Blog > Lhivera’s Library > Mage Issues
Wednesday, September 2, 2009 • 2,473 views • Save & Share

A Ridiculously Comprehensive Compilation

This is a list of problems, both general and specific, that currently affect the Mage class. It was inspired by a couple of recent forum events:

  1. In the Mage Q&A and in a follow-up thread, Ghostcrawler seemed to miss the point of some questions. One example was the question of Frost and Fire specs reaching Torment the Weak: Mages asked if Blizzard felt it was OK for them to have to waste so many points on useless talents to reach the talent. GC did not answer this question, he simply said he didn’t have a problem with the talent being mandatory. The point of the question was not Torment the Weak itself, but the points spent to reach it. So we’re hoping to lay out our issues in a very clear, point-by-point fashion to prevent such confusion.

  2. A poster asked in his own thread why Mages complain so much. It occurred to us that a comprehensive list might help to answer this question.

The purpose of this document is not to suggest solutions; only to track the current status of class problems. It will also tend toward very general issues where possible, so as to avoid bias toward or against any particular solution. For example, Frost PvE performance can be fixed in a number of ways, so specific talents and spells will not enter into the description of that problem.

We make no claim that everything listed here is a Crucial, Game-Breaking™ Problem, only that in a perfect world, everything here would be addressed.

If there is an issue not listed, please feel free to post it to the Forum thread or send Lhivera a private message on the Elitist Jerks site for addition to the list.

If Blizzard is on record as saying something on the list is not a problem, please feel free to post a reference link, and we’ll make note of it.

Contributions and Suggestions by…

Affix
Dagmagi
Faxmonkey
Gottlos
Kyth
Lhivera
nathanbp
Solisa
Zaldinar

…and of course hundreds of Mages on various forums over the years.

Big thanks also to dedmonwakeen for making SimulationCraft, and making it Mac-compatible.

Some Explanations
Problems Key

Some abbreviations are used:

  • ROI: Poor return-on-investment. This simply means the talent has low value per point. In the case of a pure DPS talent, this is easy — less than the “standard” 1% DPS per point. In the case of a mixed talent, it’s somewhat subjective.

  • Func: Functionality problem. This talent or spell does not reliably do what it seems intended to do.

  • Bug: This talent or spell is bugged and does not function as described.

  • QOL: This talent or spell generates frequent “quality-of-life” complaints.

  • Tip: Tooltip error.

  • WTF: Self-explanatory.
Talent Assessments for PvE or PvP

Talents are assessed not based on whether they will fit into a PvE or PvP build, but based on its actual effects. For example, Arctic Winds is considered a PvP-focused talent, even though it often will not be fit into a PvP build, because the +5% damage is equally useful in PvE and PvP, while the -hit% chance is more useful in PvP. This can highlight problems, either where a a build cannot fit a talent that it perhaps should be able to (should Frost PvP builds have more free points so they can take Arctic Winds?), or where a talent should be retuned to be more useful to the spec that can actually use it (should Arctic Winds be retuned as a stronger PvE-focused talent?).

A talent is considered PvE-focused if it has significantly more benefit for PvE than for PvP.

A talent is considered PvP-focused if it has significantly more benefit for PvP than for PvE.

A talent is considered Mixed if it has roughly equal benefit for both PvE and PvP.

A talent’s value can be affected by how likely its usefulness is to come into play in PvE or PvP. For example, the extra crit provided by Focus Magic is nice in both PvE and PvP, but in PvP, dispels are likely to reduce its uptime considerably. More mana is highly useful in both PvE and PvP, but PvP gear carries much less spirit, so Arcane Meditation provides much less benefit there.

These evaluations are obviously highly subjective, and are open to modification after discussion. The purpose is primarily to provide a simple indication of whether a tree’s proportion of PvE and PvP talents is reasonably balanced, which may help to show why a tree does well in one and poorly in the other.

Reference Material

Where possible, useful, and relevant, links are provided to blue posts that provide supporting information for the problems described here, and/or developer positions on the status of these problems. Most links will be to the official forums; in the case of posts made during WotLK Beta, links will be to an offsite archive such as MMO-Champion.

The most recent SimCraft output used in sections of this document may be found here: http://www.manoutoftime.org/misc/simcraft/

Information on the popularity of talents was datamined from the Armory; the latest data used was retrieved on May 13, 2009. Complete data may be seen here:

General Class Issues
Arcane PvE
  • Scaling issues may be causing it to drop too far behind Fire.

