Frost, Fire, or Arcane

[Akraen]

I said why I used the 60th percentile---snip

You don't realize who you're talking to do you? He's an incredibly intelligent person who cares enough about mages to get on IRC and chat theorycraft and other things.

I'd recommend not making personal insults. When the mods wake up they'll see it. Maybe edit it before they do, omitted your quote intentionally.

[Frost1129]

First off, yeah who would care about 60th percentile DPS? Or any of your statistics? It's all bad data, all of it. It's just the only data you have so you're scrambling to draw conclusions.

I'm not saying that frost is best. It's definitely in 3rd place still.

My argument is on the margins and the apparent difference at the skill level 99% of players are at. Each time someone makes a thread like this I grit my teeth, because it's absolutely pointless.

Until you can sort and filter based off categories of boss strats, uncheck raid-comp exclusive buffs like tricks, stormlash, banner, and normalize all the kill times then I will not accept any of your raidbots data and I don't think anyone else should either.

But you won't, these threads will keep happening-- they'll never be productive, and they'll always result in arguments like these.

Your justifications for it all being bad data are ridiculous. What IS good data to you? Data is never perfect, you work with what you have. It's the best resource we have for this type of analysis beyond "LOL IN MY RAID I'M THE BEST" type of anecdotal evidence that gets us nowhere.

You realize that the entire point of using large data sets like this is to remove noise from the odd (minority) groups which have weird group comps, excessive amounts of tricks/stormlash/banner, and different strategies, correct? That's the entire reason why this type of data is better than anecdotal evidence - that type of noise is overtaken by the actual differences in specs. Literally the entire reason. If we were talking about very small sample sizes, I would agree with you. But we're not. We're talking about thousands and thousands of repeat kills.

I'll more clearly lay out why I chose the 60th percentile, heroic players: (EVEN THOUGH THE AVERAGE AND MEDIAN AND HIGHER PERCENTILES HAVE THE SAME TRENDS)

1) These players are not cutting edge. They are not in groups which are clearing 14/14, but they are probably decent enough players. They are not employing bullshit cheese strats, and the odds of them class stacking stuff like banner/tricks/stormlash are very low. If they were having these types of advantages, they would not be in the 60th percentile. These are your average heroic raiders.

2) These players data will not suffer from asian ilvl inflation (or rather, that is my hope)

3) These players are probably in average guilds, using the strats we get from Fatboss and Icy veins. They are not innovators, they are followers. The odds of them using some oddball strategy is therefore less (and masked anyway by large numbers, but whatever).

4) Again, these are average players. They are not you or kuni or whoever else you want to list that have a deeper understand of how to min/max the individual specs. These are your 99%.

I could look at every percentile barring the absolute worst and find the same exact trend. From the worst to the best and then to the actual average/median. The trend holds. This means at every skill level, there is a large difference between the lowest (frost) and highest (arcane). It is not marginal.

None of this means people shouldn't play Frost though. I like Frost, I prefer it to the other specs at the moment actually. It just means we need to be honest about the way things are and not tell people Frost > *, like that one dude did.

- - - Updated - - -

You don't realize who you're talking to do you? He's an incredibly intelligent person who cares enough about mages to get on IRC and chat theorycraft and other things.

I'd recommend not making personal insults. When the mods wake up they'll see it. Maybe edit it before they do, omitted your quote intentionally.

Who he is doesn't matter much to me. The mods are welcome to see what they will, I don't mind.

[Akraen]

1) That's an assumption
2) They absolutely do suffer from asian ilvl inflation, there are asian parses all throughout the ranks
3) Fair, but I'd also say they include first kills, which rarely require focused min/max DPS. Have you ever been a guild/raid leader who knows he's operating at 80% of his potential because he's herding 24 cats? Yeah that's why I only ever get nice parses when things are on farm, and even then my "average" 11/14h guild has kill times of bad guilds because we play it safe. If you want to see me not playing it safe, look at Heroic Spine from last xpac and you'll see my name. It was totally skill that got me that rank, not threatening gkicks if anyone touched bloods but me.
4) So if they are our 99%, wouldn't it be better for your argument to be play frost since it's easiest? You'd have a much more solid argument saying the problem is that the likes of Kuni and I play frost to begin with, we should be fire, right? Kuni and I aren't idiots, we both know arcane and fire will yield higher numbers, but not high enough numbers to say it should even feel required.


