Polarization and the effect of skill, DR#279

Multiplayer Discussion
1

This is the second installment of a series. The first, which explains how the effect of skill on winrate — specifically, the skill difference between the average Diamond 4-1 player and the average top 1000 Legend player — while factoring out the effect of meta differences, can be found

This report covers Standard from after the Azurite Snake hotfix to before the balance patch changing it to 7 Health steal.

PART 1: ON POLARIZATION

However, for this installment I’m also including polarization. Polarization is a measurement of how lopsided matchups are. Usually, it’s calculated by looking at the difference between winrates in a matchup — for example, a 55-45 matchup has a polarization of 10%, regardless of which side of the matchup you’re on. In the results below, I call this raw polarization. That said, being a deck that wins a lot or loses a lot in general means that your matchup winrates tend to be further from 50%. I’m looking at some decks with winrates below 40% here, which get a high raw polarization score just for being bad. Thus, I also made an adjusted polarization score that factors in how far the deck’s overall winrate is from 50%, isolating polarization that can’t be fixed with nerfs or buffs.

I’ll always be calculating polarization for Diamond 4-1. There isn’t enough data published by VS to do satisfactory analysis at higher ranks.

PART 2: THIS WEEK’S RESULTS

Spreadsheet of all calculations link:
https://docs.google.com/spreadsheets/d/1YVadtfOz2i6h450x3ymGqH2zhsI9Nb_DodmQi9yxT68/edit?usp=drivesdk

Class Deck skill winrate diff % skill vs meta
Rogue Miracle 3.55% 85%
Priest Reno 2.72% 53%
Priest Control 2.66% 57%
Shaman Reno 2.05% 85%
DH Naga 2.03% 98%
Paladin Reno 1.63% 42%
Mage Rainbow 0.85% 38%
Rogue Ogre 0.70% 36%
Shaman Elemental 0.45% 28%
Rogue Thief 0.34% 28%
Mage Rommath 0.27% 51%
Druid Reno 0.14% 9%
Warrior Control 0.12% 10%
Rogue Mech 0.12% 25%
Mage Secret -0.01% 1%
DK Reno -0.18% 47%
Priest Undead -0.19% 33%
Warrior Enrage -1.02% 31%
Shaman Totem -1.15% 56%
DK Unholy -1.47% 35%
DK Plague -1.62% 73%
Hunter Reno -2.04% 73%
Paladin Pure -2.11% 47%
Warlock Snake -3.40% 89%
Hunter Arcane -3.40% 88%
Druid Dragon -4.62% 87%
Hunter Cleave -4.64% 87%

The average effect of skill last week was about 70% relative to the effect of the meta’s 30% last report, but this report it’s about 50-50 (somewhere between 46.95% and 53.25%, depending on how we weight the average). I’ve had running debates within the community where I claimed that the effect of skill is consistently greater than the effect of deck popularity, and that doesn’t hold this time. It’s worth noting that both of my previous analyses, the first using a more primitive method, looked at relatively stale metas just before an expansion release. This report covers a much fresher meta, which might explain the difference. That said, fresh metas matter too (arguably might matter more) so it does appear that my hypothesis there is disproven.

Control Priest, whether in original or Highlander flavor, remains the deck that rewards skill the most, but that still just takes the deck from bad to mediocre at best. I suspect that the introduction of Miracle Seller has made Rainbow Mage slightly easier to play, but still rewards skill more than the other Mage archetypes for which VS collected data. Cleave Hunter and Dragon Druid reward skill exceptionally poorly.

