Also, I still have issues with the hypothesis that:
“Higher percentiles show higher skill and differences in percentile can demonstrate scaling due to optimal play of the class”.
This doesn’t make a whole lot of sense to me. If we’re talking about universal scaling from gear than skill differences at the same gear bracket shouldn’t show a marked difference. The percentage of benefit should be roughly the same - as the benefit is from gear and not skill.
Instead, what you see in the higher percentiles is a narrowing of the gap - this doesn’t represent better scaling but rather how well a class performs at peak. It’s well known that Fury benefits from optimal play. And if we want to argue that Fury is doing okay compared to different classes this is a serviceable metric to use. But I don’t think it shows scaling.
Well, the data I was just looking at was all percentiles at different ilvl brackets. (Although the highest ilvl bracket is also the highest percentile overall, as they are correlated).
I don’t think it’s too big a problem to look at higher percentile parses (although the extreme top parses is often a poor choice due to other factors). It generally ensures that you are comparing people that are largely optimized in the fight and gear.
Lower parses can lose sight of what the class can do balance wise, and other stuff like relative skill requirements become a larger factor in the DPS comparisons (where Ret has a distinct edge because it’s super easy to play near optimally).
I personally prefer to look at upper percentiles for those reasons, and remember that most of the balancing decisions we are dealing with now are as a result of people over a decade ago playing this game at a level that would be a gray/green parse today, which skews things pretty mightily.
Looking at near top potential (but not absolute top) helps to focus on on what the class is capable of and factor out some of the extraneous stuff that gets in the way. (For example, probably don’t want to be including a green warrior parse as a negative for the class because the guy was on track to legendary parse but died halfway through the fight, but that stuff is all included in lower bracket ranges.)
As far as scaling goes, the people at the top right now are expected to be the most geared right now, which is generally holding true. So while it isn’t proof positive of scaling, it does show that the gear helps move the 99th percentile upward as you move through the gear brackets.
1 Like
I think a better way to measure that is through iLvl brackets.
While I understand your view in terms of higher performance players being a better quality measure, I don’t fully agree with it on two counts.
-
I don’t think it accounts for classes with a higher rng damage cap (like Warriors)
-
I don’t see it as a relevant variable for measuring gear scaling when used in combination with iLvl brackets.
On number two, I think percentile can be used as a poor proxy for analysing iLvl brackets, but not in combination. Combining both to reach a conclusion seems like massaging the numbers. Basically the scaling from gear for those at the same iLvl should be roughly the same. So cutting the data into iLvl groups and then cutting it again by performance percentile seems to be over working it to me.
I can see a point for using percentile as a proxy for iLvl as you did in the past when iLvl data wasn’t available. It’s not a strong case but I can see a case there.
But Kelliste’s method of breaking it down by iLvl and then breaking it down again by percentile makes no sense to me at all to be honest. Not as a measure of scaling from gear.
Percentile as a measure of skill difference has no bearing on scaling from gear. Or at least I don’t get the connection.
As a proxy for higher gear levels there is some value but not strong value.
That’s what I did, you see it consistently moving upward.
Looking at a set percentile across brackets is a good way to see how that level of performance is impacted by gear.
Across all of the percentiles (both the fixed high ones and the all percentiles at once) you see fury warriors damage increasing, which is the scaling from gear.
That’s why it can be a mistake to look at 99-max percentiles too much, but 95th is usually a pretty decent indicator as it includes many that are just getting average luck but are playing largely optimally.
Even for warriors though, the top parses for them usually isn’t that much of an outlier from the rest of the pack, so I’m not sure that RNG is too heavily featured in their parses. The longer fights in Ulduar help a lot to lower the impact of pure RNG on parses.
And I accepted that. It does look like we are scaling better than Ret - though we double dip on the weapon damage upgrade so I expect that to even out over the phase.
I don’t agree here. It represents a marked shift and it is correlated with things like slam procs. The box plot shows it pretty clearly that there is a significant peak in performance that is not normative and much more so than for many other classes.
I agree that the 95% mark is much more normative and a good measure though.
Edit: I don’t think I disagree that much with your position actually. But we are emphasizing different things - perhaps representing our respective biases.
I get a bit tilted though when people make definitive claims that the data isn’t exactly clear cut on - especially this early in the phase. Not being open to discussion on these points at this stage seems particularly closed minded and a bit arrogant (this is not pointed at you btw). Especially when they aren’t initially open about what numbers they use to reach these conclusions.
As a spec we get our fair share of people making extraordinary claims about how we supposedly scale and how good we’ll be in ICC. You can’t really blame me for challenging how strong the foundations for those claims are. I mean in some peoples heads Warriors in ICC are the boogie man. I’m just trying to lead to more grounded discussion on this topic rather than all the myth and assertions that fly around masquerading as “fact”.
Do we scale well with gear - yes we do. How well? That’s an open question. How will that pan out in this phase? Also an open question.
