Misconceptions on Instancing: Jason Booth

Misconceptions on Instancing

Brad posted a bit of a manifesto about instancing, in which he basically claims that instancing != MMO != Vangard. Lum fires back with some salient points. I’m only going to talk briefly about why Brad’s points are incorrect, and instead spend most of this post focusing on many of the things you can do with instancing, and many of the reasons that instancing is a valuable game design technique. But first, the rebuttal section:


“Instancing is a form of player contention management”


While I can see how someone would think this, it’s not really true, and it’s not really the prime reason for instancing. Much has been said about EQ’s camping problems, so I’m going to assume you’re familiar with the same type of player contention Brad is talking about. We’ve seen those issues greatly reduced by features such as distributive camping (AC1) and round robin looting (Guild Wars), and guess what, even in instanced games you still need many of those same solutions because many of those contention issues still exist (ninja looting).

So, instancing doesn’t really provide much in the way of contention management except in cases where you’ve placed an extremely rare item on an extremely slow dropping spawn in one location. While Brad implies at times that this type of design makes the reward worth more, I’d personally feel that it just makes players resent you and your design more, and that the cost is not worth it. I consider it bad design to make the player hate you.


“Instancing is not MMOG!”


Massively Multiplayer Online Game. Hmm, I see nothing in this term that negates instancing, or implies some type of Raph Koster style world model. All I see in this definition is an implication of having more players online playing together than in a traditional online game. There’s no mention of persistence, either, btw.

Now lets consider a traditional MMP server. Like all computers, it’s limited by a combination of CPU speed, Memory, and bandwidth (both internet and between given CPUs). Even if the client could render any number of players (which it can’t), the server can rarely handle over a certain number of players in a given area. The slower and less interactive your game is, the higher that number is likely to be. Usually, with good performance, you’re looking at 60 to 100 players in close vicinity to each other tops. With worse performance, maybe 200.

Add physics and real game play and that number drops considerably, since both of those usually have an n^2 performance cost on the hardware.

You have the same problem on the opposite end on the spectrum though; if everyone is equally spread out on a server (ala ac1’s darktide server), the server has to keep track of a much larger play space, and needs to load a lot more data. Suddenly your running into memory problems instead of CPU and buss contention issues.

So, to sum up, the ideal case for an MMP server is a reasonable number of people in each area. Not too many, not too few. To accomplish this, we create designs that encourage these numbers to exist; and sometimes enforce them with hard limits (AC1’s portal storms, for instance). Sometimes we let performance handle it (it’s too laggy in this zone) for us. Most of the time, these numbers work out to be not much greater than a 64 player battlefield 2 server.

So, how massive is MMOG really? Is it really different than interlinking several BF2 servers together? That’s basically what EQ’s server structure was, hence all the zoning. Remove physics and reduce game play speed, and I’d bet a BF2 server could hold quite a few people. So in my mind, there’s very little difference between an instanced based game that allows 60 players in an area and a traditional MMP.

So Brad, if you want people to stop referring to Vangard and Guild Wars using the same term, you should choose a term which doesn’t apply to Guild Wars to describe Vangard, because I see nothing which makes Vangard any more MMOG than Guild Wars, by definition or implementation.


The true power of Instancing

The true power of instancing in an MMP space is the power to control the difficulty curve and put consequence back into the game design. The latter should be very appealing to those in the Raph Koster camp of world models; because that is by far the hardest thing to have in a traditional MMP environment. Lets examine both of these for a moment:


Difficulty Curve

In a traditional MMP, the difficulty curve has to be controlled on a very small scope. That is, generally, the to-hit equation. As long as you can hit it, you can kill it. Get enough people together and even the toughest monster goes down.

Often games use a ton of hacks to work around this issue: con levels and experience rewards being tempered by them, many on one end-game models, and brutal to-hit equations which make combat feel terrible (miss… miss… miss.. I see my sword going right through him!).

Rewards also must be tied directly to things like damage. You cannot reward someone for sneaking past a monster, healing a party member, etc. Instead, you come up with alternate grind ladders or complex equations to try to balance out this fundamental flaw.

But in an instanced environment, you can remove any or all of these problems. Want a tight limit on the number of players? Want to reward based on smart play instead of just killing? Want to use a bunch of low level creatures and swarm the player instead of a single high level one and still provide a decent reward? Want to blow up half of the dungeon when someone pulls a lever? No problem. You have complete control over the game experience.


Consequence

Now this is the real crowning jewel of MMP design. You hear about it all the time; players want to change the world; they want to have effects on each other; but we can’t allow them any significant change without ruining someone else’s play experience. This is, to me, where next generation instancing can really shine, and the per-adventure versions of it we’re seeing now aren’t really exploiting this potential to the fullest.

