Reading over Brian’s article on the trinity, and (moreso) the responses made both on Gamasutra and his blog, and its clear, there’s a lot of ill-formed thought about the trinity and what’s right or wrong with it. So let’s start with the obvious.

Myth: The Trinity exists to justify healers.
Fact: The trinity exists to justify the tank.
Healing is not less necessary in a trinity – in fact, it is moreso, as anyone who has done warsong gulch or one of the ‘aggroless’ boss fights in WoW knows. It is also much, much more challenging to pull off, as your healing now has to hit unpredictable targets, including ’soft’ targets like squishy mages.

On the flip side, it is almost impossible to justify a defensively-bent damage-soaking meat shield without some artificial help. Most designers would tend towards having these guys do less damage than an average DPSer, and even if they did comparative damage, it always makes tactical sense to kill soft targets faster than hard targets. Which is to say, the problem with making defensive-minded classes is that they will almost never have aggro – which kind of undercuts the heroic vision of the defensive-minded class. If you’re playing a walking tin can, you WANT to be taking hits.

Myth: The whole trinity thing would go away with perfectly realistic AI.
Fact: Perfectly realistic AI sucks.
Here’s a secret – it would be trivial for designers to code monster AI that did the smart thing every time. Kill the healers, then kill the DPS, from highest DPS or squishiest to tankiest. If someone casts crowd control, don’t let them cast it twice. Ignore damage output, and target troublesome classes. And above all, don’t divide your damage – have the whole group of enemies burn down one target. Yes, this strategy is effectively what PvP groups do in the arena.

The problem is: so you think its hard to find healers now? It becomes 10 times harder when the healer experience is looking at your corpse on screen for 90% of the fight (this is, incidentally, the experience of most newbie healers in the arena). Take healers out of the equation, and you move the problem to whichever DPS is squishiest. Which is to say, the problem with realistic AI is that it’s NOT FUN. And fun is still, you know, kinda the point.

Note: many other games in many other genres have AI that would be considered borderline retarded if a player played that way. This includes shooters, strategy games, RPGs, GTA, etc, etc, etc. Stealth games are the worst offenders – the guards stop looking for me if they don’t find me after 20 seconds? The role of AI is not to be realistic and not to crush the player, but instead to challenge the player and be ultimately a solvable puzzle. (I’ve written about this before: You don’t want realistic AI) The point is not that the trinity is the only way to solve this problem, but rather that giving ’smart’ AI to the mob that bypasses the meat shield is going to result more in frustration than in interest. Take out the trinity, and another ‘puzzle’ needs to take its place.

Myth: Collision would get rid of the whole artificial ‘tanking’ thing.
Fact: Um, no.
The whole ‘you can’t step past the tank’ idea only works in tight corridors – specifically corridors that are one man wide. Great if you’re making Diablo. Lousy if you’re making a game set on wide open worlds. Or boss fights in wide rooms. Or, you know, a dungeon crawl with wide corridors.

Then, if you want to be a realism nazi, there are further questions. Can a flying creature just hop over the tank? Can a giant step over him? Can the tank still block the corridor once he’s knocked on his ass?

And then you have the thief problem, which is to say a backstabbing thief has to be right next to the target to do the dirty deed. The smart thing for any AI to do is to turn around and clock him – and there’s no collision shape to hold him back.

One of the things you have to be very careful of as a designer are solutions that only work in certain instances or locations. While it is good to do so to create interesting puzzles or one-off challenges, it is not a good fulcrum for the AI design as a whole, as it limits the designer entirely too much to the sorts of challenges he can design.

Myth: Eliminating the special roles and effectively making all classes DPS would get rid of the trinity.
Fact: It would also, effectively, make all the classes the same. And class homogenization sucks.
Roughly equal DPS, roughly equal survivability – sounds attractive cognitively, but in practice, players actually prefer when classes are strongly different from each other. This, instead, pushes them to be roughly equivalent in practice. If they aren’t, and no specialty roles are needed, then you will end up with one or two classes being dominant.

We already have a working model of this, by the way. It’s called Diablo 2. It is, admittedly, superior for single-player gaming, but in multiplayer PvE, it’s got roughly the same tactical challenge of a slice of swiss cheese. Oh, sure, its fun to zerg a dungeon, but as a general rule, I always felt less like I was working with party members than I was fighting near them.

Myth: The problem is that most fights in MMOs are party vs. single-big-target.
Fact: Almost all modern MMOs have (and should have) a wide variety of fight-types.
There’s a huge misconception that most boss fights in MMOs are ‘tank-and-spanks’ – the party vs. one big boss monster. In practice, these fights are considered simplistic by both players and designers, and have been on the decline by designers for a while now. Most newer MMO boss fights tend to have waves of additional trash (’adds’) or have multiple bosses which requires multiple tanks (or one very savvy and tough tank). And, of course, most ‘trash’ is also multipulls as well.

Whatever your AI puzzle, you want it to be as flexible as possible. You want it to be able to handle one big boss, a ‘council’ fight of several smaller bosses, or waves of hundreds of little scrubs. Tank-healer-mage, in the hands of a good design team, has that flexibility. Diablo 2 doesn’t, and it forces sort of a sameness into their encounters.

Myth: Creating more special roles is a way around the trinity.
Fact: Works great in raid situations, is problematic in smaller groups.
Designing boss encounters in a role-playing game is incredibly tricky, since you don’t know what character (and by extension, what combat capabilities) the party brings to the table. Take, for example, curses in WoW. Only a couple of classes can dispel curses. But a five-man group in WoW can’t count on having one of those classes in them. As such, the desigers have to design their boss fights in such a way that curses are minor headaches that can be healed through or worked around. By contrast, it is reasonable to assume a 10 or 25-man raid will have a curse dispeller in the group, so the designer can put in far more lethal or devastating curses that demand immediate attention.

As a general rule, these specialized roles create much more interesting and varied fights, largely because they create -unusual responsibilities-, which help change up the player experience. Yes, usually as a warlock you DPS, but on Leothinas you need to ranged tank (he can’t be tanked from melee range). Yes, usually as a priest you heal, but on Iron Council, dispelling this debuff before it blows is a higher priority. The problem on five-man is that you can’t count on those roles being filled.

Tank-healer-DPS is simplistic, true, but more to the point, it allows the designer to create more interesting five-man encounters by making a pact with the player: “hey, if you bring a competent tank and a competent healer to this party, it’s doable.” It makes it easier not only to design encounters, but also for players to create groups without knowing the dungeon intimately – which is good.

Myth: The real problem is that there’s not enough tanks and healers.
Fact: Yes, but this isn’t just a tank-healer problem.
Anytime you have varied roles in an MMO, it carries this inherent risk – the risk that I can’t have my fun because the personnel mix on my server isn’t 60%-20%-20%. But saying this is a trinity problem is misleading – it’s a problem based on having any sort of unique roles. And it could certainly go the other way – if 25% of the server were tanks, you’d have tanks grumbling how there isn’t enough DPS. If you had no tanks but your fights depended lots of crowd control, you’d just move the problem there.

But the problem can exist outside of the trinity system – and does, even in WoW. As an example, when I was last doing lots of pug dungeons in WoW, there were too god damn many rogues. At the time, rogues did great DPS, but crap utility, whereas most other DPS (mages, shaman, warlocks) brought better buffs and crowd control to the table. Many groups didn’t want to run groups with two rogues, which meant that rogues had trouble getting dungeon runs due to class demand not matching player class preference. The trouble finding tanks is just another facet of this problem.

Original comments thread is here.