Tier Lists: HSReplay vs ViciousSyndicate

General Discussion
1

I’ve watched discussions about tier lists pretty much since I started playing Hearthstone. Often people discuss that archetype A is T1 in one site, but T2 in another site to make a point. Other people, trying to be as nice and informative as possible, opt to say the tier in at least two websites. I want to clear some of that stuff up so that we can have more productive discussions.

First, lets start with something that is rarely mentioned but is extremely important: the definition of a tier is completely arbitrary. This means that if I start a website today I can make up my own definition and it’s never right or wrong. Each website has its own convention to what a tier is and how an archetype gets into a tier.

As an example, TempoStorm’s definition is 100% expert based. In other words, it’s an opinion, an informed one, but an opinion still. HSReplay and ViciousSyndicate are data based. However, tiers are defined differently.

ViciousSyndicate’s tiers are based in exact win rates. As an example a T1 archetype is one that has a win rate higher than 52%. HSReplay’s tiers are based in win rate distribution, meaning, given all win rates of all archetypes, how “distant” are they, in a gaussian distribution) from the average 50% win rate. A T1 archetype is one that has a win rate that is 1 standard deviation higher than 50%.

So… which one is better? That’s a tough question to be honest. I’ll explain the implications and you decide for yourself.

ViciousSyndicate’s tier list thresholds are static. This makes it consistent and immune to meta variance. It also provides some pretty good info, for instance during the Un’Goro meta you saw a period of time with no T1 decks. This told us somethings, for instance, that there was an enormous balance. Even the most hated deck at the time was actually pretty balanced overall… its problem was polarization.

HSReplay’s tier list threshold vary because they depend on how much win rate variance exists. Depending on how often the calculations are made, they may very well vary every second, although I doubt that. While this doesn’t make it easy to compare over time, it is pretty strong for the immediate analysis. In a very tight meta, with a very little separating archetypes, HSReplay allows an immediate identification of the over performing decks.

So, it’s up to you. Do you want a stable, longitudinal way of looking at tiers? Or do you prefer a dynamic, carpe diem kind of way? There’s no right or wrong choice, as long as you don’t compare the tiers.

2

I prefer the clarity of VS. At least there you know that a T1 deck has reached a certain winrate threshold. In HSR, a deck can keep the exact same winrate, but go from T1 to T2 over the course of a few weeks. Makes no sense to me.

3

what about the win rates on HSR? are they real in the sense of the decks actually winning that much or are they relative as well? because they seem to be always a couple of percentage points higher than on VS.

4

I don’t trust either source, their stats clearly don’t apply to me and I at least feel like there are always better lists than the ones shown there.
It is irrelevant for people like us to know that priest is bad and hunter is good and other things like that, the only useful information is what cards are popular in certain decks and which decks are popular, winrate is just an abstract notion that people like to speak about like characters in cartoons like speaking about (speculative) science and “philosophy”

In my case I take lists from pro players besides streamers and Zalae who’s dumb. Stats about winrate from experience and stats about popularity from those sites.

Hope this was taken as an unpopular opinion not as the words of an A’hole

5

those stats are mighty useful when trying to reason with certain people.

6

If you want to win an argument, if your goal is to arrive to a conclusion that helps to both parts growth and well being then not really, in certain cases sure but you, Vlad, and the other people here relying on those stats to win their arguments might lose sight of this when debating in those 70 replies posts.

1 Like
7

hard not to take this personally. the nice thing to say is that you don’t understand what we’re trying to do. the other things I’ll keep to myself.

8

https://www.reddit.com/r/CompetitiveHS/comments/8ekl7h/reading_numbers_from_hs_replay_and_understanding/

Should be an educating read for those people who blindly refer to stats all the time when the cluelessness is so obvious.

9

should have kept adding this kind of statements to underline that I don’t speak to offend here.

I just think that if there are people constantly saying how people here (seem) to speak to defend blizzard rather than to have a discussion this is a thing to consider.

While sure, when people say “You always loose when Boom is played” using stats is a start (though I’d rather try to address the feelings than the facts when people make this kind of statements, but again, I consistently loose debates for these reasons) but it will lead to more salt from that person, it is justified though, that’s the point, even if not beneficial, that’s the other point.

10

TL&DR: VS aims to reduce bias of their data and focus on the overall meta, therefore, their reports are more “trustworthy” for most people than HSR. However HSR is more feature rich and constantly updated. Speaking from professional experience, VS’s has all the features that I would expect from top notch data analysts, engineers and scientists. I’ll happily give them a seal of approval based on my experience.

Keep in mind that the point of this thread was to shed some light on tier list thresholds only.

Now for the long version, for those interested.

Yes, that is completely true. As far as I’m concerned what I like about HSReplay is that it’s constantly updated and feature rich. VS is a lot more conservative, even if you have the golden subscription which I’m not sure how it’s working right now since I don’t have it.

They are real, just calculated differently from VS. I remember discussing it with people from VS more than a year ago and their objective was to reduce self reporting bias, meaning, that they understood that people that made their data available to them were a very specific population. As far as I know HSR doesn’t take bias into consideration, it’s just pure data. VS made some decisions based on that to reduce that bias, hence, it is “safer” to trust VS reports. The problem is availability. HSR has more features and is updated continuously.

1 Like