  • Lower efficiency and utility than Fire seems like it should be balanced by producing higher DPS than Fire, at least on short fights.
Arcane PvP
  • Ineffective for competitive play.
Fire PvE
  • Too reliant upon random chance.

    Note that the symptom of this problem is not Mages falling behind other players; it is any single specific Mage producing highly variable DPS on different executions of the same encounter. It is possible to gauge, within reason, the consistency of play quality by looking at the number of casts per spell in a given amount of time. If play quality is consistent while DPS output is highly variable, that is an RNG problem, not a play quality problem.

    Blizzard has only discussed this problem as if it were a matter of players comparing themselves to other players, rather than players comparing their own performances against each other; they have yet to address the issue of same-player variation. (Reference)

Fire PvP
  • Ineffective for competitive play.
    Blizzard considers it a problem, but low priority. (Reference)
Frost PvE
  • Ineffective for progression content. A significant increase (about 15%) is required to bring the spec within 9-10% of Fire’s DPS on most encounters; this is comparable to the difference between Demonology and Destruction. Note also that Frost loses more DPS to movement and high mana demands than Fire and Arcane; if it were 5% behind Fire on a 5-minute completely stationary fight, it would be 10% behind on a 7.5-minute fight with 10% movement.
    Blizzard considers it a problem, but low priority. (Reference)

  • DPS process too simple. Frostbolt is usually the only damage spell cast, except on rare occasions when Brain Freeze procs are needed for extra mana. Unlike permanent pets, the Water Elemental requires minimal, if any, management, usually usable in a simple fire-and-forget manner.
    Blizzard evidently considers this a problem, since they want to add Ice Lance to Frost’s rotation in the process of buffing its performance (Reference). Mathematically, however, this seems a highly improbable solution, since very large changes to Ice Lance result in negligible changes to overall performance (Reference).
Frost PvP
  • Poor damage. Frost Mages in PvP, even those in top competition, now feel that Frost’s damage output is far too low to make it a reasonably successful spec; Frostbolt damage, even when the Mage is left alone to chain-cast, is not considered a serious threat to any team with a healer.

  • Poor itemization. Frost Mages gain very little from crit, due to a low crit multiplier and a high crit rate. Their best PvP stat is haste. There is very little haste available on current PvP gear; while it is possible to build a set with healthy quantities of hit, spell penetration, and crit, it is not possible to build a set with healthy quantities of hit, spell penetration, and haste. PvP Frost Mages would like to see PvP set pieces with haste replacing much or all of the crit on the current sets.
Frostfire Spec

Rather than being a roughly “half-and-half” build that has a different playstyle than deep Fire or deep Frost, this is essentially just a variation on deep Fire with a virtually identical playstyle.

PvE Spec Bloat

The PvE specs for all three trees are bloated, in the sense that very few, if any, points remain for optional talents after maximizing DPS. For all three of these specs, the problem is the Arcane tree:

  • Arcane Mages simply must spend an unusually high number of points within their primary tree on basic DPS/efficiency talents.

  • Fire and Frost Mages must spend nearly all of their “extra” points on the first three tiers of Arcane to reach Torment the Weak, and the only talent available there that makes for an interesting optional talent is Magic Absorption.

The optional or “flavor” talents tend to be the ones that give a spec its personality; once you’ve fulfilled the basic requirements for dealing damage, these are the talents that allow you to do the extra little things other specs can’t do.

Unfortunately, the current design results in nearly all of these points being locked into the secondary tree, which prevents the Mage from having any “flavor” appropriate to his spec apart from basic damage-dealing. A Fire Mage can bring a stronger Amplify Magic, for example, but will have difficulty taking Firestarter. A Frost Mage can have Magic Absorption, but is highly unlikely to have anything but a minimal snare on his Blizzard. An Arcane Mage is more likely to have Frostbite than he is to have Magic Attunement.

See more details in the talents sections.

Wildly Different Scaling Values

The amount of DPS each spec gains from each stat is too different to sustain any kind of reasonably comparable scaling as the Mage gears up.

The numbers here are generated in Simcraft. We recognize that Simcraft and all simulators have limitations when applied to real-world situations. We believe, however, that this will give a reasonable, rough assessment of relative scaling factors.

