If you think matchmaking is rigged, read this

Multiplayer Discussion
41

As for the OG poster of the thread, you ever see the data for classes vs classes? It’s a better outlook than deck composition. I tell you this from experience I had over two years ago.

I played a ton of Renolock, an it was basically the only deck in the game I would play. So just nonstop Warlock.

Eventually Playing as that deck an class I saw ONLY the META of wild. The BEST decks, there was nothing ever anything else when I queued as Warlock in Wild. Didn’t matter ranked or casual mode.

The minute I swapped to another class I lost the Warlocks matchups in both ranked an casual. They would only need roughly 3-4 games before that class worked it’s way in the meta.

This was actually quite useful if you wanted to dodge a deck in ranked. However the game play at Diamond 2+ Had over a minute wait time if you tried this, before the game would try to just give you someone in queue. (example Secret Mage) a meta deck.

So it would be most telling if we had more information on class queuing’s. Because that is a stim of the game is rigged for people. Why does X class have a different meta than when you switch for a few games with something else.

Look how many times you see people mention exactly that. Even I can vouch for it from experience. Someone in this thread did there best to explain at certain rank points you would see more of the meta. Which is true as you would expect from a developed meta game. But it acts up when you decide to stop using X class.

So once again the story continues. Because I think they’re might actually be a confirmed statement about Classes not likely to see each other as a part of the Matchmaker system. I would have to look around for this as I think this was announced in a Developemnt blog or something of that nature.

42

That’s the thing though. It doesn’t work that way. We like to try to notice a pattern there, but there isn’t one.

The pool the game pulls from for your opponents doesn’t check what you queued with. You may notice small streaks that makes it look like it did, but it’s not your action that caused the result.

It’s what everyone else decided to hit “play” with at the moment you did. Over long stretches of queueing, you’ll see your matchup spread with every deck start to trend toward the standard meta of that bracket that you see in the weekly VS reports/HSR data.

There’s no system in place to adjust your matchups based on your deck selection. If there were, HSR would very quickly pick up that different decks see a different meta.

1 Like
43

This discussion between Sc0t and Schyla is even more interesting and entertaining when you imagine that it’s Millhouse and Elise talking about this in a college classroom.

3 Likes
44

I literally turn it off and don’t return for hours or days the first time this happens. Literally hours or days. Even mid match. The first time it makes me angry or annoyed I bolt because I don’t waste my play time to feel crappy. It doesn’t make be buy anything ever. It breaks my interest completely. Idk what kind of moron spends money to feel crappy.

Then what, exactly, is your actual point?

Making the UI pleasing and stimulating is good design.

100% no.

There are so many common myths that people swear are true based on anecdotal observations and coincidences.

Absent actual, falsifiable, evidence it is impossible to assert that the match is based on anything other than what we have been told, which does not include deck list.

As an aside, I have never noted the phenomenon of decks being matched to different decks, but I also tend to play a list for an extended session rather than trying to game the meta. By that I mean I pick something with a good spread versus the most common decks and play my games.

1 Like
45

Under what conditions are you expecting to see concurrent random calls within a game?

46

This is a very nice simulation and it’s worth looking at, but I think it falls short of accomplishing its goal.

The people who believe that the matchmaker is rigged are wrong, but they’re not all irrational or stupid. They’re basing their conclusion on two things: their experience and a specious intuition about probability.

They’ve experienced situations where their win streaks have been followed by match after match of hard counters. Their intuition regarding probability tells them that at any fixed rank, the probability of facing a certain deck this match is independent of the probability that you faced it in the previous match. Thus, the counter deck streaks seem very improbable.

Their assumption regarding the independence of each individual match is somewhat true, even if the conclusion they’re drawing from it is not.

The data you provided is not going to convince them because it doesn’t explain the fundamental difference between their observed reality and their understanding of how probability works. The data replaces paradox with sorcery. It’s giving them nonintuitive results without explaining why their intuition is wrong. This rarely convinces people.

What’s more, the narrative you’re using to describe it might add to the confusion. It’s not the case that the way the system is built gives a player more counter decks after a win streak. Any given individual’s matches are still random events that are not quite independent (because their rank changes and deck distribution changes per rank), but not not very correlated either (at least not when the matches are in close proximity).

The heart of the matter that needs to be explained first is the difference between between the play session of one individual player and the experience of all players in aggregate.

The system is built to sort players by skill level. At the end of the month, each player should end up at about their correct skill rating. The story of how that happened differs from player to player, but because there are so many players then some common stories will appear many times. The process of finding your skill rating involves ranking up, but also ranking down. There are only so many ways that ranking down happens, and a common one is to face counter decks. The game isn’t giving you counter decks on purpose, it’s just that when you aggregate the data you’ll see a lot of cases where that happens.

It’s fairly simple to give examples of how this sort of thing happens if people are interested.

