I really, really not trying to make this a tin foil hat thread, and I am not really salty about this either, because the results weren't horribly bad, but ...
I just cannot see how some recent events can just keep getting written off as coincidence. I really need to start a spreadsheet on this so I can gather more accurate data, but it is prevalent enough that it has become predictable.
Just using one instance, last evenings play. I was playing a Pirate Warrior deck ... fast aggro looking to hopefully climb to a new rank floor on the ladder. It's a homebrew deck, but sticks pretty close to "typical Pirate Warrior" (Honestly, how creative can you really get with that archetype anyway) ... but thats not important for the topic.
I played 9 games, I believe, and of those 9, all but 1 were other fast aggro / midrange decks ... several Hunters running the hybrid Spell / Minion setup, a non-evolve Shaman, and a couple Paladin's. No problem ... the deck can hold its own, but they were essentially just mirror matches in a way. I just assumed that was the micro-meta I was in at the time, so I decided to switch decks to Control Warlock.
After switching decks, next 6 matches saw Warlock, Mage, Priest ... not a single aggro deck to be found. In fact. the first match was a near exact mirror match with another Control Warlock ... 32 minutes ending in fatigue, and losing by 1 HP.
And it is not just last night ... I can almost count on when changing decks, that the classes I face will change as well.
Here is the big thing .. I don't mind it if the game does match specific decks with other specific decks (I should say archetypes, to be more accurate), I am just ticked off because what seems to be such an obvious thing is either denied by the developers, or just not confirmed. (I honestly don't know if they have ever outright denied it, or just haven't ever confirmed it).
If we knew this information to be true, we could take action by making the use of tech cards, and not feel as if using them was a waste.
Rollback Post to RevisionRollBack
I wanna glide down, over Mulholland I wanna write her, name in the sky I wanna free fall, out into nothin' Gonna leave this, world for awhile
I can add another story to your bucket ... i played a Cthun deck again recently (like after several months) and first two matches i had were vs another Cthun decks ... i havent played vs Cthun decks for like half an year :D
But rly, i dont believe there is any secret matchmaking system, that puts certain types of decks together. You would need more data than u can collect on your own, to even try and see a pattern. Couple hundred games u can do yourself is still way too small sample size for u to make any kind of assumptions so ... yeah.
I'm one of the people who has argued pretty extensively here that Blizzard is almost certainly not lying about how their matchmaking works. That doesn't mean you're wrong though.
The last time one of the developers talked in detail about how matchmaking works in Hearthstone was January of 2017. Here are a couple references where they state that only star rating is used in Ranked mode.
This does suggest that experimentation or tweaks to the matchmaking system may happen over time.
So, a couple thoughts. First, for a game like Hearthstone, a year is a very long time in their development cycle. There's almost certainly been a lot of development, discussion, experimentation, and maybe changed minds about certain things since January, eleven months ago.
Also, here's something I heard while attending Blizzcon 2017 in one of the engineering panels that's not conclusive, but it is very interesting. The transcript is linked with my comments below.
One of the server-side battle.net developers said this:
"Derek: From a mathematical standpoint, what I really enjoy working with our BI [Business Intelligence] team with is on the matchmaking for Hearthstone, and that’s where I think the rubber meets the road; because they actually look at all the numbers, they do the analysis, and then they say: “Okay, in order to adjust our matchmaking so that we have this nice perfectly formed curve that we want these are the numbers that you need to put into the system.”
So, as a software engineer you understand how to craft code, but the more higher mathematical understanding of what’s going on in the meta, it’s very important that we have them to help us do our jobs."
When I heard the bolded comment, my first reaction was that sometime in the last year they may have made changes to matchmaking to try to drive various deck choices as close as possible to 50%. The result of this might well be that selecting a relatively more powerful deck makes it disproportionately likely to be matched against decks against which one is weak. Selecting a weak deck might result in being matched against opponents against which one is strong.
Since the comment was oblique and was in response to a question about something else (the developers' use of business intelligence data) it's not clear how such a system might work exactly. It's also not evident that the system is stable. (For example, there may be multiple methodologies in place being A/B tested against each other, or the methodology might be changed or tweaked over time based on collected data, so that your experience one day may not match someone else's experience a week later.)
Confirmation bias is definitely a thing, and anecdotes like "oh I got matched against twelve pirate decks when I switched to murloc paladin" don't prove anything, since there is a lot of randomness in the system. Without extensively collecting and analyzing data, or without further Blizzard comment, I'm not sure we'll be able to get a clear idea of what the matchmaking is actually doing.
