Our original design for multiclassing was based on D&D 3.X's approach, where players pick classes level-by-level. You could build a character as a wizard for 4 levels, then add 2 levels of fighter, then switch back to wizard for 3 levels, and so on. Because 3.X multiclassing has some mechanical problems with how spellcasters progress (barring certain prestige classes like the Mystic Theurge), I designed a system that separated each class' power source advancement from its level advancement. It was designed to create a more reliable advancement curve for each class' abilities. It more-or-less succeeded at that, but we noted some problems early on, including after the announcement:

• People had a difficult time understanding the relationship between individual levels and power source progression.
• Selecting powers worked pretty well if you knew exactly what you were doing, but you could easily wind up with casts per encounter from levels of spells you had no spells for (e.g. 2 third level casts but no third level spells).
• There were some shortcomings with building characters level by level in different orders.
• Displaying power source progression in the UI was either confusing or took up an inordinate amount of space.

After talking things over with the other system designers, we discussed what the most important aspects of multiclassing were.

• Allowing people to realize hybrid class character concepts. "I want to be a fighter and a wizard."
• Keeping the overall power of the character competitive with single-class characters. The character should be viable. It's okay if it winds up over- or a little under-powered compared to a single-class character as long as it's not fundamentally weak.
• Allowing players to emphasize one aspect of the hybrid more than others. "I'm a fighter and a wizard, but more of a wizard."

The original design allowed the first and the last aspects, but the middle aspect suffered because of the high degree of flexibility. It was still easy to make non-viable characters. A non-viable character can be part of a viable party, but still feels bad to play. The high degree of flexibility also strained the first aspect, the basic character concept. A character with 18 levels in fighter and 2 levels in rogue is less of a character concept and more of a strategic build choice.

I went back to the drawing board to revisit an idea I had around the same time as the original design, which was based on AD&D 2nd Edition-style multiclassing, where the player chooses to opt into multiclassing at character creation instead of selecting classes level-by-level. In such systems, the core concept is established from the beginning. A player who says, "I want to be a fighter and a wizard," can be that (a battlemage) from the beginning instead of picking one class and then alternating to the other later on. Progression is also easier to understand from the beginning as access to abilities and the increase of their power is consistent from multiclass to multiclass. A fighter/rogue (swashbuckler) gains access to 2nd level abilities for both classes at 4th level, as does a priest/monk (contemplative), barbarian/chanter (howler), and druid/ranger (beastmaster).

In a strict sense, the new system allows for less overall flexibility, but multiclass characters now get two abilities each time they hit a new power level, one from each class. This means that a multiclass character starts with more abilities and will always have more abilities than a single-class character of the same level. However, multiclass characters get access to each power level later than a single-class character and their abilities progress in power at a slower rate. A wizard has access to Fireball at 5th level, but a battlemage (fighter/wizard) does not gain access until 7th level.

A multiclass character also uses the average of their classes' base health and defenses. If one class begins with 48 health and the other begins with 30, the multiclass will start with 39 health. If one class gains 14 health per level and the other gains 10 per level, the multiclass will gain 12 per level.

In actual playtesting, multiclass characters have terrific flexibility even though they lag power-wise compared to their single-class counterparts. In my personal experience, they are a lot of fun to play and there is a huge amount of variety to how a multiclass character can be built even before subclasses are taken into account.

Subclasses add an additional dimension to character conception and development. As in our original design, players are allowed to choose a subclass for each of their classes. The only classes that are required to have a subclass are paladin and priest. Subclasses all have trade-offs, though some subclasses change the core playstyle of the class more than others. A sharpshooter plays similarly to the pure ranger, but emphasizes the ranged aspect more and suffers more in melee. A stalker needs to stay close to their animal companion to avoid penalties and take advantage of the subclass' melee-oriented benefits. A ghost lodge ranger plays much differently from the pure ranger because their animal companion is only present as a spirit summoned in combat. There are very few restrictions on multiclassing combinations. Only a few paladin orders and priest deities are restricted from combining for mechanical reasons (i.e., contradictory Dispositions that affect their abilities). Otherwise, the player is free to combine classes as they see fit.
​