This -will- be nerfed. (Hollow Hound)

Multiplayer Discussion
303

Nope, not what I said… at all.

But it’s a sample of highly skilled players. And one thing about samples is you can’t always make generalized conclusions based on your sample. The conclusions are a function of many factors, including how your data is collected. Specifically, we can’t tell if the win rate changes because of skill of players or because specific players stop playing the deck.

How many highly skilled players don’t bother grinding top 1k each season? How many players who ended up top 1k actually played hound hunter in the games recorded at D4-1 - and how does that impact your position? How many top skill players changed decks right after they passed the legend threshold to play something they find more fun?

There are lots of answers here beyond “skill” that explain this.

What is your proof?

Are you confidently asserting that preferences don’t matter at any rank? That people don’t change decks for any reason other than wins?

In a pocket meta. They make it viable in a very small sample of the spectrum, which is exactly what happens at lower levels and you easily discard - the specific mix of decks can make things look better or worse.

No, it isn’t. You’re just scrambling to justify yourself.

Again, you didn’t comprehend the math. I showed you explicitly, EXPLICITLY, that change in play rate was a bigger factor in overall win rate than the change in the match up spread. Clearly, you don’t understand maths.

What you keep failing to grasp is that this goes both ways. Small changes in the meta AMPLIFY the changes in performance MORE than the changes in performance. I literally gave you a mathematical example showing you that a FIVE PERCENT swing in win rate in a give match up is less important that double the play rate of that match up on overall win rate, which is how tiers are decided. The meta is the determining factor when looking at tiers, not the match up.

304

And yet, even with all of those “what ifs,” the individual deck matchups have significantly shifted amongst the top tier skill levels in the game. So while sure, maybe the slightly better players among them have moved to other things, the best players using hound hunter’s are getting less wins out of it in many individual matchups.

I had someone else dig up some matchup data from HSR for me.

Unholy DK goes from a 52.2% win for the hunter to a 42.7%
Big DH 54.4% to 48.1%
Outcast DH 56.5% to 42.2%
Pure Paladin 51.4% to 47.7%
Miracle Rogue 55.4% to 46.2%
Secret Rogue 62.5% to 50.7%
Naga mage 64.6 to 50.1%

That’s 5 different decks that hunter beats at diamond but loses to at the highest levels of play, and 2 that go from heavily favored to tossups.

Zero decks flip the other way for hunter.

Without matchup flips,

Blood DK
Frost DK
Burn mage
Control Priest
And undead priest get slightly better for the hunter

Relic DH
Spell DH
Thadd Druid
Spooky Mage
Thadd lock
And Enrage warrior get worse

Even in the absence of any meta change, hunter would be significantly weaker in top legend just from that.

So, hunter gets worse in 13 matchups and better in 5

Those aren’t meta forces at play.

305

I don’t know what they gave you, but that’s not accurate information.

Second, none of that is relevant in a vacuum. It only matters when you look at the meta because even the biggest change there 14% doesn’t affect win rate as much as double the play of enrage warrior does by itself.

None of your numbers mean anything without weights. None of them are relevant to hunter without knowing how much of each deck you see. None of it matters. None. I can’t explain this in crayon on the forum, but weighted averages are only relevant to weights.

Every time you reply you confirm that you completely don’t understand any of this. None.

It all depends on how much of each match up you see, my dude.

306

I wonder since this thread got so much tread, if this will help speed any decision along.

Thank you guys for all this input.

307

I don’t want a fair game I want to win by default D:

308

That’s not really true though, because you have some good matchups also increasing in popularity. You have to look at the whole picture. Yes, doubling enrage warrior hurts. But it hurts more because it’s also getting better at the matchup at higher skilled play.

You have to look at the entire picture.

Unfortunately, HSR lacks enough data on several decks to get a full comparisons in the meta (about 22% of the decks played in top legend don’t have enough games against hunter for the hunter page to show a win rate average against the deck, so the weighted score I’m calculating can’t be directly compared to the diamond only one).

https://docs.google.com/spreadsheets/d/1fTwk9410dOibKHD6eX-0FU5vfdqPYuwCpoSHaNJ_fdQ/edit?usp=sharing

So while we can see that hunter’s score is dropping a total of about 13.4 points between the diamond meta and win rates and the top legend one, We can’t identify exactly how much of that drop is from JUST the meta because the top legend score is missing some matchup data (every X in win rate) that we would need to do that calculation.