Note that there are potentially solid gameplay and business reasons for implementing such a system. The gameplay justification is that matching decks against each other more evenly helps distinguish quality of play within a match from deck choices in the ranking system, and since information on the internet is such that good netdecking can't be distinguished from good deckbuilding, ranking may as well be more about strong execution once the game starts.
The business justification would be that beginning and f2p players can't always choose their first viable deck, so bringing them closer to even match-ups in terms of archetype is likely to improve their win rate and therefore their retention.
Edit: Note that it's also possible that matchmaking is weighted by class rather than by deck. If this were true (and it would be much simpler to implement) then a class with two or more widely differing strong decks, like druid with aggro-token and Jade recently, might be able to strongly outperform just because class-based matchmaking can't distinguish between those two decks.
I see no single reason why matchmaking should not be rigged, esp. from business perspective.
I agree with you the way that the term "rigged" is thrown around here, although I would also say that automated matchmaking is so artificial that it almost doesn't make sense to call any particular choice "rigging" as long as it's trying to achieve a 50/50 win/loss rate.
That said, there IS a business reason for Blizzard not to say "we're not doing X" and then go do X, which is that there's business value in community goodwill that derives from speaking truthfully. However, as I've pointed out, their developers haven't commented on this much in the last year, and the situation could well be completely different now.
Failure to come out with an updated statement isn't the same as suggesting they lied when they spoke back in January, though some people have accused them of that. I believe that their comments were at least accurate at the time, because it's easier to refuse to comment than it is to deal with cleaning up something revealed to be knowingly false.
I played Jade Shaman almost exclusively since the start of the expansion, and I experienced the same change in what decks people play. At the start people played a lot of Hunter, then I saw the meta switch to control with Priest/Warlock. What you're experiencing is in some ways a coincidence, but it's also just how the meta flows, especially this early in the expansion. People will try something new and cool like Spell Hunter, people will see them doing that and copy their deck, then they end up getting stomped by some other deck so they switch to that one, and so on.
If you switch decks to deal with the metagame, why wouldn't everyone else be doing the exact same thing?
Also, keep in mind how random distribution works. Even in a perfectly random system, there will always be a few players experiencing rare events that may seem arranged in some way, but in reality they're just outliers. For example you could imagine every person on Earth tossing 5 dice three times. Most of those people will just get random arrangements of numbers, but a few will get lucky and roll a 6 on every dice one, two or even all three times. Now of course the people who rolled random numbers aren't going to go to forums and write about their experiences, but the people who got perfect rolls may very well do that because they had a rare event happen to them. That is why conspiracy threads like this one become so common and seem plausible, because out of all of the millions of players, only those who experience these rare events end up telling others about it.
I agree very much that it is essential to disregard personal and anecdotal experience in evaluating this, for all the reasons you describe. Unless someone like Vicious Syndicate or hsreplay is going to do a mass data analysis to identify whether something unusual is going on, we're stuck with parsing Blizzard comments and not really knowing unless they decide to share.
I see no single reason why matchmaking should not be rigged, esp. from business perspective.
I don't even see it as "rigged" .. ."if" it exists, I don't even disagree that it should. It makes sense to try to match one archetype against another, for many reasons. If the local meta is heavily populated with slow decks, I can just run a fast deck and I most likely will win an abnormally high number of games, just because I chose "deck x".
It also allows the player to not feel as much pressure to play an archetype they may not want to play ... if the local meta is primarily fast decks, and I want to play that C'thun deck I haven't played in a while, it would make sense for the system to match me with slower decks, where I might stand a chance.
To me, it just seems too obvious. Coincidence can only be assumed so many times. Much like what Sinti described, I have swapped to some very rarely played decks as well, only to have similar decks come up in the first couple games.