You are 100% correct on the trend that raidbots presents, but the trend is blown a bit out of reasonable proportion due to the skewing that I insist is there from varied kill times, sample sizes, WCL, and one major thing you forgot: the amount of frost mage parses without water elementals associated, which happens any time a guild begins logging after the mage summoned his elemental.

My career is analysis, and it all hinges on knowing significant data from erroneous data. The last thing I'm going to do is sit here and tell you that the trend is wrong, or that frost is the best, or that all those problems with raidbots even change the outcome. But what I will tell you with absolute certainty is that unless you normalize fight duration, fix the water elemental bug, and exclude asia-- attempting to cite raidbots in any argument is misleading.

Had you just quoted and corrected that outlandishly wrong person I would never have chimed in.

[DarkstarG]

Who cares about 60th percentile dps rofl

Because that is where most players will be on average.

[Deleted]

Data can be quite useless if you make the wrong generalisations. Why do you assume that if you're, say, 6/14 HC with 568 item level you are automatically a "decent" player? Because they're not 14/14? Your data is extremely polluted as well, there is no way you can exclude Asians by simply selecting the 60th percentile. I really don't understand why the 60th percentile matters.

What people need to do is look at their gear, think carefully, and then change a spec accordingly. Hell, you might even want to try them out. If somebody's found their way here to ask which spec they need to play then they should be capable of doing that at least.

[Huevos]

In response to the "who cares about the 60th percentile", there is a pretty good chance the OP cares about the 60th percentile, since he's at 561 and is admittedly new.

If you want to refute a post that draws some conclusion from mathematical evidence, back it up with your own data so the community can make an informed decision. There is no perfect set of data, we work with what we have, but if the trend is the same at various percentiles, I don't see how you can say there isn't any relevance unless you back up your claim with your own data.

It's easy to just tell people to play what they want because for the most part it's the right answer. But a lot of people, especially those like the OP who don't have the experience in all three specs, aren't going to be able to decide what they play the best without a lot of reps. If it were me, I'd want to be pointed in the right direction.

[Akraen]

Because that is where most players will be on average.

No it isn't, that's where the 60th percentile players are.

You take x players, line them all up in order from highest to lowest.

Then you put 10 equally spaced lines between them.

60th percentile is the same quantity of players as the 50th percentile, or the 70th percentile, or the 100th percentile.

[Deleted]

Basically, the real answer is: Play whatever you're most comfortable with. Except at low gear levels, don't play fire at low gear levels. 565+ (or so) you can basically choose any spec, and with perfect play be within 5-10% of any of the 3 specs.

[Akraen]

In response to the "who cares about the 60th percentile", there is a pretty good chance the OP cares about the 60th percentile, since he's at 561 and is admittedly new.

If you want to refute a post that draws some conclusion from mathematical evidence, back it up with your own data so the community can make an informed decision. There is no perfect set of data, we work with what we have, but if the trend is the same at various percentiles, I don't see how you can say there isn't any relevance unless you back up your claim with your own data.

It's easy to just tell people to play what they want because for the most part it's the right answer. But a lot of people, especially those like the OP who don't have the experience in all three specs, aren't going to be able to decide what they play the best without a lot of reps. If it were me, I'd want to be pointed in the right direction.

I did. I normalized fight times and drew everyone a much better picture here.

I've had this same argument in the same way a hundred times. Just search it.

[Frost1129]

I did. I normalized fight times and drew everyone a much better picture here.

I've had this same argument in the same way a hundred times. Just search it.

