snaggle
Dalayan Beginner
I've spent many hours in the ol' gorge of the one-eye testing variables like atk, critical hit %, weapon skill mods, and rbow on my dps. I'm kind of a nerd. I wanted to see what effect each had exactly, and which to prioritize when I can't max it all.
The setup:
You can see my fomelo here:
http://shardsofdalaya.com/fomelo/fomelo.php?char=minuette
I'd say I'm about tier 7, with some higher and lower loots. I wanted a consistent testing environment, so I went with the cyclops elders in the gorge of one-eye. They are all pretty similar in level, there's plenty of them, and I can kill them without having to fight anything else so the log file is easier to parse. I'm using Gamparse, which has cool overlay features so I can watch dps in real time. I killed at least 8 for each test, without the pet or casting. I used a shammy to slow it so I wouldn't lose cycles casting.
I had pretty much the usual raid buffs, except I was using swift like the wind for haste. Focus, rbow, sv, eot/eotw, bihli, cotb, rageo, jb, sltw. I didn't cast spells or kick to keep things simple. I used normal stance.
Variability of the data
The first thing people should know about parsing dps is how wildly the numbers will vary. There are a ton of factors that effect each combat round, and most things are a modification of a random 'roll' like in d&d, so the occasional crit on a proc can really throw off the numbers. The key is the law of large numbers - the more rounds of combat involved in testing, the closer to the true mean the numbers will be. Each test consists of me doing at least 100k in melee damage.
As an example, here is a list of the dps numbers for one of the configurations I tested:
110.3
79.1
79
86.9
74.5
79.8
89.9
89.4
88.6
71.5
79.2
72.7
88.3
Average: 83.78
Standard deviation: 9.87
Basically, if the average dps for scenario 1 is within one or two of another scenario, chances are it's not statistically significant.
The data
http://www.fourthwc.com/files/10.25.10dps.jpg
(It's a little wide to post in the thread)
I ran 12 different setups, but I'll start with 4 for now. If anyone besides me cares, I can post the rest. I can also post the actual logs if anyone wants to out-nerd me. This set was to see what effect +hand to hand mod items have. I used a h2h weapon in the offhand, along with a 1hb in main so I could parse each type of damage separately. I couldn't remember exactly what +weapon skills are supposed to do, nor could I find it on the forums/wiki. I think my data shows that they do very little.
In configs #1 and 4, I used my new 100k charm, which has 25 atk and gets +5 h2h from my seeker rune. In configs #2 & 3, I swapped in the charm of the brute, which gives +15, so with the rune brings it to +20 h2h. I expected to see a significant difference in some aspect of h2h performance when I made this switch, but really didn't see much of anything.
Rbow
I've seen some weird stuff with Rbow. It's supposed to simply add a proc to melee attacks, but seems to do more. The parser views procs as 'dirdmg', so the procs aren't going to show up directly in my data. If you exclude the procs, there should be no difference between using Rbow or not. In configs #1 &2, I had Rbow, after that it faded. Configs #1 & 4 are the same, except for Rbow, and it seems like #1 was significantly higher in dps. It's hard to tell, given the high standard deviation.
+Hand2Hand mods
These appear to be worthless. Configs #1, 4 & 3 are pretty close in terms of atk, so the only variable left should be the extra 15 h2h. #3 does worse in comparison to both in terms of h2h dps, with no change in accuracy and slightly worse in % crit damage. Adding 15 more h2h doesn't raise the minimum normal hit, and might slightly raise the max crit hit.
Conclusions
* +h2h skill is not worth pursuing, and maybe all weapon mods are equally insignificant.
* Worn atk is fungible - it doesn't matter if it comes from the charm or other items. (my other parses deal with spell atk like savagery)
* adding 40 worn atk doesn't make a huge difference in dps, as shown by comparing #2 & 3. 1hb went up and h2h went down, so either the effect is lost in the noise or h2h isn't working properly.
* Focus effects like affliction enhancement and damage increment shouldn't be swapped out for atk gear for bsts.
* One should never accept conventional wisdom, and should try parsing a bunch of stuff to optimize their character's dps.