The following are the scaling values for the four major specs, as used in the standard Simcraft config files. We are including here only the four major DPS stats, and excluding Intellect and Spirit, partly because the contribution of primary stats is relatively low, and partly because the simulator can produce some severely skewed results for primary stats at certain fight durations.

The percentage values listed next to each number is that number’s value compared to the base reference line (average on the first table, Fire/TTW on the second). For example, if crit were worth 1.5 to Fire and 0.5 to Frost, it would have a percentage value of 33.33%. Ideally, for all specs to have similar scaling, all these values should be reasonably close to 100%.

Spec Spell Power Hit Rating Crit Rating Haste Rating
Average 1.5925 2.8650 1.1450 1.5250
Fire/TTW 1.72 (108.01%) 3.14 (109.60%) 1.51 (131.88%) 1.65 (108.20%)
Frostfire 1.69 (106.12%) 2.84 (99.13%) 1.63 (142.36%) 1.54 (100.98%)
Arcane/IV 1.49 (93.56%) 3.11 (108.55%) 0.81 (70.74%) 1.59 (104.26%)
Frost/FFB 1.47 (92.31%) 2.37 (82.72%) 0.63 (55.02%) 1.32 (86.56%)

Here is a second comparison, using Fire as the baseline (since it seems to be the spec that is scaling properly):

Spec Spell Power Hit Rating Crit Rating Haste Rating
Fire/TTW 1.72 3.14 1.51 1.65
Frostfire 1.69 (98.25%) 2.84 (90.45%) 1.63 (107.95%) 1.54 (93.33%)
Arcane/IV 1.49 (86.63%) 3.11 (99.04%) 0.81 (53.64%) 1.59 (96.36%)
Frost/FFB 1.47 (85.47%) 2.37 (75.48%) 0.63 (41.72%) 1.32 (80.00%)

One particularly obvious problem is that Blizzard has given Fire and Frostfire both the highest crit multipliers (254.45% and 330.75% respectively, with CSD) and a significant crit-based DPS proc (Hot Streak). As Arcane and Frost have lower crit multipliers and no crit-based DPS procs, the disparity in value of crit rating is particularly large.

Blizzard has acknowledged that this is a problem, and has planned a pass over the talent trees to improve the situation. (Reference)

Mana Management

Non-Arcane Mages don’t really have mana management. For Fire, Frost, and Frostfire specs, max efficiency is achieved with max DPS; you will virtually always produce the highest total damage done in a fight by using your maximum burn process, using all your mana options up as quickly as cooldowns will allow, and then standing around waiting for regen if you go OOM before the fight ends.

It has been suggested that we should “pace ourselves” in longer/more expensive fights. This could make sense if we did higher DPS than other classes in max burn mode and lower DPS in “pacing” mode, but instead we seem to be balanced around our maximum DPS; this means we underperform when pacing ourselves, without compensatory overperformance when in max DPS mode.

Furthermore, the only method we have for “pacing” is to stop casting, which is not a particularly compelling playstyle.

Blizzard has a significant disagreement with players on this subject. (Reference)

Required Subspeccing

Mage design philosophy is a bit self-contradictory. On one hand, we are not intended to spec very deeply into a single tree; Blizzard wants to require us to spec into a secondary tree. But on the other hand, each spec is intended to focus very heavily on 1-3 spells, all within the same school, and each tree is heavily devoted toward improving that one specific school.

This results in problems such as the one seen by Fire and Frost Mages: they are required to subspec into Arcane to get Torment the Weak and Focus Magic, but since the majority of the talents in Arcane are dedicated to improving a deep Arcane spec, they have difficulty finding useful points to spend on the first three tiers. Arcane Mages have similar difficulty in the first two tiers of Frost.

This results in a significant number of “wasted” points in each spec — points that neither improve basic DPS/efficiency, nor provide optional functionality appropriate to the spec (or, in some cases, any significant benefit whatsoever, such as Tier 1 Arcane for Fire and Frost or Ice Shards for Arcane).

Blizzard considers required subspeccing to be a good thing, but despite repeated direct questions, has not addressed the issue of the number of wasted points. (Reference)

Arcane Subspeccing

The Arcane raid spec subspecs into Frost, with the goal being Icy Veins on tier 3; we will examine the value of talents on tiers 1 and 2.