1 Like
47

The kind of morons who think that an entertainment company is lying about cheating its players out of wins so that they feel bad.

3 Likes
48

They are only “nonintuitive” results if one clings to the notion that matchmaking is random.

“Deck” choice (if we consider paper rock and scissors to be “decks”) has an effect on winrate — different decks have different favorable, unfavorable or even matchups. Matchmaking is by winrate. Therefore deck choice is not matchmaking neutral. It’s just that the algorithm doesn’t need to, and in fact does not, “deck scan” to have this effect.

Perhaps the best way to demystify this would be an analogy. Let’s say that a particular employer earnestly set out to make it’s hiring process about choosing the most educated people for the positions available, and achieved this, truly hiring by that standard and that standard alone. Would this mean that all ethnic groups have an equal chance of employment? Most likely not. In order for all ethnic groups to have an equal chance of employment, all ethnic groups would need to have identical levels of educational achievement, which simply is not the case in real life. There is inequality between ethnic groups in regards to educational achievement.

This analogy is not an employer hiring randomly — that’s actually a pretty despicable practice if you think about it, imagine a hiring manager choosing between applicants by rolling a die. There’s nothing random about the hiring decisions here. But I don’t think it’s difficult for people to understand how a hiring standard that has absolutely no overtly racist component can nevertheless result in a racial disparity when implemented. In the same way, a matchmaking algorithm that has absolutely zero deck scanning can nevertheless create deck-archetype disparities, because it selects according to a standard — winrate — that is not equal among various deck archetypes.

2 Likes
49

Interesting experiment. I’d be interested to see what happens if you factor in a non binary winrate (say 60%)

50

As polarization decreases, the rate at which archetypes bunched together decreases. This is of course, assuming that players don’t change archetypes , which is generally a pretty fair assumption given that most players don’t have the budget to create multiple archetypes. As long as matchups are not perfectly 50 50, the sorting effect will occur, it just might be extremely slow

It is worth noting, but because Hearthstone resets ranks monthly, this effect would be most noticeable at the end of the month and least prevalent in the first few days of the month. Not including win streak star games

51

I meant I’d be interested in seeing the charts :slightly_smiling_face:

52

I don’t quite agree that this is the sole cause of what we’re seeing here. You’d get streaks of counter decks even if deck win rates were constant throughout the ranks. It’s obviously a factor, but not the only one (and this is also reflected in your simulation).

53

Well, I just don’t have as much free time as I used to. Maybe I’ll try making a more complicated sim later this month. I’m switching apartments, and the new one is currently very vacant, so I’m going to have that day where I sit around all day waiting for all the utilities to get hooked up. Mandatory day off.

The two main factors here are an algorithm which is sorting archetypes into like clusters over time, and a random starting position. But it’s important to understand that the random starting positions are random because I decided to have them start that way. The randomness in the results is there because I put it there, and the nonrandomness is there because that’s what the matchmaking algorithm does. It cleans up randomness over time. It transforms chaos into order.

I’ve consistently stated in this thread, multiple times, that a standard matchmaking algorithm, that matches by MMR, isn’t random. You and Schyla have seemed very reluctant to accept this.

:clap: MATCHMAKING :clap: IS :clap: NOT :clap: RANDOM.

I am really starting to understand why the “is matchmaking is rigged or not?” debate never gets resolved. It’s because both sides are wrong. No, matchmaking isn’t rigged; no, matchmaking isn’t random. It’s a conspiracy theory fighting a strawman.

1 Like
54

This reminds me of math and chemistry. They commonly get overly simplified explanations from the laymen, but then you talk to a chemist or mathematician and you realize the process was way more complicated than it appears on the surface. Then you realize your previous conclusions were completely nonsensical.

I will admit, when I played a lot of ranked HS in the past, it certainly felt rigged. Lol

1 Like
55

Pretty much always since all Hearthstone games will run concurrently on any given server.

56

Spot on.

That’s what my previous post was so frustrated about. I mean, the game isn’t rigged, but the “explanations for the laymen” are oversimplified to the point of falsehood. Like they’re chosen less for truly educational purposes and more for propaganda purposes. “B-b-ut They™ think [insert ridiculous belief here].” Okay, so what if they do? How is replacing one falsehood with another an improvement?

We’ve been doing the laymen a disservice. Well, I don’t know, maybe we don’t owe them anything. But regardless of whether they’re owed the truth or not, I feel we can and should do better.

57

This is all assuming they are even running into a string of hard counters after their win streaks. It’s unlikely they are running into their single worst matchup 5 or 6 times in a row. Most decks have lots of bad matchups too, is every unfavored matchup being grouped together as a counter?

If it’s against anything except their worst matchup, why would the algorithm go easy on them when it’s “supposed” to be forcing a loss?

Why did the algorithm wait until the 7th win instead of the 6th, or the 5th or the 12th???