Rollback Post to RevisionRollBack
I wanna glide down, over Mulholland I wanna write her, name in the sky I wanna free fall, out into nothin' Gonna leave this, world for awhile
In this thread from a couple weeks ago, I posted a quote from Blizzcon that seemed to suggest that some type of meta-based matching (either by class or by deck contents) had been adopted. Here's a little data that suggests there might be more to see:
I've been playing a dragon priest deck a fair amount recently, taking me from rank 5 to 3 (and in the losing streak represented in this data set, back again.) Here's the distribution of classes I've faced:
Here's the latest Vicious Syndicate meta report win rate chart for the three major Priest archetypes. (The four darkest red boxes on the Dragon Priest row, representing low win rates, are aggro paladin, murloc paladin, Kingsbane rogue, and pirate warrior) :
Finally, here's the overall class distribution in ranks 5-1 from the most recent VS meta report:
Note that dragon priest is relatively strong against everything except aggro paladin, murloc paladin, Kingsbane rogue, and pirate warrior. Meanwhile, my number of matches against paladin on a dragon priest deck (eight out of 20) far exceeds the two or three game expectation value for a 20 game series.
This suggests that matches are selected in such a way as to push the result toward 50%. It's not absolute proof, of course. A completely random series of matches could look the same. However, it certainly seems to warrant a closer look.
Here are a couple thoughts on what the implications of such a choice would be on Hearthstone from a business standpoint, and for us as players, which might shed some light on why it might make sense to choose to do something like this:
Players like winning more than they like losing, and losing too much is probably highly correlated with disengaging from the game. This is why they've stated that they've biased matchmaking for new players toward other new players, for example. So, from a business standpoint, attempting to account for choice of class/deck somehow in trying to get each individual player as close as possible to a 50% win rate makes sense. The most players will stay engaged if the winning is spread around as evenly as possible.
Another possible argument for it is that it helps to factor out the coarse decision of what class/deck to play from measurement of skill. By adjusting rates of matching against other decks to reduce overall dominance of one deck/class choice, the remaining variation is more likely to be driven by quality of play, which is what the ranking system (particularly at ranks 5+) is meant to measure.
One potential downside to such an approach is that if (as implied by the Blizzcon quote above) matchmaking is driven by business intelligence data, it means that changes that might drive the last business intelligence data drop out of sync with the current game (such as releasing an expansion, buffs/nerfs, or major new unexpected deck-building insights by players) could allow exploitable class/deck choice advantages for a window of time until the data is again updated.
I should note that I don't consider using such matchmaking strategies "rigging," to the extent that that term connotes unfairness. There are valid fairness reasons to want to minimize the impact of deck choice on measuring player skill, such as that the availability of aggregate data to players like I've posted here makes choosing a strong deck archetype pretty trivial.
If the game (ladder mode) is supposed to be competitive, you should be matched against someone of your same rank at random, period. Anything else is completely unacceptable from a competitive standpoint. Not all players will have a win rate above 50% - it's impossible since every game produces a winner and a loser (or 2 losers).
Any 'MMR' beyond straight ladder rank is tantamount to giving someone a win they don't deserve at the expense of someone who earned it. Sure, sometimes you get a favored match, but it should be random, not manipulated to aid a player on a loss streak.
Rollback Post to RevisionRollBack
Free to try and find a game, dealing cards for sorrow, cards for pain.
Please, if you have a moment, take a look at my follow-up post. Sounds like you disagree but I wonder if you'd like to offer any thoughts about why my arguments don't hold water for you.
Edit: My TL;DR about why it is possibly a reasonable thing to do is that deck archetypes will always have varying strength and NOT to do something like that rewards people who slavishly follow widely-available data (from sites like Vicious Syndicate or HSReplay.net) in choosing what to play, as opposed to people who master whatever deck they choose and play with care and forethought (which are usually what people think of when they talk about "skill" at Hearthstone.)
Edit 2: From a practical standpoint, as a player, it would seem to me that one benefit is that it makes it more viable to play a wider variety of decks. Something that's not as viable against all decks in the meta averaged together can still be worth mastering and playing well, and that means more different decks that can perform well on ladder.
My main issue with the 50% argument is that I am a player and not a Blizzard executive. My loyalty is for the game, not profits. If I owned stock I might be more inclined. I do understand the challenge they face - Hearthstone is a 'first' CCG for a lot of people. People unfamiliar with these types of games might be inclined to think that a 'good' player win almost every match they play. It would be easy to understand a lot of them quickly becoming demoralized and uninstalling the game. I think the ranking indications exacerbate this - I mean half of players are in 2 of the 21 ranks (ignoring 25-21). I guess on the other hand, people who think that they are doing good at the game don't buy cards either.