Tier 1: Three useful points, two wasted points.

  • Frostbite (not useful): Arcane Mages rarely cast Frost spells, and uncontrolled root procs are considered dangerous in a raid situation.

  • Improved Frostbolt (not useful): Arcane Mages virtually never cast Frostbolt in PvE.

  • Ice Floes (useful): The improved Icy Veins and Ice Block cooldowns are occasionally useful in PvE, and the improved Frost Nova cooldown can be a useful quality of life improvement outside of a raid environment.

Tier 2: Three useful points, two wasted points.

  • Ice Shards (not useful): Arcane Mages virtually never cast spells that are improved by this talent.

  • Frost Warding (not useful): Ice Armor and the ward spells are used very rarely in PvE.

  • Precision (useful): The hit chance increase and mana cost reduction are both useful to Arcane Mages.

  • Permafrost (not useful): Arcane Mages very rarely have a chill effect on a target.
Fire Subspeccing

The Fire raid spec subspecs into Arcane, with the goal being Torment the Weak on tier 4. We will examine the value of talents on tiers 1 through 3.

Tier 1: Zero useful points.

  • Arcane Subtlety (not useful): Fire Mages do not cast Arcane damage spells, and dispel protection is not particularly useful in PvE.

  • Arcane Focus (not useful): The only spells this affects that are ever useful to Fire Mages in PvE are Polymorph, Spellsteal and Counterspell. However, since these spells have the same chance to hit as his damage spells, they will be capped through Hit Rating.

  • Arcane Stability (not useful): Again, this talent affects only spells that Fire Mages don’t cast.

Tier 2: Seven useful points.

  • Arcane Fortitude (not useful): Mages do not particularly want increased Armor in PvE.

  • Magic Absorption (useful): The increased resistance is situational, but very useful when the situation arises. This talent can be useful filler for climbing Tier 3.

  • Arcane Concentration (useful): This talent is very valuable to Fire Mages.

Tier 3: Four useful points.

  • Magic Attunement (not useful): Fire Mages don’t value a range increase on their Arcane spells. The value of the improvement to the Amplify/Dampen buffs is very small, as they are rarely used; additionally, such an improvement doesn’t really fit the flavor of a Fire Mage.

  • Spell Impact (useful): This is a strong talent for Fire.

  • Student of the Mind (not useful): Each point provides less than 0.4% DPS and less than 20 MP5 in T8 gear, a pitiful return on investment.

  • Focus Magic (useful): This is also a strong talent for Fire.
Frost Subspeccing

The Frost raid spec subspecs into Arcane, with the goal being Torment the Weak on tier 4. We will examine the value of talents on tiers 1 through 3.

Tier 1: Zero useful points.

  • Arcane Subtlety (not useful): Frost Mages do not cast Arcane damage spells, and dispel protection is not particularly useful in PvE.

  • Arcane Focus (not useful): The only spells this affects that are ever useful to Frost Mages in PvE are Polymorph, Spellsteal and Counterspell. However, since these spells have the same chance to hit as his damage spells, they will be capped through Hit Rating.

  • Arcane Stability (not useful): Again, this talent affects only spells that Frost Mages don’t cast.

Tier 2: Seven useful points.

  • Arcane Fortitude (not useful): Mages do not particularly want increased Armor in PvE.

  • Magic Absorption (useful): The increased resistance is situational, but very useful when the situation arises. This talent can be used as useful filler for climbing Tier 3 if desired.

  • Arcane Concentration (useful): This talent is very valuable to Frost Mages.

Tier 3: One useful point.

  • Magic Attunement (not useful): Frost Mages don’t value a range increase on their Arcane spells. The value of the improvement to the Amplify/Dampen buffs is very small, as they are rarely used; additionally, such an improvement doesn’t really fit the flavor of a Frost Mage.

  • Spell Impact (not useful): This talent is worth less than 0.2% DPS per point, and only if you have already taken Brain Freeze.

  • Student of the Mind (not useful): This talent is worth less than 0.2% DPS per point. As Frost Mages have no Meditation-like talent, it provides zero mana regeneration.

