Gearing a ranger

[bcbrown]

Snaggles kindly offered to send me a couple parses with Swarmcaller and the two epics against a froglok forager. Here's the results of a quick analysis.

First, I've been doing some parses on my own ranger, and found some interesting patterns. Here's a damage histogram from a Fellspine's Tail in mainhand vs Bloodmaw, 574 hits:

[You must be logged in to view images. Log in or Register.]

The most obvious observation from this graph is the huge spike at 22 damage, which exactly equals 2*damage + bonus, or 2*7+8. There's a min value of 9 (or 1+8), and a max value of 34. The other thing I noticed was that there's more hits above this modal value (46%) than below (15%), with 39% at the modal value. I'm not going to go into the rest of my analysis here, as that's far too much for a single post, but I wanted to use this to illustrate my expectations when I started to look into Snaggles' parse.

I expect to see a similar graph for the Swarmcaller, with a modal value of 2*29 + 34 or 92. For Earthcaller + Swiftwind, since both weapons are slashing, I won't be able to pull them into two separate graphs, so instead of a single modal value I expect two, at 2*14 + 11 = 39 for EC and 2*13 = 26 for SW.

If you normalize those modal values by delay you get:
Swarmcaller: 92/41 = 2.24
Earthcaller: 39/24 = 1.625
Swiftwind: 26/21 = 1.24
Earthcaller + .75 * Swiftwind = 2.55

So the epics should be roughly 2.55/2.24 = 13% more DPS than the Swarmcaller.

On to the data

Swarmcaller: Three fights, with durations of 1:44, 1:42, 1:44 (from timestamp of first swing to last swing). Total 168 hits, 0 misses, 168 swings. Total damage is 17720, average hit is 105, DPS is 57.16.

[You must be logged in to view images. Log in or Register.]

Earthcaller + Swarmcaller: Two fights, with durations of 1:18 and 1:24. Total 255 hits, 0 misses, 255 swings. Total damage is 11494, average hit is 45, DPS is 70.95.

[You must be logged in to view images. Log in or Register.]

Discussion

Both graphs are more or less as expected, but there's some interesting wrinkles. The first is the secondary peak at the max hit value; I haven't seen that on any of my parses. The Epic DPS is higher than Swarmcaller as expected, but it's 24% higher rather than the 13% higher I expected.

I'm very surprised that there seems to be zero misses in the logs. I had a 90% hit rate (at level 51) against a level 5-8ish mob, but Snaggles had a 100% hit rate (at level 60) against a level 45 mob. His strength is something like 230 while mine was at 138.

Given the delays and a 75% dual wield success rate, I'd expect the Earthcaller to be about 54% of the total epic swings and the Swiftwind to be about 46% (1/24 vs 1/21 * .75), but the modal peaks imply the Swiftwind has about twice has many hits as the Earthcaller (7% at 26 damage vs 3.9% at 39 damage).

Finally, I'll note that there were 168 and 255 hits for Swarmcaller and Epics, while my Fellspine parse had 574 hits. Given the results I've found so far I think you need about 500 hits to avoid the data being excessively noisy. So I'm not too surprised that not all the projections line up exactly. The secondary peak at exactly the max value is very interesting, though.

In the ranger AC thread I laid out a theory for defense calculations: you'll have peaks at min and max hits and how much AC you have determines the relative proportion of min and max hits. We already knew there was some player-specific calculation that uses the "damage table" values to smear out the damage values, leading to values above what would be the max value for an NPC. The parses I ran suggested that's likely accurate, where the underlying distribution is past the squelch point such that all the min/max values are at the max value, and the "smearing" can lead to up to 30% of all hits being past the modal value, up to twice as high. Snaggles' parses suggest a similar conclusion, except with enough "something" (strength, +attack, skill levels, the "extra offensive bonus" rangers start getting at 55) the "smearing" gets capped at a ceiling leading to a secondary peak at the observed max damage.