get "only" 250k players? Now the endgame doesn't work because everybody has access to everything. Scarcity of goods is nonexistent; there is no shortage of resources - no economics. There is nothing to fight over. Having nothing to fight over at the endgame is just as bad as having too many people fighting over the same few items. If reaching the goal gets too easy it will no longer be a goal. As long as you're going to have an "endgame", it better have some replay value.
Not to mention - how are they going to handle the character data? Already most games only save your character once every hour or so. Everything else that happens is handled purely in memory (the number one reason for time warps and dupe bugs). And this is not only a database problem - there are bottlenecks in hardware and the operating system as well. And I very much doubt that these people have found a revolutionary new way of accessing databases. I know there are solutions out there that could allow them to get an acceptable performance (as in; less than ten times worse than the current games), but those will be very expensive to set up for this many concurrent users.
If you scaled for 5k you can just put up more servers, and the gameplay will remain more or less the same. Scaling up is easier, although maybe a bit more expensive, than scaling down.
These are obviously not the only things to consider when you decide on how many players you want on a server, but I think they are important and often ignored. At least hardware and software requirements should not be the only determining factors when you calculate server population.