  • Focus Magic (useful): This is a strong talent for Frost.
Arcane Talents
Talent Count
General
  • Maximizing raid DPS in an Arcane spec leaves only about 1 point for use in optional talents.
    Blizzard agrees that Arcane is bloated. (Reference)
Specific
  • Spell Impact (Bug): this appears to be one of only two “increases damage by X%” effects that works additively rather than multiplicatively. For example, combining Spell Impact with Fire Power results in +16% damage; by contrast, combining Piercing Ice with Fire Power results in the expected +16.6%.

  • Torment the Weak (Func): 3/4 raid specs rely upon this talent for a massive 12% of their DPS. However, in any fight with significant target-switching, or in which a melee class cannot consistently stay on the target, the required debuffs cannot be maintained.

  • Arcane Barrage (QOL): Many Arcane Mages would like to be able to use this in standard PvE rotation, rather than only while moving.
Fire Talents
Talent Count
General
  • Maximizing raid DPS in Fire spec leaves only about 2 points for use in optional talents.
    Blizzard has not commented on Fire spec bloat.
Specific
  • Ignite (Bug): Damage can be lost or gained improperly when crits happen nearly simultaneously.
    Blizzard is working on this problem, but has said that it is difficult to solve. (Reference)

  • Pyroblast (QOL): Not particularly useful unless also specced for Presence of Mind, Fiery Payback, or Hot Streak.

  • Improved Scorch (QOL): Many Fire and Frostfire Mages complain about having to spend 7.5 seconds (minus haste) stacking the debuff with a low-DPS spell, particularly in fights with much target-switching. Additionally, buff priority/overwriting issues with Improved Shadow Bolt and Winter’s Chill cause some frustration when two or more of the relevant specs are present in the raid. They also feel it is unfair to have to spend a major glyph slot to improve this situation, given the relative ease and speed with which Improved Shadow Bolt can provide the same effect.

  • Playing with Fire (ROI): 1% DPS dealt per point, minus 1% increased damage taken per point, equals substandard value per point.
    Blizzard feels the vulnerability is useful in providing distinctive flavor to the Fire Mage, but has not commented on the value per talent point. (Reference)

  • Blazing Speed (QOL): Many Fire Mages believe that this talent has little or no value, since melee classes can frequently nullify it by (often automatically) reapplying snares before the Mage can exit melee range.

  • Combustion (ROI): Because this talent’s value diminishes as natural crit rate increases, and because its charges are consumed by low-damage glyphed Living Bomb ticks, the DPS value of this talent is rapidly approaching nil.

  • Living Bomb (QOL): This hybrid single-target/AOE spell is somewhat bizarrely weak for AOE situations due to its inability to detonate when the target dies.
    3.2 changes have greatly improved the value of Living Bomb in AOE situations.
Frost Talents
Talent Count
General
  • Maximizing raid DPS in a Frost spec leaves only 1 point for use in optional talents, which will usually be used to take Ice Barrier.
    Blizzard has not commented on Frost spec bloat.
Specific
  • Improved Blizzard (ROI): Considered by many to be of poor value, especially as 3 points now get you the same benefit as 2 points used to.

  • Improved Cone of Cold (ROI): Considered by most Mages to be of poor value due to infrequent use, high cost, and low base damage of Cone of Cold.
    Popularity: About 3.4% of Frost Mages, excluding 0/0/71 builds

  • Frozen Core (ROI): This talent is inferior, point for point, to almost all similar talents; it is not part of any common PvE or PvP build.
    Popularity: About 1.1% of Frost Mages, excluding 0/0/71 builds

  • Winter’s Chill (QOL): Buff priority/overwriting issues with Improved Shadow Bolt and Improved Scorch cause some frustration when one or more of the relevant specs are present in the raid.

  • Arctic Winds (QOL): PvP Mages can’t really fit it into a typical build. PvE Mages feel that half the value of this talent is wasted for their purposes.

  • Fingers of Frost (Bug): It is often possible to cast a third, instant spell that benefits improperly from the buff.

  • Brain Freeze (ROI): This talent provides less than 0.15% DPS per point on stationary fights of less than around 10 minutes, which is the great majority of encounters. It provides less than 0.67% DPS per point on mobile fights of similar length.

  • Brain Freeze (QOL): Frankly, it just feels plain weird to many of us to be lobbing Fireballs as a Frost Mage.

  • Chilled to the Bone (QOL): Similar to Arctic Winds, PvE Mages feel that a significant portion of this talent’s value is wasted for their purposes.