I do agree to some extent about the deck archetypes angle. Some will say nerf cards faster or make a ban, but neither of these is a realistic option. Then again, lopsided decks are somewhat self inflicted. They print Drakonid Operative, then publicly state it was to make keep dragon priest viable for the expansion (I'll have to find the blue post). Then when everyone plays it, they throw you a few extra bad matchups to keep things level? My issue with all of this is that well it may be well-intentioned, it is not in the spirit of fair play or legitimate competition. I would go so far as to encourage Blizzard to clean it up and hire a 3rd party to certify that the system is unbiased - like American Idol did during its run.
Rollback Post to RevisionRollBack
Free to try and find a game, dealing cards for sorrow, cards for pain.
Although I don't have any conclusive evidence or enough data to prove something statistically, personal experience tends to point matchmaking through specific cards (or even decks). Similar to one of the posts above, I played a C'Thun Warrior out-of-whim one day to be matched to another C'Thun deck. And on another instance, I used a buff priest similarly out-of-whim (i.e. Shadow Ascendant) only to be match with the similar archetype.
I wouldn't be surprised if matchmaking is tiered -- something like: ranking -> specific keycards (commonly used cards, legendaries, DKs) -> length of active play. By putting the community into subsets, it'd be easier for Blizz to match specific players. I brought up length of active play because recently (maybe since Oct/Nov?) I've been paired against gold players frequently in Standard matches even if I don't have a single gold hero and hardly ladder at all. Using one-off weird decks still pits me against common meta deck archetype until I start having 5 consecutive losses.
Blizz has enough data to map a model to retain a business-oriented mmr, and at the same time balance it with the competitive scene. What they do not have a direct control of are forums such as these, but given a lot of people tend to net deck or use in-fashion decks from pros, Blizz can use pro player decks as templates to match deck specific deck archetypes against each other.
P.S. Hopefully what I said made sense... I'm just having my coffee right now. But please keep up the conversation. :)
Fair enough. I would say some of how reasonable it is depends on specifics of how it works. It's possible they're just doing a coarse class vs. class adjustment, in which case clever deckbuilding details can still get one an edge. It's also possible they're trying to identify archetypes by looking for specific cards or card groupings. They might even be using some type of automated means to identify groupings of decks or archetypes, something like Vicious Syndicate's "class radar" feature.
All that said, I do personally like the idea that ranking should be a test of execution and not "what do I put in my deck." As a player, that would mean to me that the optimal strategy might be finding a difficult-to-play deck and getting really good with it, rather than just going for easy and overall powerful. (Arguably, like Spiteful Summoner Dragon Priest.)
Edit: BTW, statistics above are from my lunchtime games at work. I'll check statistics at home later today to see if they agree.
All that said, I do personally like the idea that ranking should be a test of execution and not "what do I put in my deck." As a player, that would mean to me that the optimal strategy might be finding a difficult-to-play deck and getting really good with it, rather than just going for easy and overall powerful. (Arguably, like Spiteful Summoner Dragon Priest.)
I feel guilty, but do agree with your example. Yesterday, I tried the Spiteful Summoner Dragon Priest and had enough sticky minions plus huge tempo swings by turn 6ish. Throw-in Spellbreakers and you can aggro with the big minions and win by turn 8.
Anyway, back on topic, yes, it would be ideal for players that ranking should be a test of execution. it's refreshing to see powerful decks that need expert piloting to be successful (e.g. Patron Warrior back in the days -- there were one or two win conditions, and the key strategy is to survive until all pieces are together). But a balance has to be struck somewhere, mmr can't be solely be based on ranks -- it'd be greatly frustrating to new players. Conversely, maybe frustrating to old players who want to play fun decks, but given their length of play may be more tolerable and acceptable to losing.
More data is fun, please do post your findings when you get more!
I get the impression that the MMR (assuming it exists), is mostly archetype based, with possibly a smaller emphasis on class. Since some archetypes are more prevalent to a specific class, this can explain the skewed numbers of same class matchups as well.
My games last night consisted of mostly Control Warlock, with again, the majority of my games being against Warlock by a more than 2-1 margin.
I then switched to a Big Spell Hybrid Hunter. This deck almost seemed to "confuse" the system. The deck essentially plays out like same old Mid-Hunter, with the exception of Summoner's and 2x Call of the Wild. The matchups there were nearly all aggro / midrange.
Playing Control Warlock, my matchup breakdown for 10 games was 5 Warlock, 3 Priest, and 2 Rogue.
