Preservation Evoker in Mythic+
Heal the party before the damage lands. Other healers watch a health bar drop and answer it. You can play Preservation that way, and it is the weakest way to play it.
Put heals on people who are still at full health. Dream Breath and Reversion keep paying out after you cast them, so a healthy target banks the healing instead of wasting it. Echo works differently. Its own heal is instant, and it then sits on the target waiting to duplicate whatever you cast next.
What you press
Living Flame is the most pressed ability you actually choose. It is one button that either heals an ally or damages an enemy, and a cast record cannot tell the two apart. Disintegrate is next, and Disintegrate is always damage.
The shape of a key
A run is roughly half an hour. Every Season 2 dungeon has a median length between 26.8m and 32.8m.
The pull count is the number that matters. Each dungeon has 12 to 21 pulls that recur often enough to be worth learning by name. You do not need all of them to start.
Some dungeons hit harder than others. Ruby Life Pools has the worst recorded five seconds of party damage and Temple of Sethraliss the mildest, so that is where your practice time goes first. The figures, and the gap between them, are on the dungeon comparison. Every damage figure on this site is the whole party added together, mostly landing on the tank, so none of them is a healing target.
Pull lists, with what was cast before each spike: dungeons.
Ramping
Learn the set of abilities, not the sequence. The set holds up across runs. The order does not. Take one recurring pull in Kings' Rest: the ability that appeared in every single recorded ramp for it opened almost none of them, and no ability opened more than a handful. Which buttons appear is stable. Which one goes first is not.
What appears where, pull by pull, with the counts behind it: ramping.
Damage
Keep casting damage while you ramp. Once procs are excluded, damage is 0% of the things you actually press in the 30 seconds before a spike. Disintegrate and Fire Breath carry most of it.
What to press and when: damage.
Cooldowns
You store Stasis far more often than you spend it, at least inside the window that gets recorded. The release may simply be landing outside that window, so the gap is not proof of wasted charges. Obsidian Scales and Tip the Scales turn up in most runs.
Everything known about what players bank in Stasis comes from a small number of uses in two runs, both Kings' Rest. Read the caveats before acting on it: cooldowns.
Talents
Most of your tree is not a choice. 56 talents appear in 95% or more of recorded builds. That figure is overall and is not broken out by key level.
22 talents do move with key level, and some move a long way between the lowest band and the highest.
Every band percentage: talents.
Flameshaper and Chronowarden
36 of the 36 runs with a build recorded are Flameshaper. 0 are Chronowarden.
Every number on this site therefore describes Flameshaper play. Chronowarden abilities are correspondingly thin. Do not treat those as measured.
This is not evidence that Chronowarden is bad. These runs came off a PTR build and a World of Warcraft Logs leaderboard. That is the top of the spec rather than a cross-section, and the top of a spec converges hard on one build. A handful of runs cannot tell you whether the other tree is competitive.
Where to go next
How these numbers were measured
Runs come from Warcraft Logs. Preservation Evoker only, Mythic+ only.
Only one report counts per key. Two players in the same group each upload the same run, so the raw data holds that key twice. 27 of 352 logged rows are a second copy of this kind, and only one of each pair is counted. That is where 325 comes from.
Runs whose event stream was cut short are excluded, because their totals would be low by an amount nobody can measure. None have hit that yet.
A pull is listed only if it appears in at least 60 percent of that dungeon's runs. What you see is the dungeon's spine, not one group's exact route. If your group skips a pack listed here, or takes one that is not, nothing is wrong.
An ability is called part of the ramp only if it appears in at least 60 percent of the runs with ramp casts for that pull. One player's habit on one key does not reach the page.
Casts are only collected from the 30 seconds before a damage spike. That is a setting, not a finding, so nothing here can tell you whether a ramp starts earlier than that. Where you see a window, it is the window we looked in.
Ramps are worked out one run at a time and then compared, rather than by taking each ability's average timing separately. Doing it the second way produced ramps of six casts inside four seconds, which is faster than the game allows anyone to press buttons.
Damage figures are the whole party added together. Nothing here records who took the damage, and most autoattack damage lands on the tank. Use these numbers to compare one pull against another. They are not an amount you have to heal.