Why are there tons of threads of people complaining of losing constantly to their counters (losing streak), only to switch decks and keep losing? Where are their free wins to stay at 50%?

There are also tons of threads about the actual games being rigged. Mulligans, draws, RNG outcomes, disconnects…These are also peoples experiences from which they have drawn their own conclusions. So why on earth would Blizzard bother to design a MM algorithm to scan decks and force bad matchups while also manipulating the game itself to dictate a winner?

If your answer to any single one of these questions is “They don’t rig every single game”, then you have no way of knowing which, if any of the games you have ever played was actually rigged.

And even if they are rigging some of the games in some of the ways mentioned on this forum, there is zero chance it’s being rigged in exactly the same way as every single person has complained about (if only because they frequently contradict). Some people would 100% be wrong.

58

Except the above has been known for years and Is not the problem.

Under the old rank system where their was less fluctuation in the second half of the month.

Yes it was quite common to have a bunch of say druid. then whatever could easily beat druid say rogue, then whatever could easily beat rogue. etc etc.

When people are talking about rigging what they are talking about is say you know x deck is 20% of the player base. but you can play for days and not see it/barely see it.

your both winning and losing games so its not a bad thing. Then immediately on swapping decks you see multiple x decks

59

It might be more likely than you’d think.

After running the simulation for 64 rounds, archetype “clumps” tended to be about 3 ranks thick. As the number of rounds in the simulation increases, both the “purity” (a rank being closer to 100% a single archetype) AND the clump thickness increase. If I gave this simulation more rounds, I could get it to a clump thickness of 5-6.

So basically, the later it is in a month (that is, the more time since ranks were last reset), the more believable such claims become. I don’t know if, in real life where bad matchups aren’t 0-100 levels of polarized, if 28 days is enough time for clumping to get to 5-6 thickness in Ranked; it might be something that would take multiple months and thus is essentially prevented by resets. But I wouldn’t just assume this to be the case.

Maybe they’re not always full of it.

Assuming a paper rock scissors metagame, a ranked matchmaking algorithm will create clumps of rock then paper then scissors then rock again as increasing in rank. Keep in mind that one rank can be partially one clump and partially the next clump.

So let’s say you’re playing Rock. You quickly climb through Scissors, then climb through Rock mirrors, then you hit the Paper wall. Maybe you win occasionally, but the point is you keep beating your head against that wall. Then you decide screw Paper, you’re switching to Scissors. The problem is: you just lost your last round, lowering your rank from the Paper clump back to the Rock clump. Oops.

Then if you’re stubborn enough and stick with Scissors, you might just moonwalk your way through the entire Rock clump.

Well they very clearly don’t. We have tens of millions of games of data on sites like HSReplay. That data conforms with what we’d expect from no rigging. So the only two possibilities are: there is zero rigging, or there is rigging on such a miniscule, highly targeted scale that the evidence is lost in the vast sea of data and discarded as variance.

Like for real it’s impossible to think it’s rigged for everyone, or for most, and understand basic statistics concepts at the same time.

In an example above this, in this post, I explain that too. If you’re playing Rock and you’re stuck between the Rock clump behind you and the Paper clump in front of you, you’re not going to see much Scissors (your good matchup) even when it’s about a third of the meta.

×××××××××××

Matchmaking only by rating + paper rock scissors metagame PREDICTS running into wall of multiple counters in a row

Matchmaking only by rating + paper rock scissors metagame PREDICTS running into multiple new counters if you switch archetypes after running into wall of multiple counters in a row

Matchmaking only by rating + paper rock scissors metagame PREDICTS not seeing certain popular archetypes at all within a couple of ranks

If you’re a “rigging skeptic” you shouldn’t doubt other players who describe these experiences. And if you’re a “rigging believer” you should understand that no rigging is necessary to create these experiences.

60

I think we are using different definitions of random.

Let me explain how I’m using the word (broadly speaking and oversimplifying a bit).

Each player is assigned an MMR. When they win, it goes up. When they lose, it goes down.

When a player joins the queue, the matchmaker finds a pool of potential opponents who have a similar MMR. It then selects an opponent out of this pool. This is the selection that I claim is random. The algorithm is trying to match the player with someone within a certain skill band, but exactly which person in that skill band is chosen via a random process.

Under these circumstances, you would expect the probability of being matched against a specific archetype in any given game to be close to the play rate of that archetype in your skill bracket. This is what I believe is confusing the “game is rigged” people. This local information about individual matches still invariably leads to the global behavior you’re observing in your simulations, even if it’s not intuitive for everyone.

To my understanding, when you say that matchmaking is not random (all in capital letters with clapping hands), you mean something else by the word random. Possibly that any person can be matched against any other person with equal probability at any point. It’s certainly not random in that sense, but that’s a rather restrictive definition of randomness that doesn’t address the problems with the view point you’re trying to argue against.