For the record, it’s this claim (that didn’t have any accompanying evidence) that first got me pushing on this - because it’s flat out wrong as a blanket statement:
Yeah, and the extent of it will take a long time to figure out precisely outside of just reading sims.
So far it has been looking like fury doesn’t need intervention for their DPS this phase though. It certainly starts off in a bad spot, but the higher gear warriors are looking pretty decent overall, and people should be trending toward that over time.
Yeah - and I’ve tentatively changed view on this. I’m now in the hold off wait and see camp.
To be precise - I think our performance in this phase will give us a good basis to judge how we’ll perform and scale in ICC and whether we can make up the difference.
Yeah, warriors are a bit of an exponential scaling class as your damage feeds your rage which feeds your damage further.
I’m just glad to see both ret and fury doing better than they were in phase 1.
I’m not as convinced about this in WoTLK. But I’m open to being convinced. Thus the wait and see approach.
Feels good being passed over as Warrior dps because now not only do Rets have comparable DPS, they still also have 100x more utility.
Raid comps with 1 War snd 3-5+ Paladins “Sorry, full on Warriors.”…
This phase will be putting people in ICC 10m ilvl or very close to it. There is no wait and see, the data is showing an upward trend which has already been proven.
Cool Warriors scale well.
Still being passed over for Paladins because of their insane utility and now comparable DPS.
Been looking for an Ulduar pug on my War for over an hour and I either just get ignored or told no right on the spot.
Can log over to my Paladin and get spammed invites before I can even close the LFG tool.
Which was the main issue people had with Rets getting a buff to begin with. They were never really shorted of raid spots.
Whens the last time you saw people spamming LF1M Warrior compared to LF1M Paladin?
Big problem is alot warriors dont know how to play. Synced weapons 1 munch deep wounds 2 clip your oh. If you know how to heroic strike cancel then you might have a chance. Pls dont tell them to ruin the class cuz scrubs dont know how to play. Warrior is actually the deepest class in classic when played 100%. If you dont know heroic strike cancel is when you queue hs it gives you oh yellow roll meaning it wont miss and can crit ect. It completely changes the class and gives you control of your rage. id say less then 5% warriors know about it but top warriors been doing it since vanilla. I have been pushing it on warrior discord. Also the gap will not change until late toc.
If you dont know about heroic strike cancel then you will never ever be orange parser come icc. You might be able to out dps some classes but top warriors will destroy you.
By putting /stopcasting at end of heroic strike u turn it on always and only turn off when you drop below 20 rage and cant BT/WW. Alot warriors say they hs around 40 but that is if you dont know how to do this. This was done in classic/tbc/ and now especially in SoM it was sweatest of sweat. I highly recommend learning class better go to warrior discord. Also heres link to EJ fury breakdown it explain literally everything.
https:// web.archive. org/web/20100926014150/http:// elitistjerks. com/f47/t37804-warrior_fury_compendium
And I’m yet to see the proof.
You claim you have it but you haven’t supplied any.
You can’t either, because we don’t actually have enough iLvl data to prove a reliable trend.
Not only me but others have shown it as well.
Well I’m talking with you and you haven’t.
I’ve supplied data for why I think you’re wrong but it’s based on the data I think you’re using as you don’t actually supply any.
And the whole “we all have the data” is just you speaking to the popularity of your case. You are not supplying data by saying that.
How hard can it be though? Others have supplied data and in some cases it changed my mind. In your case you just assert “facts” and claim “the data” is there. Cool where is it?
The one piece of data you did refer to had a category with exactly 39 Fury parses. Excuse me for discounting that. I mean slice the data enough and eventually you’ll get a bit right?
You said there is a trend? Show a trend then. With a reliable sample size. Else admit that like me, you don’t know how Fury is scaling yet.
I’m actually not understanding why Rets like you and others are so invested in showing that Fury is scaling well and shouldn’t be looked at. The data for and against is weak why not just wait? Why proclaim we “know” stuff we actually don’t know?
By the way, this all makes me look defensive, but let’s go back to the original point. If we look purely at item level it has Ret not Fury pulling ahead. It’s only because I’m being fair and acknowledge that sample is screwy that we are here arguing about your conjecture.
If I apply the same standard to your data it’s just as screwy. So, I look defensive but I actually have data that you would be pretty confident in if my cause fit your bias.
The fact is neither of us can say yet with confidence how the classes are scaling because strong realiable data isn’t available yet - wait.
1 Like
This has been the concern of a small Ret community aside from the vocal forum majority demanding changes…the larger and more long-term concern was the net-negative precedent it sets for the broader community to ask for (in spite of Blizz stating this was “not a goal”) and perhaps expect similar review and changes. This thread in some ways proves us right, without directly commenting on the state of our plate (but Light-challenged) brothers.
1 Like
“We set class mechanics to 3.3, but under no circumstances will we adjust knock-on effects of wild imbalance that impact social interaction and limit choices and reduce the value of this resurrected product because we believe in the original game’s integrity – to say nothing of our limited resources – so TTYL guys, we need to overhaul every single dungeon plus Ulduar”