|
Having established an upper bound of the MP5 equivalence at 200 MP5, (Every heal an overheal, zero lag, spam casting) it may be useful to establish a lower bound, or move towards a formula for estimating its value.
Well, the lower bound is zero-- no value if you're not casting. Let's set up some conservative median values:
I usually overheal for 30-40% on my GHeals over the course of the night. (Adjust this number to fit your healing style.)
Now, the best case for the set bonus is that every heal I land is an overheal by ~30-40%. The worst case, though, is that 70% of my heals have no overheal and 30% are fully overheal. That means that only 30% of casts receive mana back.
This calculation is very sensitive to casting frequency. The assumption I usually use for theorycrafting is a casting period of 3s-- this is fairly rapid casting, but not spam, allowing for some regen breaks or just timing. (I'll check this assumption against some logs later tonight.)
For a 3s casting period, you get mana back on 100% (best case) or 30% (worst case) of the casts.
100% is (100 mana / 3s = 33.3 mana/s =) 166.6 MP5
30% is (30 mana / 3s = 10 mana/s =) 50 MP5
That gives you a range of values.
Now, to get a general estimate, let's try some statistical handwaving:
Treating individual casts as independent random events, it is likely that the distribution of overheal per cast will be normal, with a mean equal to your overall overheal percentage.
That is, if your overheal percentage on the night is 40%, then your average heal will be one that heals for 60% of its value with 40% overheal, and you will have roughly as many casts over this as under it, with most samples near it. (This is complicated by the fact that overheal is capped between 0 and 100%, but I think we can set up the assumptions for that not to matter.)
Now, the question becomes, given a certain overall overhealing percentage, what percentage of casts will contain no overheal?
Getting the actual answer would require figuring out what the variance is within a given sample of heals, which is complicated even given my huge assumptions about the underlying distribution.
But, we can use that assumption to make a guess-- namely, that the mean will generally lie halfway between the best and worst case scenario, which is a feature of normal distributions. For a 30% overhealer, an average of 65% of heals will contain some overhealing. For a 40% overhealer, 70%.
Given 3 seconds between GHeals, that's (65 mana / 3s = 21.6 mana/s =) 108 MP5 for a 30% overhealer, and (70 mana / 3s = 23.3 mana/s =) 116.5 MP5 for a 40% overhealer.
If you asked me to put an MP5 equivalence to the set bonus with a gun to my head and no access to real data sets, that's the model I'd use. Anyone with (a) access to good data, or (b) better knowledge of stats want to comment?
|