Class Deck adjusted raw
Druid Dragon 20.71% 20.72%
Hunter Cleave 17.74% 20.72%
DK Plague 16.58% 16.57%
Druid Reno 16.18% 15.98%
Shaman Totem 15.55% 15.85%
Warlock Snake 14.95% 13.58%
Warrior Enrage 14.86% 16.46%
Paladin Reno 13.96% 13.19%
DK Unholy 13.68% 19.90%
Mage Secret 12.88% 14.06%
Hunter Reno 12.47% 11.61%
all average 12.30% 13.99%
Rogue Miracle 12.29% 23.06%
Rogue Ogre 12.17% 17.21%
Mage Rainbow 11.56% 11.29%
Priest Reno 11.46% 14.32%
Shaman Reno 11.41% 11.46%
Warrior Control 11.30% 13.91%
Paladin Pure 11.14% 15.92%
DH Naga 10.93% 10.71%
Shaman Elemental 10.84% 19.89%
Priest Control 10.35% 16.03%
Mage Rommath 9.12% 12.97%
Priest Undead 8.62% 9.04%
Rogue Mech 8.55% 8.69%
DK Reno 8.25% 11.27%
Rogue Thief 7.91% 23.36%
Hunter Arcane 7.83% 7.90%

The average match is currently 57-43 in favor of one player or the other.

Dragon Druid and Cleave Hunter are both exceptionally polarizing decks. Generally speaking any polarization above 20% is practically doomed to become a sentiment outlier; if your average match is 60-40, then half of your matches are even more polarizing than that (approximately, it’s a mean not a median). This, combined with how easy the deck is to play, is enough for me to deem the deck the second coming of Standard Questline Warrior. It doesn’t seem overpowered in terms of winrate, technically speaking, but the other two factors make it bad design. Cleave Hunter was almost as bad, but it, unlike Dragon Druid, already received a nerf tap.

2 Likes
2

If you keep telling people Control Priest requires skill, you’ll instantly lose credibility from at least half the forum goers.

Despite the fact you’re right and that my experience with it and other decks agrees. Bad, Non-Control Priest players won’t ever understand why but ask people who actively spectate me when I play it and I think it becomes obvious.

I like the challenge, but at the same time it’s so frustrating that I have to play so perfectly while these other decks can reward sloppy terrible plays with wins and I’m only rewarded with a mediocre deck at best like you say.

Thus, I took Cleave Hunter and can take Aggro Paladin to legend quick with little to no resistance and it feels like I put the game on easy story mode.

Before I read your stats though, I was wondering if Rainbow Mage would take some skill because I played 10 games with it and found it was deceptively skilful to know when and how to gain or find lethal. I would have put it close to Control Priest before reading the stats. Seems like I wasn’t too far off.

Overall, it looks like if you have a deck that has to find a way to win or calculate how to get the win, it’s much harder to play than a deck that just spits out a win simply by playing your cards on curve like Aggro Paladin.

4 Likes
3

While all that’s true, my point here is that you’re not alone. The control priest population isn’t large but they’re mostly in the same boat. It’s definitely a deck that I would argue for buffs, because if it got a 2-3% bump in winrate then Reno Priest would be a skill testing, below average polarization deck that’d be between 49% and 51% winrate at both Diamond 4-1 and top Legend.

Although Pure Paladin is a deck that doesn’t reward skill very well, it’s still a full tier better in that category relative to Cleave Hunter and Dragon Druid. The main thing that was gross about Pure Paladin before the nerf was its overwhelming winrate in general. I mean, when a deck like that is the most popular at top Legend you know someone on Team 5 messed up.

Last report it was even closer. And I don’t think it’s very deceptive, Mozaki Mage was also a deck that heavily rewarded skill and Sif is a rather similar card.

1 Like
4

Having lost only to pure Paladin consistently on a rainbow mage sweep from bronze to legend (in one day), I’m fairly certain that a large chunk of that win rate, both before and after nerfs comes from countess and the fact she is tutored consistently via order in the court.

I wonder if they will ever nerf that card. In any case, I think it may be worth teching the snake oil shuffle into opponent deck into rainbow mage to shore up it’s only discernable weakness afaict.

5

I summon Neonghost to take down the sophistry .

3 Likes
6

Lost me at this is a skill based game, this game takes very little skill, some luck, and mostly the right deck. How are you attributing increased win rates to skill and not just better luck of the draw with cards?