My career isn't strictly analysis, but error analysis of my experiments plays a large role in it - if we want to throw that stuff out there. Your link is very old and many changes have happened.

I'll do what you just did on a larger scale with more recent data when I get home and post the results, taking into account all the different factors you require.

[DarkstarG]

No it isn't, that's where the 60th percentile players are.

You take x players, line them all up in order from highest to lowest.

Then you put 10 equally spaced lines between them.

60th percentile is the same quantity of players as the 50th percentile, or the 70th percentile, or the 100th percentile.

What it means is that approximately 60% of the playerbase will be approximately within the information he gave. It has nothing to do with the remaining 40% because they will either be lower or higher depending on gear/skill.

[Frost1129]

1) That's an assumption
2) They absolutely do suffer from asian ilvl inflation, there are asian parses all throughout the ranks
3) Fair, but I'd also say they include first kills, which rarely require focused min/max DPS. Have you ever been a guild/raid leader who knows he's operating at 80% of his potential because he's herding 24 cats? Yeah that's why I only ever get nice parses when things are on farm, and even then my "average" 11/14h guild has kill times of bad guilds because we play it safe. If you want to see me not playing it safe, look at Heroic Spine from last xpac and you'll see my name. It was totally skill that got me that rank, not threatening gkicks if anyone touched bloods but me.
4) So if they are our 99%, wouldn't it be better for your argument to be play frost since it's easiest? You'd have a much more solid argument saying the problem is that the likes of Kuni and I play frost to begin with, we should be fire, right? Kuni and I aren't idiots, we both know arcane and fire will yield higher numbers, but not high enough numbers to say it should even feel required.


You are 100% correct on the trend that raidbots presents, but the trend is blown a bit out of reasonable proportion due to the skewing that I insist is there from varied kill times, sample sizes, WCL, and one major thing you forgot: the amount of frost mage parses without water elementals associated, which happens any time a guild begins logging after the mage summoned his elemental.

My career is analysis, and it all hinges on knowing significant data from erroneous data. The last thing I'm going to do is sit here and tell you that the trend is wrong, or that frost is the best, or that all those problems with raidbots even change the outcome. But what I will tell you with absolute certainty is that unless you normalize fight duration, fix the water elemental bug, and exclude asia-- attempting to cite raidbots in any argument is misleading.

Had you just quoted and corrected that outlandishly wrong person I would never have chimed in.

Part 1 is an assumption. But I think it's a valid one.

I have been a raid leader, actually. So I do understand that PoV.

No, you shouldn't be fire / arcane. I'm not talking about ease of play, or what specs people feel the most comfortable with, or the spec people find the most fun, or what spec should be required to play. Additionally, I don't think that you or Kuni are idiots. I'm saying that ignoring that stuff, the specs have an arcane > fire > frost trend, with good sized gaps in between, which we see in the raidbots data, regardless of how you parse it.

For the record, I prefer you chiming in over someone who doesn't even bother to read my posts but feels the need to respond to them.

[Akraen]

[Deleted]

I concur wholeheartedly. Hopefully this will help people understand how percentiles work.

[Frost1129]

Data can be quite useless if you make the wrong generalisations. Why do you assume that if you're, say, 6/14 HC with 568 item level you are automatically a "decent" player? Because they're not 14/14? Your data is extremely polluted as well, there is no way you can exclude Asians by simply selecting the 60th percentile. I really don't understand why the 60th percentile matters.

What people need to do is look at their gear, think carefully, and then change a spec accordingly. Hell, you might even want to try them out. If somebody's found their way here to ask which spec they need to play then they should be capable of doing that at least.

It was an attempt to word around the asian lvl problem. If you have a better solution let me know and I'll apply it.

[hatchetman240]

There still isn't enough information to exclude raidbots as a data source, not even close to enough. Assuming asian data skews DPS higher, it will skew it higher for everyone pretty much evenly. The only point presented concisely is the statement that water ele is not represented for frost mages, great! That's really useful, but it doesnt account for the 25% DPS gap between arcane and frost in heroics. Every other variable should impact all specs equally (strat differences, ilvl AND player skill). WOL/raidbots is perfectly fine for citation. Demanding people stop citing for personal reasons is very strange.

