Arcane Mage Mana Management: Why 93%?

[Komma]

Many popular Arcane Mage guides have the following line in their rotation description:

Cast Arcane Blast, if you are above 93% Mana before your start casting.

The same logic can be found in SimulationCraft's Arcane Mage APL:

arcane_blast,if=buff.arcane_charge.stack=4&mana.pct>93

This has lead to people asking where the 93% "magical number" came from, which I plan to address in this thread.

The TL;DR answer is "Using ABar instead of AB at >=93% mana leads to wasted mana". Technically speaking, 93% isn't completely on-point, because it's a rounded number. The exact number is actually closer to 92.78% for non Arcane Orb users.

Here is a spreadsheet of how an Arcane Mage's mana fluctuates as you dump Arcane Charges and rebuild. As seen in the first section, the highest mana level (highlighted in green) is observed just before you finish casting the second Arcane Blast (AB1, which stands for Arcane Blast with 1 Arcane Charge).

What this means is that if you cast Arcane Barrage at any point above 92.78% mana, you will reach 100% mana when rebuilding Arcane Charges. This means wasted mana. Another way to look at this is any excess mana above 92.78% when casting Arcane Barrage is wasted, because you're already at the highest level of mana possible after rebuilding Arcane Charges.

Note that due to Nether Attunement, the spreadsheet applies to all Arcane Mages regardless of haste levels. The numbers are determined by player regen %, base mana costs and the effect of Arcane Charges, not by any character statistic such as haste or mastery.

How does Arcane Orb change things? By shortcutting the first 2 Arcane Charges, the player can now stay at a slightly higher mana level before casting Arcane Barrage. But as you see in the spreadsheet, the difference is very small; 93% is still a valid number.

The current metagame for Arcane Mage DPS can be interpreted as "Cast as many Arcane Blasts at 4 Arcane Charges as possible". To do this, you want to be efficient with your regen, stay close to maximum mana, while casting the maximum amount on 4 Arcane Charge spells. This is why we avoid mana capping - wasting regen leads to a lower amount of "excess" that can be spent on AB4s.

In SimulationCraft, "93%" was chosen not only by spreadsheet analysis, but also through trial-and-error. After experimenting with numbers between 90-95%, it was chosen because it offered the best results.

[Pappus]

Wouldn't it make more sense to use AO to get the last 2 charges instead of the first 2?

[Komma]

Wouldn't it make more sense to use AO to get the last 2 charges instead of the first 2?

No, it doesn't. Optimizing DPS by avoiding casting AB0 and AB1 which have very low DPET, is much more important (and rewarding) than optimizing DPM.

[elovia]

What about wasting mana via AM regen? The AM threshold (wherein you will be at 100% mana after casting 1 AM + 1 AB) is around 91%. However there is nothing in the APL that cares about this threshold. Furthermore, as I was typing this post I looked into how SimC estimates mana regen and this looks really odd to me.

http://i.imgur.com/dmSjMjW.jpg" target="_blank

I assume in this example the 95% is what mana% the character is at when it begins the cast, not when the cast is complete? On my character it's usually

91% - Begin AM
95-96% - Complete AM/Begin AB
99-100% - Complete AB

Assuming all of the above is correct, has it been tested for situations where you have 1-2 AM and are at a mana level above AM Threshold (~91) is it better to AM and waste mana for a stronger AM or cast AB and not waste any mana? If not I can run it myself (and probably will anyways for funsies) and report back

Edit - after fiddling in SimC, saving AM until below the threshold is only a ~20-30 dps increase; nothing to write home about.

[Pappus]

Wouldn't it make more sense to use AO to get the last 2 charges instead of the first 2?

No, it doesn't. Optimizing DPS by avoiding casting AB0 and AB1 which have very low DPET, is much more important (and rewarding) than optimizing DPM.

The idea is based on the thought, that you can delay ABar by one AB dipping below 93% mana if your AO is ready to get charge 3 and 4 and therefor avoid the big hits to your mana pool.

The difference is probably small, but I feel there is a good return possible if you are allowed to gamble one more AB every 30s. It might lead to another AM into another AB etc after all.