1 Like
7

No way! Haven’t you heard what our stat overlords told us? The best win rate cards for pally are their early game, which went untouched! Cards like countess or order win rates just couldn’t compete! The win rates can’t possibly change that much!

Pallies must just suddenly lost over 9000 skill points because reasons. How could they possibly lose if they just curve out 1-2-3?

1 Like
8

It’s largely just that pure control in general is typically more skill testing. If it can find wins, it usually requires pretty solid planning through all phases of the game to do so consistently.

We all remember the nightmarish games where priest drew the right removals, and copied the perfect threats to end the game in a way that feels super bad.

What we dont see are the many more games where the removals dont line up right, or decisions between using a specific removal now, or holding it for later is the difference between a win or a loss.

Every card in priest feels like it’s designed specifically to make your opponent rage quit, but getting wins with them don’t happen freely.

I don’t think it’s necessarily a fresh meta effect you are seeing here. I think that high polarization tends to just lead toward your skill mattering less overall.

Sure, you could maybe play your cards better to get 3-4 more % against dragon druid, but if they were beating you 60% of the time and are a large chunk of your matches, you are really just better off playing something else.

Sometimes skill differences can overcome polarization seen at D4-1 as you climb. We have frequently seen it before with hunter and paladin decks.

Sometimes it doesn’t, as the plays you are up against just don’t have reasonable answers that can be employed by a variety of decks, and you start seeing the game shift toward a more meta driven, rock paper scissors setup.

9

Again with the skill, most people reach maximum skill level needed to play this game very quickly. Its a deck based luck of the draw type game, most match ups you can concede after round four because the game is decided by then

The decks are designed to be played a certain way, where is the skill in just doing what you are suppose to do? If you get the right cards you win, if you dont you lose. Knowing what to do with the cards is a very small learning curve. The decks are suppose to make you feel clever by doing simple math and combos and it works.

10

I wouldn’t call removals not lining up right skill testing. It’s luck of the draw.

It’s the equivalent of aggro top decking the right card to keep winning on board (e.g removing enemy board like a big taunt for cheap), or land lethal, or buff their board to be awkward to remove, or reflood/refill, etc

11

You know that i can use this text to non ironically tell that cassino decks are the most skilltesting decks of this game don’t you?

Because good luck knowing what to do with they before the game starts.

12

I have no idea what a casino deck is I believe you though

All I am saying is that sure skill makes a difference to some extent, a very very small difference, but most of this game is straight luck of the draw and deck build. This is not a game that is difficult to master, the fun of this game is suppose to be in part the randomness of the deck but most of these decks are now autopilot and play themselves.

If you draw the right cards as an aggro pure pally nobody is beating you. If you dont draw the right cards it becomes more difficult to win , the worse the cards the more skill needed, but the cards are random luck so the main factor influencing winrate is luck of the draw and deck.

13

There is no such thing as maximum skill level.

Will a lack of skill feels like is: there’s an opportunity for a play presented to you, and you simply do not see it. It is not abnormal for you not to see every single play that you could, because no one sees every single play. But the difference between you and a wise person is that you believe that you can see everything that is.

14

Yes there is. Your skill is limited to maximizing the deck based on the cards draw. Which is a very easy skill level to hit. What do you mean there is no maximum skill?

I do not see it? What exactly are you finding so difficult to master? You have mana crystals, you play cards that use mana crystals, most decks are designed to pull ahead in the match by turn 4 based on draw. What am I missing? There really isnt a choice in what cards you play to open, there are the correct ones for your deck and the incorrect ones , you either get them or you dont, if you dont get the openers you need and your oppenent does you will lose.

I any luck of the draw card game you have to just be good enough to get lucky. In this game its very easy to be good enough to get lucky. I think you are seeing a level of skill depth that does not exist here.

Put me up against the best player in the world and I will beat him pending on the draw.

