This is meta data from HSReplay, in D1-D4, as of yesterday’s patch.
Winrates:
0202402024 https://imgur.com/gallery/CxczDaV
Popularity:
02242024 Popularity https://imgur.com/gallery/l1MvNHz
Thoughts? Impressions? Decks? Discuss.
Where’s Warlock? The nerfs supposedly did nothing.
I only have my cellphone right now, but you know what? Lemme see if I can take a decent screenshot of playrates.
I don’t think D1-D5 winrate stats matter for any discussion.
If you want to progress, those stats are meaningless and not a representation of the true power of decks
I see no sludgelock on the list, and streamers and I are still spamming it in legend
/criticism
Now for the most pressing part - how to counter plague DK? I’ve negative win rate against it with all my decks
I like how the highest played priest deck is “other priest”, lol .
What rank do you think is best used here? I can upload whatever. Lemme know. I like D4-D1 because it’s not the highest echelon of play, but a competitive environment populated by players more competitive than most others. Kind of a snapshot of mid-level players (most of us).
You’re right, probably, I was just warning that those stats shouldn’t be taken for granted as a real representation of the decks’ power
The stats for other ranks are already included in the link, as far as I’ve seen, so it’s all good, at least for popularity
Anyway, just be careful when interpreting data like this
For example, Totem shaman appears to be the 2nd strongest deck in d5-d1, but that statistic is useless because its’ pick rate is 0,78%
The less games are played, the higher the variance, so the win-rate is more skewed/polarized, and that is misleading
That’s kinda like which countries have the greatest GDP per capita - ofc the smaller the country, the higher the GDP per capita will be, but that stat is meaningless
1 Like
Lol, you just kill it. The longer you let it hang around the worse your odds of winning.
Smash them in the nose and gg.
That’s not how stats work.
The take away is people are sleeping on totem shaman, not that the deck’s power is wrong.
Edit: what’s totem shaman’s match up with plague dk, btw? Is it feasting on that deck to be strong? I don’t see the numbers, but just a thought. It’s the type of deck that should chew up the plague decks and spit them out.
Quite the opposite, actually. The variance is smaller, not higher. The more played the deck is the more variance, with variance measuring the gap between the best and worst players with the deck.
1 Like
Too early to tell. Not enough data yet.
2 Likes
that’s not what variance is xD
In statistics, variance is a measure of spread between a number and the average, in this case, between a deck winrate and the average winrate (50%), and that’s higher the LESS games you play, if everything is balanced around the average properly.
But wouldn’t the smaller quantity of games make the representation of power less accurate
?
1 Like
Variance is the deviation from the expected value, but you’re making an assumption about the expected value that isn’t supported by your data.
If you want to tell me that you think this is unreliable because it’s not representative of the deck’s power, that’s a separate question from variance and sample distributions.
Of course. This is only a place to discuss the meta as it is right now. We aren’t making definitive conclusions whatsoever. We shouldn’t be, anyway
2 Likes
it’s a statistical law, regression to the mean
The higher the sample, the closer to the mean the variable will go
The lower the sample, higher the deviation
For example, 0,78% have a winrate of 57% vs 12% have a winrate of 53%
You can’t possibly assume the 0,78% of the total population can be as representive as the 12%. It’s just too weak of a sample. Maybe only hardcore shaman players play it, and they are experts with that class. Maybe it was just a luck/variance in a couple of matches. Maybe it’s both.
But one thing is sure, if that deck got picked by 12%, the win rate would be a lot lower than 57%
That’s not what we’re talking about and that’s a simplistic understanding of the topic.
There’s nothing stopping this small sample from being completely accurate - a perfect hit on the mean.
What happens with repeated samples is you have additional confidence in your conclusions.
That’s not what’s happening here, no.
Why? If that small number is the total of the population, then it’s a perfect representation of that deck in the meta.
You’re confusing different phenoma here that don’t apply.
I mean, simply seeing more mirrors would have that effect, yes, but these numbers are a snap shot of a moment in time. None of them vary independently. Totem shaman has X win rate only in the specific moment you look at it. Change the rates of play of any other deck or all other decks, and everyone’s numbers change again. These aren’t forward looking statistics, they’re backward looking historical facts. They aren’t predictive in any real sense.
This doesn’t make sense because you’re talking about multiple different problems at once, and I don’t really want to devote more time to it.
I just grabbed the ‘meta’ page for Diamond through Legend, not significantly different from D4-D1 but some shifts. Sludgelock decline is real.
1 Like
I’ll make sure to add more data tomorrow, like legend, and top legend.
I want to, again, make clear that I am not trying to make predictions or anything. I just want discussion on how the meta is at the time of posting. 
I have encountered a few casino mages again and hate it with passion.
I wish we could delete all RNG classes from the game, starting with f***ng mage.
When they say win rate: do they have a cap of min games; because the statistical significance drops a lot below 200 games; though personally I think it’s subjective cause I respect some decks that only have 50 games because they may do something very interesting and original I was looking myself to create etc.
PS: Pro tip: if you right-click and “open as image” you can link the image URLs directly (maybe(sometimes they redirect even that)).
PPS: If you want a free source that does something similar (and maybe even better sometimes) check d0nkey.top
I really don’t like HSR stats because the winrates are always from the tracker perspective. So winrates tend to be inflated a bit because people playing on PC with tracker installed are a little better than average.
That said, while I wouldn’t pay much mind to the stated win percentage, I do think that the data can tell us something about the decks relative to each other. So like it might be the case that Rainbow DK is actually at 55% winrate and Aggro Paladin is at 53%, but it’s very unlikely that Aggro Paladin is outperforming Rainbow DK.
I couldn’t disagree more strongly. I think that’s exactly what should be considered for all balance purposes. And I do mean all.
Um, no. To both of you.
When we’re dealing with binary data — and that’s what winrate data is, assuming no ties, it’s just ones and zeros — then variance increases if we add a loss to a winrate above 50% or a win to a winrate below 50%, and variance decreases if we add a win to a winrate above 50% or a loss to a winrate below 50%.
Special exceptions: if the loss or win results in a new winrate of exactly 50%, the variance doesn’t change. And if the old winrate is exactly 50%, adding a game of data decreases variance regardless of win or loss. Finally, if it’s ALL 1s or 0s the variance is 0, even if the sample size is small; this never happens in practice but I’m just covering all the bases.
Generally speaking though, as sample size increases variance decreases. As in, if you have five games where A has a 60% winrate against B, and you add another 5 games of data with the same winrate, variance decreases. But you can’t just ASSUME that the new data will just be a carbon copy of the old data like that. So Neon is more wrong here.
I’ve covered this already earlier in this post, but on the one hand correct, and on the other hand so are you.
One the one hand, yes, the overall winrate for any deck is the sum of the products of matchup winrate against an opponent deck and the popularity of that deck, so if you change popularity of ANY deck you change overall winrate. But on the other hand, there is no alternative to this. You’re still criticizing the best method.
They’re predictive in the sense that the meta of tomorrow tends not to stray too terribly far from the meta of yesterday. It’s like, if I see where you are standing at this very moment, I have a fairly rough idea of the places you might be five seconds from now. I might be off by a few yards, but you won’t be on the other side of the planet.