What we CAN do with this is see that among the decks that hunter does have matchup data for, if we were to apply how hunter performs against those decks in diamond to the play rates in legend, hunter would get significantly better, and would see about a 10% boost in win rate overall.

So skill is ABSOLUTELY a major factor here in why hunter is performing worse. It’s not just the frequency of decks (which I can’t directly compare due to missing data from HSR).

But here’s the thing, the difference in each individual matchup is a major factor (the biggest) in why the deck choice at top legend is different than in diamond. You would never see a major spike in outcast DH with the individual deck matchup performances it sees in diamond. The deck isn’t played well enough to win there.

Hunter has a TON of matchups that just aren’t good for it anymore in top legend that are good matchups in diamond. Those decks increasing with diamond play skills would make hunter perform better in the meta, not worse. The fact that they stop being good matchups for the hunter is why hunter now actually sees counter decks at top legend.

Only ONE of Hunter’s naturally bad matchups in diamond actually increases in play rate. The rest of them are newly bad for hunter when people play the game properly.

309

Also did someone put “Mythical Terror” on the “Problematic” list?

Haha, no. Especially with “Always a Bigger Jormonger” coming out for Hunter to combine with Hollow Hound.

Really, I’d love more synergies for Big-Demon DH.

310

This comment is the absolute pinacle of (Bovine excrement).

The idea that only 1000 people know how to “properly” play the game is complete horse apples.

That being said, let’s talk about your spread sheet and how it misstates facts.

Most decks have their win rate go down in higher brackets because higher brackets everything gets closer to 50/50.

You note a whole bunch of match ups that went down, but don’t say how much they changed. Take control mage, that hunter, according to your numbers, beats 70% of the time. If that went down 3% to 67% what do you think that means? The bigger influence is that the deck went from 1.7% of the total decks to .2%. that’s the bigger loss in wins right there than the 3% you think is so key.

You can do this for most of the decks you list as “went down” and their corresponding play rates dropping. It doesn’t mean hound hunter is worse, it means those decks aren’t there.

The spell dh match up went down too… but the play rate tanked, according to your calculations, and that’s a bigger change than the impact of the change in win rate.

But more importantly, you leave out 8 cells in the column weighting for top 1k, but the column totals between the first measure and the second are identical to three digits .938. That’s pretty amazing that you hit the exact total in both columns despite different number of cells. You know what that suggests?

Faked data. Huh.

311

This is largely because people are playing better, and the worst decks get filtered out of the meta entirely.

This doesn’t explain ANY of the flipped matchups.

Both numbers are on the sheet. You can clearly see how much they went down if you read across the line and do a subtraction.

I most certainly do not. I included every single matchup there was data for.

Uhh, I didn’t. The second breakdown totals to .887 due to a different number of decks that fall under the “other” category and don’t have enough games to register as a specific deck.

So, no weird coincidence there. (I actually hoped that the redistribution of decks at top legend led to a closer sum so I could directly compare the two metas’ effect, but there were too many hunter matchup data missing, so it suppressed the top 1k score, making them impossible to compare directly)

Yes, and then there are are no less than EIGHT new decks that hunter now loses to that it didn’t at diamond, and NONE that it beats now that it didn’t before.

That’s pretty irrefutable that the deck is just worse at top legend.

8 new bad matchups, 11/17 matchups up there got worse, which directly creates a meta that hunter is bad in.

Shifting the play rates around in the same way in diamond doesn’t create an unfavorable environment for hunter. It only has 4 negative matchups there. The number of bad matchups for hunter nearly TRIPLES at high legend, and those are the same decks in both brackets.

So yeah, that’s pretty irrefutable, and you can verify the data. It’s all available on HSR (with minor tweaks that have shifted since I pulled it earlier today). It’ll always paint the same picture.

But go ahead, continue to be bad at the game and blame the meta for your struggles against hound hunter and not the real issue: your skill at the game.

