The Resto Decision Tree

Resto Druid theorycraft in a bit of a spot right now compared to say, Holy Paladin. Largely this is due to our very peculiar itemization, where Haste/Crit/MP5 are all 'meh' and very hard to valuate, and Spellpower/Int/Spirit are not things that one can generally 'stack' in the proper sense outside of gemming. Even in gemming the answers are somewhat abstract, and generally come down to 'If you have mana troubles/the socket bonus is good use Luminous, otherwise use Runed'. There just isn't much to quantify in the current healing environment.


On the other hand, Resto Druids do have something that is generally unmodeled: a quite complex decision tree. Past actions influence future results, and what options are available, and their values, at any given time depend on what choices were made in the past.

There are six tools generally available in the resto druid toolbox, ignoring significant cooldowns: Noursh, Regrowth, Swiftmend, Lifebloom, Rejuvenation and Wild Growth. Healing Touch is not included due to various issues which make it very hard to utilize outside of Nature's Swiftness. These six abilities all possess significant synergies with at least one of the other six, two possess slight cooldowns, and all have fast casting speeds, with three generally hovering near the 1s GCD cap due to talents. These factors combine to create a plethora of decisions that a druid is faced with whenever a spell finishes casting, each of them altering the values of future actions.

Take the following examples:

If you choose to reroll Lifebloom on a tank now, the future value of doing so again drops, then immediately begins to rise as the Lifebloom approaches expiration. On the other hand, choosing to not reroll Lifebloom means that the value (and urgency) of doing so will continue to rise until it reaches 0s in duration, at which point you lose, you fail, you goofed. Unless that was your plan in the first place (which it always is when you let a lifebloom expire, right?).

The value of Nourish in the present is dependent on if you've chosen to HoT the target in the past (and with 4 piece T6, how many HoTs you've placed on them). Also, if Nourish crits, it decreases the cast time for the next (via Nature's Grace) thus increasing the value.

The value of Regrowth in the present is dependent on if you have a Regrowth already on the target (via Glyph of Regrowth), and to a much smaller extent how close you are to clipping a tic of the HoT from a previous cast. Additionally, casting a Regrowth in the present also puts the HoT on the target, providing future healing, enabling/continuing the +20% bonus, allowing for Swiftmending, providing a 5% or 20% to Nourish, and like Nourish, allowing for a Nature's Grace proc (at a 46% higher chance) to make future casts of both Nourish and Regrowth faster.

The value of casting Rejuvenation on a target is dependent on whether or not you already have one on the target, the time until tic on said spell, and the duration remaining. The value of casting one on a fresh target depends on future expectations of damage, how smooth/spikey that damage is, while also allowing for 18s of Swiftmend opportunities and providing a 5% or 20% boost to Nourish for 18s.

The value of casting Swiftmend on a target depends on whether or not you've cast a Rejuvenation on the target in the last 18s, or a Regrowth in the last 27s (and is zero if neither of those are true), which of those two is present, the current health of the target, and expectations of near future damage on the target. Casting Swiftmend now takes it off the table of options for the next 15s.

And so on for other options such as Wild Growthing (raid or tanks), single stack Lifebloom for raid healing (topping off or proactive).

I have very little idea how to model this in a practical way (if it can be even done at all), as throwing together a decision tree, weighing each branch, and then choosing the optimal branch, is something that would likely be too complex even for a mod.