  • Deep Freeze (QOL): PvE Frost Mages dislike the fact that, in a tree so heavily weighted with PvP talents, and with a dearth of PvE talents, their 51-pointer proves to be more of the same (though of course PvP Frost Mages like the talent very much).
Spells
Arcane Spells
  • Amplify/Dampen Magic (QOL): A group version of these buffs is sorely desired by raiding Mages.

  • Amplify/Dampen Magic (Tip): Buff tooltip inaccurate when talented.

  • Blink (Func): There are places where it simply doesn’t work properly due to pathing issues.
    Blizzard has requested that we file bug reports describing the problem locations. (Reference)

  • Evocation (Func): This spell’s usability is very poor in certain types of encounters, which is problematic as it is absolutely essential to performance for every spec.
    There have been mixed messages from Blizzard about Evocation, with occasional acknowledgment that it may require a change to make it more reliable, but most recently it seems they are OK with Evocation as-is. (Reference)

  • Mage Armor (Tip): Buff tooltip inaccurate when talented/glyphed.

  • Mana Shield (QOL): Many Mages feel that the total mana cost of this spell is disproportionate to the value of the health it preserves.

  • Mirror Image (Func): Part of the problem here is we aren’t entirely sure what the spell is intended to do. If it’s intended to provide extra DPS, it seems be be fairly feeble. If it’s intended to be a threat-management tool, it seems like it should not restore all threat when it expires. If it’s meant to be a confusion tool, it seems like it should not be so easy to distinguish the images from the Mage. If it’s intended to do several things, but not very well, then it is a success.

  • Spellsteal (Func): It’s not particularly effective at stealing useful buffs due to the number of buffs a character may have; as a pure dispel, it is very expensive.
    Blizzard is considering improvements to Spellsteal. (Reference)
Fire Spells
  • Flamestrike (Func): This spell’s tiny radius, without options for improvement, is commonly felt to make it overly difficult to use on all but the most perfectly-tanked, tightly-packed groups of targets.

  • Molten Armor (Tip): Buff tooltip inaccurate when talented/glyphed.

  • Scorch (QoL): Some Fire Mages feel that the shorter range on this spell is problematic in PvE, since maintaining Scorch nullifies the range advantage provided by their other spells.
Frost Spells
  • Frost/Ice Armor (Tip): Buff tooltip inaccurate when talented/glyphed.

  • Frostbolt (QOL): Many Mages feel that Frostbolt should no longer suffer the 5% penalty to its coefficient due to its snare. The argument in support of this change centers on Frostfire Bolt, which suffers no such penalty while having the same snare, and on the fact that Frostbolt simply isn’t scaling well.
Glyphs
Arcane Spells
  • Glyph of Mage Armor (WTF): this complaint is a little controversial. Many Arcane Mages complain that the glyph is useless because regen is capped at 100%, and 100% is achieved through talents and unglyphed Mage Armor. Some, however, argue that the glyph’s intent is to allow the Mage to drop a point or two from Arcane Meditation.
Fire Spells
  • Glyph of Frostfire (Bug): this appears to be one of only two “increases damage by X%” effects that works additively rather than multiplicatively. For example, with Fire Power, the combined effect is +12% rather than the expected +12.2%.

  • Glyph of Living Bomb (Func): Fire Mages feel that this glyph should not interact with Combustion charges, as it can very quickly consume them on low-damage ticks.
Frost Spells
  • Glyph of Frostbolt (QOL): Frost Mages loathe this glyph. It is the only primary nuke glyph that you do not want to use unless you are fighting a raid boss. No other spec wants to remove the glyph from its primary nuke when it goes out to solo or quest. And quite a few other glyphs offer a similar DPS increase while carrying a negligible downside (e.g., Fireball) or no downside at all (e.g., Frostfire Bolt, Incinerate, Arcane Blast). To add insult to injury, the glyph also devalues points spent in Chilled to the Bone and Permafrost.
Set Bonuses
Tier 8
  • Four-piece (WTF)

    • This proc is significantly less valuable to Arcane than to Fire and Frostfire; when Arcane uses a Missile Barrage proc, it consumes the Arcane Blast debuff, greatly reducing the value of an additional cast.
    • The proc has miniscule value to a Frost Mage; as of the 3.2 Empowered Frostbolt change, Brain Freeze is worth at best about 0.67% DPS per talent point on a mobile fight, making the proc worth only a fraction of that.