And yes, I’m aware that the data shows hound hunter as fairly overwhelmingly powerful in diamond, to a degree that has triggered nerfs before.

That doesn’t change that the deck is significantly worse at top levels of play (and it’s still good there), and it’s not just the deck distribution doing that. Skill differences are a HUGE part of it.

312

This card needs to be destroyed. Decks for Hallow Hound are popping up. This is just shameless advertising for diablo. Seriously blizzard is pandering destroying entire games to sell other games too. Next up, can’t win at pvp in diablo because the wow mounts are too strong.

313

Hollow Hound, my guy, and these are creatures from Shadowlands. I get the sentiment that the decks for it suck but what you’re spitting is a little curve-ball-ish.

1 Like
314

Can I take my victory lap now then?

https://docs.google.com/spreadsheets/d/1ISj7J02c3QBs4CIFMLmrL810t3fY1Fo8A-5u7OE01v4/edit?usp=drivesdk
I recommend going to the Index tab first if you want to get into the “how.” If all you care about is results, straight to the Summary tab.

The conclusion is: as an overall meta average, the shift due to changing matchup winrates is about triple the shift due to changing meta. About 75/25.

Given some things I could quote from Smeet but haven’t, they may have exaggerated the extent to which skill trumps arbitrary meta shifts. I mean, it does trump it, but meta shifts still have a significant effect at 25% relevancy. “Only”/“never”/“always” statements are rarely true.

And it’s worth noting the 75/25 thing is a global average, and some decks are outliers. In particular, Neon was onto something regarding DH because the increased winrate that Outcast and Relic DH see at top Legend (relative to D4-1) is a whopping 83-89% due to meta shifts rather than matchup winrate shifts. Relic DH’s significant popularity at high Legend basically represents “the 25%;” it’s not the whole club, but it’s club president. In every meta there are going to be some decks like this; there are always going to be opportunities created by shifting matchup winrates that someone is going to be in a position to exploit.

It’s not “basic math,” it’s an assumption you made about data while you just happened to be in possession of the formula.

And it’s also vital to understand that when Vicious Syndicate publishes a table of Diamond 4 and up matchup winrates, those are NOT Diamond 4-1 matchup winrates. They do give the number of games each matchup winrate is based on, so you have to go and calculate (wD*gD-wL-gL)/(gD-gL) where wD is D4+ winrate and gL is Legend games played. The actual matchup winrate for Hound Hunter vs Enrage Warrior at D4-1 is not 45.3%, it’s 46.0% — in terms of 50% that’s -4%. That matchup winrate at top Legend is 42.0%, or -8%. The matchup literally becomes twice as bad going from one rank to the other. So even with this bad matchup gaining almost-but-not-quite triple popularity going from D4-1 to T1KL, forty-something percent of the overall performance impact to Hound Hunter from Enrage Warrior is due to matchup winrates and not a shifting meta. Barely more than half is meta shift in this one narrow instance.

By the way, overall, the change in performance for Hound Hunter is over 80% due to matchup winrate changes vs meta shifts, once you also factor in all the decks that aren’t Enrage Warrior. Hopefully the debate at the heart of this derail is now settled and I can end on this fact about Hound Hunter as a light rerail.

315

Yeah, in general my whole argument is that playing decks differently at high skill creates different matchup spreads that yield a different proportion of decks. I largely have been rejecting that the different meta is the main cause of hunter’s performance shift.

There are definitely cases where individual decks are only viable in that specific meta, but that requires it to have a matchup spread that’s specifically favored to those decks alone. This is fairly common for priest decks that appear strong at top legend but fail in the larger meta, as that deck in the past was often built as a specific counter to that meta.

That’s not what’s happening to hunter here.

Some decks are better at top legend because of skill differences, some due to meta shifts. They always are working together though, and those meta shifts largely don’t happen without some decks having notable shifts in win rate due to skill differences (even if it isn’t that particular deck, another deck dropping off can make room for it).

It’s a complex system.

Thanks for the deeper dive!

316

In this meta, Control Priest is the polar opposite. Its winrate gain at T1KL is less than 4% due to popularity shifts, and almost entirely from an improved matchup winrate table. T1KL Control Priest players are smart. Weird, but smart.