15

Eh, there’s a reason that priest consistently has better matchups in upper legend compared to lower ranks. They aren’t just lucky.

There’s a major skill component in using removals most efficiently. Doing so requires you know the opponent’s deck and their power turns very well.

Anyone can get lucky and draw the right removals to handle the right situation. There’s a lot of games that can be won even when you don’t. Priest usually gains win rate % and jumps a tier between diamond and top legend due to people doing that better.

There is an upper limit of what skill can do to improve your chances of winning, but it’s not something people will ever reach. It requires more processing power than our brains have.

16

It depends on what you mean by “small.”

I’m not concealing in any way that the skill difference between average D4-1 players and average top Legend players creates a winrate difference of AT MOST about 3.5% up or 4.5% down (players playing against the dumb decks better). It’s not just a list ranking decks relative to the other, it also has actual numbers.

That said, the difference between a Tier 1 deck and a Tier 2 deck is often about 3% winrate, and that’s not a small thing when you’re climbing. A 3% increase in winrate saves 4-10 hours on average for a climb from Diamond 5 to Legend. Squeezing out that extra couple percent is clutch.

You are looking at this from the myopic perspective of a single game. What if I put you up against the best player in the world… for a set of 100 games? Do you actually believe luck could save you there?

2 Likes
17

No I would get crushed there, one game getting lucky is what I meant.

I was just trying to say the number one factor in winning or losing is luck, number two is deck build, and skill is a distant third.

I see what you mean about over the course of many many many hours the skill starting to factor in in terms of rank so sure it can help you climb faster, but anyone can get to legend over a long enough period of time with the right deck as long as they have hit minimum skill level.

1 Like
18

Not really. I made a thread a bit ago about the effect of winrate on grind time: "Brute forcing" to Legend

Assuming we’re not talking bots or people who don’t need to work for a living, it’s essentially impossible to play more than 240 hours of Hearthstone in a month, and rank resets monthly. Below a certain winrate it just becomes unfeasible given the time given.

But it is true that if you’re merely average or slightly below and you play a lot, you’ll probably make Legend. My point is that if you’re horrible then you won’t have enough time.

19

I wouldn’t put luck that high. It’s not most games that are determined by draws so lucky/unlucky as to determine the game outcome on their own.

Skill also mitigates the luck factor, as you are more likely to win games that aren’t being handed to you by having your perfect power plays on curve (or ever).

Deck build / opponent matchup can sometimes outvalue both of those though. Sometimes matchups are polarized enough that there’s not much room for skill expression, or even luck to swing the game in the opposite direction.

The weight of the things impacting games aren’t really fixed values, and the ranking of them can swing wildly from match to match.

This makes me glad that over my 20k or so hearthstone games I have averaged a 54% win rate.

Although in the diamond brackets my win rate is usually significantly higher.

20

Regarding the difference in skill between average D4-1 players and average top Legend players, and the difference in deck popularity between D4-1 and top Legend, we can measure the effects of those on winrate, and like I said in the opening post, in the previous report skill had a BIGGER effect on winrate, and in this report they’re about 50-50. Actually measured.

That said, you could arbitrarily say that the difference between D4-1 deck popularity and T1KL deck popularity “isn’t that big” but the average skill difference is, so it only takes a little metagame to make as big of an impact as a lot of skill. Everything in my statistical analysis is relative to D4-1 and/or top Legend, measuring the “distance” between the metas and isolating either skill or deck popularity. I don’t and can’t measure the “distance” from nothing at all to D4-1, so we’re all free to hypothesize what goes on there.

Still, I think it’s luck first. By a lot. As I recently said, the decks MOST effected by skill gain ~3.5% from D4-1 to T1KL, the average effect is more like ±1.75%. Meanwhile, the average polarization in D4-1 is ~14%. That’s a 8:1 ratio. Depending on the meta, skill and deck popularity might be 70-30 or 50-50, and some metas are significantly less polarizing, but right now it’s about 80% luck, 10% skill, and 10% deck popularity.