Future parsing tests (or suggest your own):
* Effect of /s stances
* Effect of damage increment on overall dps - procs only and procs + nukes
* Effect of affliction enhancement on overall dps
* front/behind comparison
The setup:
You can see my fomelo here:
http://shardsofdalaya.com/fomelo/fomelo.php?char=minuette
I'd say I'm about tier 7, with some higher and lower loots. I wanted a consistent testing environment, so I went with the cyclops elders in the gorge of one-eye. They are all pretty similar in level, there's plenty of them, and I can kill them without having to fight anything else so the log file is easier to parse. I'm using Gamparse, which has cool overlay features so I can watch dps in real time. I killed at least 8 for each test, without the pet or casting. I used a shammy to slow it so I wouldn't lose cycles casting.
I had pretty much the usual raid buffs, except I was using swift like the wind for haste. Focus, rbow, sv, eot/eotw, bihli, cotb, rageo, jb, sltw. I didn't cast spells or kick to keep things simple. I used normal stance.
Variability of the data
The first thing people should know about parsing dps is how wildly the numbers will vary. There are a ton of factors that effect each combat round, and most things are a modification of a random 'roll' like in d&d, so the occasional crit on a proc can really throw off the numbers. The key is the law of large numbers - the more rounds of combat involved in testing, the closer to the true mean the numbers will be. Each test consists of me doing at least 100k in melee damage.
As an example, here is a list of the dps numbers for one of the configurations I tested:
110.3
79.1
79
86.9
74.5
79.8
89.9
89.4
88.6
71.5
79.2
72.7
88.3
Average: 83.78
Standard deviation: 9.87
Basically, if the average dps for scenario 1 is within one or two of another scenario, chances are it's not statistically significant.
The data
http://www.fourthwc.com/files/10.25.10dps.jpg
(It's a little wide to post in the thread)
I ran 12 different setups, but I'll start with 4 for now. If anyone besides me cares, I can post the rest. I can also post the actual logs if anyone wants to out-nerd me. This set was to see what effect +hand to hand mod items have. I used a h2h weapon in the offhand, along with a 1hb in main so I could parse each type of damage separately. I couldn't remember exactly what +weapon skills are supposed to do, nor could I find it on the forums/wiki. I think my data shows that they do very little.
In configs #1 and 4, I used my new 100k charm, which has 25 atk and gets +5 h2h from my seeker rune. In configs #2 & 3, I swapped in the charm of the brute, which gives +15, so with the rune brings it to +20 h2h. I expected to see a significant difference in some aspect of h2h performance when I made this switch, but really didn't see much of anything.
Rbow
I've seen some weird stuff with Rbow. It's supposed to simply add a proc to melee attacks, but seems to do more. The parser views procs as 'dirdmg', so the procs aren't going to show up directly in my data. If you exclude the procs, there should be no difference between using Rbow or not. In configs #1 &2, I had Rbow, after that it faded. Configs #1 & 4 are the same, except for Rbow, and it seems like #1 was significantly higher in dps. It's hard to tell, given the high standard deviation.
+Hand2Hand mods
These appear to be worthless. Configs #1, 4 & 3 are pretty close in terms of atk, so the only variable left should be the extra 15 h2h. #3 does worse in comparison to both in terms of h2h dps, with no change in accuracy and slightly worse in % crit damage. Adding 15 more h2h doesn't raise the minimum normal hit, and might slightly raise the max crit hit.
Conclusions
* +h2h skill is not worth pursuing, and maybe all weapon mods are equally insignificant.
* Worn atk is fungible - it doesn't matter if it comes from the charm or other items. (my other parses deal with spell atk like savagery)
* adding 40 worn atk doesn't make a huge difference in dps, as shown by comparing #2 & 3. 1hb went up and h2h went down, so either the effect is lost in the noise or h2h isn't working properly.
* Focus effects like affliction enhancement and damage increment shouldn't be swapped out for atk gear for bsts.
* One should never accept conventional wisdom, and should try parsing a bunch of stuff to optimize their character's dps.
Future parsing tests (or suggest your own):
* Effect of /s stances
* Effect of damage increment on overall dps - procs only and procs + nukes
* Effect of affliction enhancement on overall dps
* front/behind comparison
