I can make it short and save you time.
Before any math happens, data sets need to meet specific criteria to make specific assertions. This one doesn’t. Doesn’t matter how much math they did, doesn’t matter if every cell of their formulas is perfect, the underlying data set can’t support the inferential claims because of how it is collected and what it is.
It looks neat, it means zero.
I see. Well, thank you for your summarization. I take it you’re the one that landed on Scrotie’s ignore list. 
I removed him from mine long enough to try to help him, and he played his usual “i’m completely lacking in formal education on topic X, but I know more than anyone else contributing here will ever know” act and I put him right back on ignore.
It’s sad, honestly. He didn’t even realize that several of the people who commented were basically patting him on the head and saying, “that’s cute, kid. Wrong, but cute.” I was just the most direct, which I thought would help him maybe question his methods, but there’s no breaching that wall.
There’s nothing wrong with speaking frankly.
There is when people aren’t ready to hear it.
Until someone accepts that they have a deficiency or gap in their understanding, nothing will help them learn, so I just stuck them back on ignore so I don’t have to see their sophistry.
The point of the thread is for others to know it.
Nope. I probably should have. I probably should. As you can see, Neon does not argue in good faith.
Differences in matchup winrates exist between ranks, and these differences are relevant.
I swear, you made a topic comparing the skill ceiling of various decks to show that control and aggro aren’t so different. I believe that your conclusion was aggro takes more skill.
This reminds me of that. Certainly, people disagreed, implying a “big brain” is required to play control. Maybe it does – just look at Schyla.
Eh, it doesn’t just jump immediately to zero meaning.
There’s enough data available to make rough estimations. As more weeks go by without tuning changes, you can expand the data set and gain more confidence in them.
There’s definitely enough data available to identify which decks don’t perform as well (or better) at high ranks compared to diamond, which is reasonable to attribute to skill differences due to the average deck piloting skill being better up there.
You can’t measure the skill directly, but you can very solidly say that there is more going on with the deck performance than simple arbitrary deck choice differences can explain.
I certainly enjoyed reading this discussion.
I’m just a nobody, with no proficiency in mathematics-
Yet…
I could still see Neon was making logical points, especially on whose burden of proof it was. He was pretty respectful too.
Actually, that thread — which had a less refined method than this one — showed that on average control decks reward skill more than aggro. However, this difference is not as large as one might imagine, and there are specific aggro decks that reward skill more than specific control decks. So it’s a general trend, not a rule
Oh, okay, I must have misremembered. Nevertheless, appreciate you.
Well I wouldn’t say you misremembered, so much as mixed in some of the misinterpretations in the replies. I said that some aggro deck or another was high-ish skill (if I remember correctly my words were “the thinking man’s aggro”) and people clowned on that. As they predictably would.
And Enrage Warrior right now is a rather “high skill” deck, too. (Decks don’t have skill, people do; what I’m really saying is that it rewards skill rather well.) It’s not Rainbow Mage or Control Priest, but it’s not that far behind. Near the bottom we have the Paladins and Arcane Hunter, but also Control Warlock and (slightly above these others) Blood DK. Control’s “skill advantage” is very slight.
Which control deck, would you say, has the “highest skill advantage” (as of right now). I’d like to try it. I also like that term; I don’t recall hearing that before.
Well here’s the answer in chart form:
| Deck | D4-1 | skill effect | meta effect | T1KL |
|---|---|---|---|---|
| Control Priest | 47.69% | 2.34% | -0.82% | 49.21% |
| Rainbow Mage | 50.85% | 1.23% | -1.45% | 50.63% |
| Enrage Warrior | 54.18% | 0.77% | -0.39% | 54.56% |
So Control Priest is technically the deck that rewards the skill difference between Diamond and Legend the most, but this takes it from bad to mediocre. Really if you’re a good pilot I would just recommend Enrage Warrior for maximum advantage, which is also what I’d recommend to bad pilots. Did I mention that it is hands down the best deck in the format?
But if it has to be a CONTROL deck then I guess Rainbow Mage.
Thanks for giving me the answer I wanted, he-he. Rainbow Mage is probably my favorite deck right now, followed closely by Plague DK and Thaddius Warlock.
@ScrotieMcB – Should I use the highest winrate Rainbow 
Mage at top 1,000? Or would you recommend using Rainbow Mage decks at my respective rank?
IDK man ask Schyla or something?
Okie dokie. I’ll see what he thinks whenever he gets online. 
No, really, it does. If you have no precision you have no specificity.
You clearly don’t follow… every data point of hsreplay ever recorded wouldn’t yield a reliable indicator of skill when used in this way. All of it. It isn’t about insufficient quantity, it’s about the insufficiency of the quality of the information.
It’s not good for any estimate of skill on a deck level. At all. Ever. In any way.
But not because of a single factor like skill. Because you didn’t actually measure any skill.
This is your assumption, which you didn’t prove, which makes you guessing, not doing a statistical analysis.
It’s only skill because you believe that’s what the deciding factor is in your heart of hearts because nothing about this analysis measured skill in the first place precisely because of exactly what it is.
Then you aren’t able to make claims about it. That’s exactly how all this works.
But you can’t describe how much, in which cases, and when… and the sum total of that is your analysis is garbage.
I mean, look, I will say this one more time: Anyone with basic proficiency in statistics will tell you that the problems with the analysis are the data, not the maths because someone with even a basic data analysis background wouldn’t have even started to manipulate this data set towards this end, knowing it was impossible to make reliable inferences.
This is the part that stats people understand that lay people are confused about. Before we did any math we had to meet conditions that aren’t met, so the math means zero.
He is talking out his backside whatever answer he gives you, so just play what you think is fun.
So I ask you again, neon: control priest versus enrage warrior at diamond 4-1 is 53-47 in priest’s favor. At top legend, the same matchup is 59-41. What causes this difference? If it is all statistical noise, then all of our data in the matchup win rate tables means essentially nothing.
The answer he will give is:
It’s impossible to know.
Which, is technically true, the numbers don’t dig deeper than surface level “it wins more.” You technically can’t prove causation with the stats we have.
But uhh… there’s really only 2 things that are going on:
RNG from draw and card effects
And
Piloting differences (i.e. skill)
There’s no reason to believe that top legend priests are just luckier than D4-1 bracket (although the larger sample size does push it toward the average luck on both sides more, we don’t see this matchup varying wildly from week to week in upper legend either).
What drives the piloting differences doesn’t really matter, it’s still skill based differences. We obviously can’t say “it’s this particular aspect of skill causes the difference,” but it’s definitely possible to say that X deck performs better/worse when in an environment where players are piloting decks closer to optimally.
The numbers are good enough to do that.