Correct. It truly is a case of top Legend players just play against the deck better than Diamond players do. Or more precisely, that’s about five sixths of it, with a different meta being the other sixth.

317

Yeah, not referencing this specific meta for control priest. There have been times where it’s been recognized only good in top 1k due to specific decks up there.

Right now it’s more well rounded for wider ladder use.

Yeah, that’s basically what I’ve been trying to get across, just struggled to quantify the meta influence.

1 Like
318

I don’t read, scrtotiums posts, but I can assure their math is 100% made up and wrong. They likely perfomed some extra calculation on the data which only they think they know about to make it work for their purpose.

Again, I am talking about a specific deck, hound hunter, and the change in hunter performance in your pocket meta is more about your pocket meta than anything about hound hunter.

Top 1k should never be used for game wide balance because it’s fake. It’s not typical or real due to the preferences of those players to play specific types of decks into the ground the second they are remotely viable.

Anyone who wants to pour over historical data will look at decks like miracle rogue and control priest and see how much more often these decks are overplayed relative to power in this top 1k festival. It’s about what they like more than anything else and 1k people should not define balance for the other half million players. That’s completely stupid.

It isn’t though. It’s that they play against all decks better and win rate in general approaches 50% for all decks as they regress normally towards that level as competition becomes more equal.

I can’t see how you calculate influence, but I’m certain you did it wrong based on how much you’ve shown you don’t know about statistics in the past. This is the reason I ignored you, you’re a pseudo intellectual with no formal training who thinks they’re the smartest person in the room.

The reality is the maginitue of the effects vary like length and width. When the playrate triples, the loss rate would have to change by the same order of magnitude in absolute terms to have the same effect size.

So if It is 3x the rate of play, the it would have to be 1/3 of the rate of wins to have the same effects. So if you win 55% and it’s 5%, the change to 15% would be the same effect on the win rate as a drop to 18.5 is win percentage to have the same absolute effect on the weighted average.

This level of change isn’t the case. The play rates are most of the changes for hound hunter - both the increase in play of bad matches and the absense of best match ups.

319

Methodology is explained in some detail in the Index tab. You (and anyone else) is free to check my work. To be totally honest, since I was copying over 2x winrate and popularity tables from VS by hand with 25 decks each, or about 1000 data points, I probably have a couple of hopefully minor typos in there somewhere; I’d give the whole thing a ±2% margin of error.

To explain the core concept in a little more detail, the goal is to take the processes of individual deck performance due to shifting opponent deck popularity, and that performance due to matchup winrate table changes, which normally occur simultaneously, and isolate them.

We can do this by applying matchup winrate tables for one meta to deck popularity data for a different meta. The real D4-1 winrates are found by multiplying the D4-1 matchup table (if, theoretically, we had perfect data for it, which we don’t, but we have close) × the D4-1 deck popularities. Same concept with Top 1000 Legend. But if we change just one of these at a time, say T1KL matchup winrate table × D4-1 deck popularities, then we isolate the effect of meta by contrasting it with the real T1KL winrates (if D×L→L×L then only the first term is changing), and we isolate the effect of skill by contrasting it to the real D4-1 winrates.

Here there is a methodological question which comes up, analogous to getting to the other side of a mountain on foot. We can’t go through the mountain, because if the two shifts happen simultaneously then we can’t isolate each factor. Each isolation is like a rest stop halfway through our trip around the mountain. But we can choose one factor to isolate first, then apply the second, so the choice of which factor to isolate first is like the choice to go clockwise or counterclockwise around the mountain.
start →→→ rest
↓↓↓ mountain ↓↓↓
rest →→→ end

So the question is: clockwise or counterclockwise? To which the answer is: whichever has more data. This is unambiguously the route of D4-1 matchup winrate table × T1KL popularity. It takes a LOT more data to make a reliable matchup winrate table than it takes to reliably measure deck popularity. Matchup winrate data at T1KL is extremely sparse even from VS. So we go that route and not the other. In a perfect world we’d do both, but frankly the data for the other route is so sparse as to be utterly inconclusive so why bother?