The server which runs a game world can spawn off an instance of any piece of it. But why not let the user spawn off the whole thing? Get a guild of reasonable size and spawn off your own copy. Want to kill off the NPCs and have them not respawn? Go ahead, it’s your world, destroy it if you want to. Want to chop every tree in the land down and turn it all into desert? Fine. Want your copy to be PvP with alternate rules? Just check the options and go.

With instancing, you have the ability to open up your game designs in ways you cannot normally. While you still need to have a main section of the world in which the players actions are less evasive on each other, groups of users can easily create parallel versions of the world with drastically different and diverging properties. Let them scorch the earth, or restore wonderment to it. Let them control how non-guild members can interact with their world. Give them the ability to affect their copy, and play with the consequences that result. If it fails, it’s not like they’ve ruined the game for everyone, or even themselves. Just their little copy of it.

In my mind, this is where the really interesting ‘world design’ type of stuff can happen. A traditional MMP is far to limiting in this sense, and we’ll always be hampered by the lack of consequence in a traditional MMP model. It seems to me that those in the ‘MMP as world’ camp should be begging for this type of freedom, and dying for instances to take them there. You’re going to have multiple servers anyway, so the only difference here is that you’re allowing people to move between them, privatize them, and affect them.


Implied Grouping

There are also many simpler instancing types that people haven’t taken advantage of yet. Here are two examples:


Boat Ride
Say, every 3 minutes a boat leaves from the harbor. It’s really just a teleport to another area, but you buy your ticket and have a little ride first. Maybe the boat gets attacked by pirates and the people on the boat have to defend it. Maybe the boat springs a leak and you have to bail it out. If you succeed, you get where you’re going and gain a small reward, if not, you all wash ashore where you got on.

What’s happening here is a form of instant and implied group, and it’s not much different than joining a side in a Counter Strike game. You don’t stand around in CS doing LFG tells, so why should you have to in an MMP?

Further more, there’s no reason the system cannot pull people from multiple boat rides into the same instance. Does it matter that we’re not being teleported to the same place at the end of the instance? No. What you really have here is a method to instantly start a new game every 3 minutes with a bunch of players without them having to specifically agree to play the game together.


Defend

A similar concept to the above would be to have an NPC which asks you to defend and outpost for a certain amount of time while re-enforcements come. An onslaught spawn causes mobs to run towards the defense point at a rate controlled by how many people are in the area. Players can come to the area at any time, and the system can spawn off new copies once the first one is full. Instant, no fuss grouping. Just jump in and play.


Non-Diku Mud gameplay


Then there’s the whole realm of possibility which opens up when you ditch the fantasy Diku-mud game all together. The play space of a game doesn’t have to be a 3d landmass; it can just as easily be mapped over the social space of the player. I did a pitch for a game back in ’98 or so which user the players social network as the exploration space, by modifying the existing play space according to social networks.

I got the idea from a friend of mine who was going to various concerts and making tails for his friends to wear. Yes, animal tails, but no, he’s not into furry. I don’t hang out with those sorts Anyway, each tail had a small tag on it which had a quote from the last person who received a tail. This way, each person was connected to the next by their quotes, and all of them connected to him visibly by the tail. A while later I saw someone at a show wearing a tail and asked if they knew my friend; before long I had a place to sleep at night. To me, that was a cool social networking feature worth exploring.

So I got the idea for tags and tails. Basically, the idea is this: a user can create an object and control how it is copied. Maybe if I give you one you can only make one copy of it, or it only spreads one degree from me. A bunch of simple options control how this object is spread through the community. Now, this object has a special ID associated with it. This ID might act as a key to a social space; my club or home, for instance. It may be like a guild emblem, free for all to see which club I belong too. However, I can also activate that ID to have it shape the world around me. Suddenly there’s a door where there wasn’t one before, the decorations in the room are entirely different, and lum is now standing their in bondage gear with a whip in his hand and a “Democrats Rock!” sticker on his forehead. Then I begin to wonder what the world would be like if I put on the tail Brad handed me earlier.

In this type of model, user control over a given space is of the utmost importance. And laying out a large contiguous world is actually a disadvantage, because the world is expanded in parallel instead. A layering of realities, if you will. Each of these is an instance of the world, which may or may not be running in parallel with other instances of that space (users interacting through two instances). It’s a wacky concept, but conceptually very cool, as any one who’s browsed friendster or myspace can tell you.


So, in the end, instancing is just another tool; and one that we’ve barely begun to use. If designers want to take some high road about certain tools as “not MMOG enough” for them, then they lock themselves out of potentially cool options within their products. The MMP space is very young (even if you count muds and such) and I hate to see designers trying to squeeze the space down to a minimalist set of approaches so early in the process. We should be open to new approaches and ideas, not closed to them and touting some kind of elitist message about which techniques are valid. No offense, just my opinion on the matter.