Preview material for Pillars of Eternity 2 got me thinking about multi-class design in RPGs.
Designing any class-based RPG involves putting together mechanically unique sequences of capabilities, and then trying to balance them all against each other using some more-or-usually-less formalized definition of balance. This is a tricky thing to do! Especially from the perspective of each class either embodying or being distinguished by some kind of “gimmick”, either mechanical or stylistic.
Multi-classing is, in effect, a (hazardous, I argue) shortcut to creating more classes. If all you have is fighter, thief, wizard, cleric, then introducing any kind of multi-classing rules more than doubles the number of classes available to your game, even if the only kind of multi-classing allowed is taking two classes.
The alternative to multi-classing is to invent a new class (or potentially several) that bridges the space between two existing classes, and therefore invent a new gimmick for that class. The oldest Elf, for example, was some times a fighting man, and other times a magic user. Adventurer, Conqueror, King has hybrid fighter-wizard elves, fighter-cleric dervishes or dwarven craftspriests, and fighter-thief explorers. The paladin is the traditional fighter-cleric hybrid, the bard a thief-wizard, and the ranger a fighter-thief, each with something unique to that class not originating from any of its parent classes.
With multi-classing, you can directly build your fighter-cleric or your thief-wizard, but (unless your multi-class system is designed to do it) with no abilities not taken directly from the original classes. This does reduce the amount of design work necessary, but in exchange the multi-classing system and each existing class needs to be carefully designed to avoid unplanned consequences. This is arguably more work, since for a game that offers N classes, simple two-class multi-classing means your design space to balance is now functionally O(N2) (yay combinatorics!), and in the worst case (3E, that’s you) O(N!) and made even worse every time a new base class is added to the game.
Some Looks At Other Systems
Like I mentioned above, ACKS explicitly recommends devising custom classes to fill in campaign-unique spaces, and offers several examples that fit in between the core four archetypes of D&D.
Dungeon World gives some classes a multi-class move that chooses a move from another class, in effect a power swap.
AD&D 2E, which I happen to have at hand, has two different systems. For humans, you can dual-class by in effect giving up your first class and starting over with a second. After advancing back to your original level, you can start taking advantage of the capabilities of both. For the so-called demi-humans, the game offers multi-classing by splitting XP between two (or for the half-elf, three) classes. Combined with the game’s different XP totals per class, 10,000 XP could net you a Fighter, Cleric or Wizard 4, a Rogue 5, or multi-classing any of those together at one level lower. A half-elf three-classer could have Fighter or Wizard at 2, with Cleric or Rogue at 3. This ignores kits, which aim to fill in with extra gimmicks for both standard classes and multi-classes.
Third Edition is a mess. Of all the possible multi-classing schemes, buying any class a level at a time is the one that works out least of all.
Fourth Edition again had two multi-classing schemes. The original was a set of feats that multi-classed a character into another class. All they offered was an initial use of one of the class’s gimmicks (multi- into a leader gave one use of the leader’s healing power), with later feats allowing the character to swap in a power from the other class. This is arguably the weakest kind of multi-classing and in my limited experience was almost never used. The upside to it is that balancing multi-classing only involved comparisons with other feats, rather than needing to balance effectively a whole new class. The other approach was hybrid characters, where each class was functionally split in half, and rules created for merging the half-classes together. I had no direct experience using this scheme, but the impression I got was that it created many combinations that came out behind, and a rare few that came out ahead. For the most part, however, 4E went the full class treadmill route, using the spaces on its role-by-power-source grid to fill out concepts.
Fifth I still can’t comment on, except to say that it seems to offer both multi-classing, and full-class options that somewhat fill in gaps, with other full classes designed to fit in those spaces as well. So there are still fighters, clerics, fighter-clerics, and paladins; but fighters, wizards, fighter-wizards and a full fighter build that grant some arcane spell casting.
Baldur’s Gate and its sequels all base themselves on D&D 2E, and multi-class accordingly. Neverwinter Nights works just like 3E. Later editions of D&D games I have not played, but I assume they model themselves on some corresponding edition of the tabletop game.
Mass Effect (to break genre) has an interesting arc across its component games. The first game largely rejected any kind of class gimmick beyond the various powers/skills. Its core three – Soldier, Engineer, and Adepts, each excel in combat, tech, and biotics (telekinesis-only magic) respectively. In ME1, the in-betweeners – combat/tech Infiltrator, combat/biotic Vanguard, and tech/biotic Sentinel each wound up getting a mix of the low-end abilities of their component areas. A full Adept could get Singularity, but the Vanguard doesn’t have an equivalent class-unique power. Mass Effect 2 and later gave the “multi-classes” each an iconic ability that, in my opinion, made them more fun to play than any of the core three.
Pillars of Eternity 2 proposes a multi-classing system that, of all RPG predecessors, looks to me most like 3E, but ideally producing a more even blend akin to the result of 2E multi-classing. On the other hand, they now have functionally 66 different classes to balance, so good luck with that! (It’s actually slightly worse, since a 4th level character could multi-class as A1/B3, A2/B2, or A3/B1, which is why I say it looks most like 3E.)
My Own Design
For my own part, I have a system that has three “classes”, each focused on one of combat, skill use, and magic use. There’s a single point of crossover with the level 6 ability of Versatility, in which every archetype picks up the first rank of another archetype’s gimmick. Since level 6 is the top of the second tier, each character has two ranks of their own archetype’s gimmick, and at the next level will advance in it again.
The other component of my approach is that all of the lifting is performed within the Arts – each skill, each fighting trick, and each spell is defined in a unit of character design that can be chosen no matter what archetype the character is. Instead of needing to define a paladin class, or a system for combining fighters and clerics, a character is chosen to be primarily focused on doing battle, and then chooses the set of Arts (perhaps Channel[Light] and Healing) that grant the styled-as-divine powers appropriate to the concept. The one shortcoming that I occasionally ponder is that there is no real class gimmick – since there’s nothing that’s discretely “paladin” in the system design, there’s no room to give paladins, say, an aura power that’s uniquely theirs. I think I come close to making up for it with the Hero Specialization, where the character gets a unique trait.
Long before I reached this current design, I was playing around with a different shape for classes and multi-classing. In that model, I would have more specific classes – fighter, thief, wizard, cleric, and so on. The various abilities of the classes would be organized into chains of powers akin to the Diablo 2 skill trees. Multi-classing would involve replacing some amount of in-class investment in these trees with another class’s tree, and part of the design would involve deciding which trees were available for multi-classing, and which were too core to the class to allow branching into.