Friday, September 17, 2010

Scaling Discrepancies

I finally got onto the 4.0 PTR last night, so at long last I have the spell coefficients for heals. Of course I could get only an approximation for spells that have actual ranges, but from the approximations it is pretty easy to get the exact value. Power Word: Shield was 6/14 and Renew was 9/14. For all spells that had ranges, the average heal over 10 casts was within 0.05 of a whole number of 14ths (with the exception of Prayer of Mending), so we can assume that the whole number of 14ths is the precise value.

Here we go:
Flash Heal9/14
Greater Heal12/14
Power Word: Shield6/14
Prayer of Mending31.4%

I think prayer of mending is basically a 9/14 and suffering from a 50% penalty because it's an area of effect. Also, currently the coefficient on Power Word: Shield is actually zero, but if you extrapolate what it was meant to be from the Glyph you get the coefficient I gave above. Presumably this bug will be fixed before it goes live.

It also turns out my spell power estimates from earlier were much too low due to two factors. First, priests get a passive ability at level 50 that gives them 5% more intellect. Second, discipline priests get 15% more intellect as a passive bonus. This seems pretty weak compared to the holy priest 25% more healing, but we'll have to see how that shakes out in the end.

These numbers make for some pretty sensible choices between spells:

  • Heal has over double the mana efficiency of the next non-Penance competitor with less than half the throughput.
  • Penance has better mana efficiency by half than the best non-heal spammable option and is close but a little less than the throughput of greater heal, making it pretty much your automatic first button if Heal/Shield is not enough, so you'll cast Flash Heal and Greater Heal only when Penance is on cooldown.
  • Flash Heal is easily the strongest throughput, beating Greater Heal by around 25%, but by far the worst efficiency, around 5/6 of that of Greater Heal.
  • A Prayer of Mending that procs only once is the worst spell you can cast. If it procs twice then it's getting pretty close to - but it still strictly worse than - Greater Heal for both throughput and efficiency, so it's still a great spell, but only when you need it and only when it is going to work.
  • Renew is around the throughput of a Flash Heal with the mana efficiency of a Greater Heal, so you should use it over those when the damage distribution makes doing so possible.
  • Power Word: Shield is always better in mana efficiency and healing throughput than Greater Heal, so it is easily your best non-Heal heal.

It seems like everything in the single-target toolbox has a place.

What's interesting, though, is the change in the way heals are going to be scaling. Notice that the spell power coefficients of all heals have been drastically reduced. Even Flash Heal, now around 64%, was a little over 80% in Wrath. They've stuck with the casting time divided by 3.5 model for most things, but removed the 1.88 multiplier that was added in when healing power was converted into spell power.

What this means for heals is that we can expect approximately half of the strength of our heals to come from the base spell while half comes from our spell power. As some examples, the base amount of Heal averages 4590 when you factor in multipliers from talents. In tier 1 raid gear the non-crit total heal will be around 9017, just under double the base. In final tier this increases to only 10928, meaning the base spell is still 42% of the value of the heal.

Flash Heals base amount after talents is 6887, the Tier 1 non-crit is is 13887. Again, almost exactly double the base. Shields base counting mastery, talents and glyph is 6496. In Tier 1 a shield will be around 12740.

The fact that half of heals comes from the base really skews the way that heals scale compared to offensive spells. I previously noted that only around 13.5% of a Fireball's damage will come from spell power. That means that as we move from tier 1 raid gear to final tier raid gear, and spell power increases by around 42%, the healing a priest does will increase by around 21%, while the damage a mage does will increase by around 36%.

This isn't a problem on it's face, but it becomes a problem when we consider that healers and damage dealers use the same coefficients to convert ratings into game effects. A mage gets 1% haste from 128 haste rating. That's probably going to be a 0.87% increase in damage. If you instead gained 128 intellect, that would likely increase your spell power by around 2.1%, which would increase your damage by 1.82%. Haste is atrociously bad. For a healer the haste does around the same thing but a 2.1% increase in spell power is only a 1.05% increase in healing. That means that haste is more than 80% as good as a spell power, instead of less than 45% as good. For crit things are a little closer together since dps get substantially more extra damage when they crit than healers get extra healing. Even then, it's going to be very hard to make up for the fact that spell power is just a much more powerful scalar for dps, and crit will likely be a bad stat for nearly everyone anyway with the current conversion values (despite the fact that they've been dropped dramatically since the first time I complained about this).

With these drastic differences in their spell power to percent effect conversion and virtually identical haste and crit to percent effect conversion, these ratings have to be either far too good for one or far too bad for the other. Something in these formulae will have to change for stats to make much sense at all.

No comments:

Post a Comment