After swapping to Hunter, my matchups were 4 Hunter, 2 Paladin, a Priest (running an aggro list ... Patches, Keleseth, Stonehill, ect ... with some dragons mixed in), and aggro druid.
In other words, 2 completely different sets of matchups, all played within a couple hours of each other, and all in 13-12 ranks. And again .. while playing Warlock, I faced 5 Warlocks, and no Hunters. When playing Hunter, I faced 4 hunters, and no Warlocks.
Rollback Post to RevisionRollBack
I wanna glide down, over Mulholland I wanna write her, name in the sky I wanna free fall, out into nothin' Gonna leave this, world for awhile
Quick follow-up: I got home and looked at stats on another 45 games of Spiteful Summoner priest, played on my home computer.
Over my total of 65 games played both at lunchtime and at home, these were the numbers for the three top-represented classes:
18 games (27%) were paladin
18 games (27%) were priest
10 games (15%) were rogue.
Priest is exactly where it ought to be based on representation in the population. Paladin, at 27%, is a bit over double its expected representation, and was overrepresented in both sets of games. Rogue, at 15%, is also overrepresented by about 50%, but the number of games is low enough that that could just be random variation.
Nevertheless, if one wanted to overrepresent classes to cancel out a wide range of other good match-ups, rogue and paladin would be the ones to match me against. I feel this picture is very consistent with the hypothesis that this is how matchmaking now works.
Rollback Post to RevisionRollBack
To post a comment, please login or register a new account.
I really, really not trying to make this a tin foil hat thread, and I am not really salty about this either, because the results weren't horribly bad, but ...
I just cannot see how some recent events can just keep getting written off as coincidence. I really need to start a spreadsheet on this so I can gather more accurate data, but it is prevalent enough that it has become predictable.
Just using one instance, last evenings play. I was playing a Pirate Warrior deck ... fast aggro looking to hopefully climb to a new rank floor on the ladder. It's a homebrew deck, but sticks pretty close to "typical Pirate Warrior" (Honestly, how creative can you really get with that archetype anyway) ... but thats not important for the topic.
I played 9 games, I believe, and of those 9, all but 1 were other fast aggro / midrange decks ... several Hunters running the hybrid Spell / Minion setup, a non-evolve Shaman, and a couple Paladin's. No problem ... the deck can hold its own, but they were essentially just mirror matches in a way. I just assumed that was the micro-meta I was in at the time, so I decided to switch decks to Control Warlock.
After switching decks, next 6 matches saw Warlock, Mage, Priest ... not a single aggro deck to be found. In fact. the first match was a near exact mirror match with another Control Warlock ... 32 minutes ending in fatigue, and losing by 1 HP.
And it is not just last night ... I can almost count on when changing decks, that the classes I face will change as well.
Here is the big thing .. I don't mind it if the game does match specific decks with other specific decks (I should say archetypes, to be more accurate), I am just ticked off because what seems to be such an obvious thing is either denied by the developers, or just not confirmed. (I honestly don't know if they have ever outright denied it, or just haven't ever confirmed it).
If we knew this information to be true, we could take action by making the use of tech cards, and not feel as if using them was a waste.
I wanna glide down, over Mulholland
I wanna write her, name in the sky
I wanna free fall, out into nothin'
Gonna leave this, world for awhile
If you add a tech card to your deck you stop seeing decks that you're teched against anyway.
I can add another story to your bucket ... i played a Cthun deck again recently (like after several months) and first two matches i had were vs another Cthun decks ... i havent played vs Cthun decks for like half an year :D
But rly, i dont believe there is any secret matchmaking system, that puts certain types of decks together. You would need more data than u can collect on your own, to even try and see a pattern. Couple hundred games u can do yourself is still way too small sample size for u to make any kind of assumptions so ... yeah.
- Click Here To Join Us On Discord! -
Confirmation bias is a heck of a phenomenon
I'm one of the people who has argued pretty extensively here that Blizzard is almost certainly not lying about how their matchmaking works. That doesn't mean you're wrong though.
The last time one of the developers talked in detail about how matchmaking works in Hearthstone was January of 2017. Here are a couple references where they state that only star rating is used in Ranked mode.
http://www.hearthhead.com/news/breaking-down-hearthstones-competitive-matchmaking-process
http://www.hearthpwn.com/news/2177-is-matchmaking-rigged-max-mccall-explains-the
Ben Brode also commented around the same time on ways they're looking at trying to make ranked mode less punishing, particularly for newer players:
http://www.hearthpwn.com/news/2135-ben-brode-the-new-player-experience-needs-work
This does suggest that experimentation or tweaks to the matchmaking system may happen over time.