However, even the “good data route” isn’t the perfect data route. I kept track of how complete the matchup winrate table data was, and VS had about 82% of the data necessary to get a flawless simulation of D4-1 skill × T1KL archetype popularity. (No individual deck archetype, no matter how popular, had more than 92% of data; about 8% of the meta at T1KL can be classified with the label “assorted jank.” Because Kibler and his barns exist.) It’s simply impossible to get more data than that, so I did the best I (or pretty much anyone else) could with what I had.

1 Like
320

No, very specifically the matchups are getting worse. Applying diamond’s win rate stats to the decks you see in high legend spikes the deck’s power by over a tier.

That’s just patently false. If the deck was truly overpowered, you wouldn’t be able to do that.

Diamond can’t do that to hound hunter, there aren’t any matchups that are truly bad for the deck there. They aren’t being played well enough to exploit hound hunter’s weaknesses.

The trend downward starts in diamond and smoothly looks worse for hound hunter until top legend, where it’s still reasonably strong, just obviously not a major outlier the way it is in lower skill brackets.

It’s fine to have the stance that the deck is too strong in lower brackets and needs a nerf because it is a bad experience for a majority of players.

It’s not really fine to just call the drop off in performance a lie, made up, or a result of forces that are clearly not the sole or even major reason the deck is significantly worse at top skill levels.

Yes, aggregate win rates tend to, but that doesn’t hide the fact that decks have inherent advantages. Those are often highlighted more at top legend.

For example, enrage warrior becomes further from a 50:50 matchup in favor of the warrior, priest becomes more favorable for the hunter.

The meta is a web of individual matchups. The vast majority of those got worse for hunter, with several going from favorable to unfavorable or a coin flip for the hunter.

No matter what you do to the deck distribution, that’s going to hugely cut into hunter’s performance. It isn’t just one deck doing this, it’s nearly all of them.

Also, the system is designed to separate players into ranks with roughly equal skill, so that 50:50 aggregate win rate is largely something you’ll see pressure towards across the board, it’s not unique to top legend.

2 Likes
321

Hmm. I wouldn’t say “never,” (didn’t I just tell Smeet I don’t like “never” statements?) but I would agree with the more moderate stance of prioritizing the D4-1 meta generally.

This is just pure raw conspiracy theory.

I actually do think that Top Legend players metagame (as a verb) much harder than other ranks. But “harder than other ranks” is not to be confused with “harder than the effect of skill on winrate changes.” Like, professional basketball players sure can jump, but none of them jump hard enough to overcome the gravity of the planet and enter orbit. You’re not comparing how hard they metagame to the right force here. You’re comparing them to normal people as if that’s what they’re fighting against. Nope, all those people are left behind in lower ranks and long forgotten. LeBron James isn’t facing off on the court against Linda from Accounting. What you should be comparing their jumping power against is a force of nature.

It’s easy to underestimate the power of skill if you’re exclusively a Ranked player. Matchmaking there is, believe it or not, designed to match you against equally skilled opponents. It does such a good job of this that large skill differences simply will not be experienced within the mode (small but significant will be seen occasionally). I was mostly an Arena player where at 0-0 you get matched against 0-0s of all skill levels, so it’s been a lot less invisible to me. I mean, I’d be exaggerating if I said I was the LeBron of Arena, but I was darn good, and my opponents were often like Linda from Accounting. Seal clubbing.

1 Like
322

I think maybe you’re misunderstanding what I did in the spreadsheet. As I explained in the previous post, the “route” is:
matchup winrate table D → D → L
deck popularity D → L → L

(By D I mean D4-1 and by L I mean T1KL.)
So the first move measures meta shifts in isolation, because only deck popularity is changing. And the second move measures the effect of skill in isolation, because only matchup winrate table is changing.

So if you mean D×D→D×L, using the"route" I took in the spreadsheet, that means you’re implying that meta shifts improved the deck by over a tier. Unless you’re looking at it “backwards”…
matchup winrate table L → D → D
deck popularity L → L → D

…in which case nevermind. Start position is arbitrary, you can go L×L→D×D or D×D→L×L, doesn’t really matter. I’m just saying which way I went in the spreadsheet.

Edit: had a lot of crap backwards. Should be good now. My bad.