[Deleted]

There still isn't enough information to exclude raidbots as a data source, not even close to enough. Assuming asian data skews DPS higher, it will skew it higher for everyone pretty much evenly. The only point presented concisely is the statement that water ele is not represented for frost mages, great! That's really useful, but it doesnt account for the 25% DPS gap between arcane and frost in heroics. Every other variable should impact all specs equally (strat differences, ilvl AND player skill). WOL/raidbots is perfectly fine for citation. Demanding people stop citing for personal reasons is very strange.

That's like saying that the DPS differences would be the same at ilvl 560 and 570. Water elemental's waterbolt is about 13-15% of people's DPS, depending on mastery, so that is actually a very large factor.

The issue is that both proponents and opponents try to generalise, because the OP requests a generalisation. I simply believe we are unable to make a fitting generalisation with the data we currently have. Unless anybody feels like making a way more advanced version of raidbots that includes Warcraftlogs, and has the option to exclude Asians.

[Deleted]

High end raiding wise the best specs would be

Frost(With good RNG on procs) > Arcane > Frost (with bad rng on procs) = Fire(With good rng) > Fire(With bad RNG)

Lmao. /10chars

[Coldvibes]

Data is going to be pretty stupid right now due to a good number of guilds doing pure cheesing in every format and in both forms of logging.

[Xomniuri]

The day i can stand still and turret almost the entire fight i will spec arcane (As it is currently with RoP), but until that day my 10man group requires more of me than just dps.
25man is a much better environment for arcane, hunters warlocks and shamans seem to be the ranged classes that do the mechanics before any of the arcane mages do

Any raid leader with a brain will try and keep any arcane mage movement to a minimum when assigning duties

Moving 10 yrds into a new ground aoe heal and having to drop a new RoP gets old after 10 minutes

i really enjoyed arcane as it was in cata, but then they go and feck it up with RoP

At the end of the day it comes down to
a) what your raid group requires
b) what you perform best as with your current gear
c) what you prefer to play as

Personally i went back to frost after being fire through heroic ToT and Normal SoO because we needed better add control for Heroic Nazgrim and there was no other class in the raid that could provide more other than me, the next time we went in adds were cake.
I considering staying as fire until we went into heroic shaman which needed better aoe for slimes, after that it just wasn't worth the effort to go back to fire, frost was performing consistently better with the duty the raid required of it.

Breaking down the benefits of frost over arcane for the first few heroics which most are progressing on
Immerseus - Frozen orb rocks for adds, ice lance spam anyone? - RoP running around all the time no thanks
Protectors - Frozen orb is very nice for rooks adds - Arcane does rock on this fight providing morons stay out of your RoP area to keep poison pools to a minimum
Nourshen - pet freeze & deep freeze the big adds 100% crit on LB don't mind if i do, also keeps dmg off the offtanks
Sha of Pride - frozen orb on adds, pet freeze and deep freeze manifestation, i can manage a LB during the freeze so not only do i get a 100% LB i also get the full amount Ice Lances from the FoF deep freeze - RoP during the last burn phase nuff said
Galakras - I'm the grunt killing bitch so freeze, LB and deep freeze adds can you say 100% LB
Iron Jug - RoP management through that fight is an effort and a half
Shamans - slims + frozen orb = cake - RoP being a bitch depends on how many tornado's are stalking you
Nazgrim - pet freeze + deep freeze + pally stun = dead shaman with no chance of a totems, brain freeze FFB or frostbolt adds, my granny can move faster, and if you need to save a FoF proc for the banners.

You get the idea, no frost mage is going to out dps an arcane mage doing certain jobs like add killing on galakras, frost is not better or worst than arcane or fire it just needs to be doing the right job the right way as with any other spec