So, a couple thoughts. First, for a game like Hearthstone, a year is a very long time in their development cycle. There's almost certainly been a lot of development, discussion, experimentation, and maybe changed minds about certain things since January, eleven months ago.
Also, here's something I heard while attending Blizzcon 2017 in one of the engineering panels that's not conclusive, but it is very interesting. The transcript is linked with my comments below.
http://warcraft.blizzplanet.com/blog/comments/blizzcon-2017-codecraft-hood-panel-transcript/3
One of the server-side battle.net developers said this:
"Derek: From a mathematical standpoint, what I really enjoy working with our BI [Business Intelligence] team with is on the matchmaking for Hearthstone, and that’s where I think the rubber meets the road; because they actually look at all the numbers, they do the analysis, and then they say: “Okay, in order to adjust our matchmaking so that we have this nice perfectly formed curve that we want these are the numbers that you need to put into the system.”
So, as a software engineer you understand how to craft code, but the more higher mathematical understanding of what’s going on in the meta, it’s very important that we have them to help us do our jobs."
When I heard the bolded comment, my first reaction was that sometime in the last year they may have made changes to matchmaking to try to drive various deck choices as close as possible to 50%. The result of this might well be that selecting a relatively more powerful deck makes it disproportionately likely to be matched against decks against which one is weak. Selecting a weak deck might result in being matched against opponents against which one is strong.
Since the comment was oblique and was in response to a question about something else (the developers' use of business intelligence data) it's not clear how such a system might work exactly. It's also not evident that the system is stable. (For example, there may be multiple methodologies in place being A/B tested against each other, or the methodology might be changed or tweaked over time based on collected data, so that your experience one day may not match someone else's experience a week later.)
Confirmation bias is definitely a thing, and anecdotes like "oh I got matched against twelve pirate decks when I switched to murloc paladin" don't prove anything, since there is a lot of randomness in the system. Without extensively collecting and analyzing data, or without further Blizzard comment, I'm not sure we'll be able to get a clear idea of what the matchmaking is actually doing.
Note that there are potentially solid gameplay and business reasons for implementing such a system. The gameplay justification is that matching decks against each other more evenly helps distinguish quality of play within a match from deck choices in the ranking system, and since information on the internet is such that good netdecking can't be distinguished from good deckbuilding, ranking may as well be more about strong execution once the game starts.
The business justification would be that beginning and f2p players can't always choose their first viable deck, so bringing them closer to even match-ups in terms of archetype is likely to improve their win rate and therefore their retention.
Edit: Note that it's also possible that matchmaking is weighted by class rather than by deck. If this were true (and it would be much simpler to implement) then a class with two or more widely differing strong decks, like druid with aggro-token and Jade recently, might be able to strongly outperform just because class-based matchmaking can't distinguish between those two decks.
I see no single reason why matchmaking should not be rigged, esp. from business perspective.
Hall of Fame (ignore list): aleathas, Baylith, cendol, DiamondDM13, Dominieq, doomr, glitterprincess, hamtarofr, Heck, Jwigg33, Kaladin, Krewger, Legend_Entomber, libertyprime, Maukiepaukie, PandarenHero, randjob, s2mikey, SchruteBucks, The_Giratina, TheWamts, ticandtac, tictactucroc, tsudecimo, WaffleMonstr
I played Jade Shaman almost exclusively since the start of the expansion, and I experienced the same change in what decks people play. At the start people played a lot of Hunter, then I saw the meta switch to control with Priest/Warlock. What you're experiencing is in some ways a coincidence, but it's also just how the meta flows, especially this early in the expansion. People will try something new and cool like Spell Hunter, people will see them doing that and copy their deck, then they end up getting stomped by some other deck so they switch to that one, and so on.
If you switch decks to deal with the metagame, why wouldn't everyone else be doing the exact same thing?
Also, keep in mind how random distribution works. Even in a perfectly random system, there will always be a few players experiencing rare events that may seem arranged in some way, but in reality they're just outliers. For example you could imagine every person on Earth tossing 5 dice three times. Most of those people will just get random arrangements of numbers, but a few will get lucky and roll a 6 on every dice one, two or even all three times. Now of course the people who rolled random numbers aren't going to go to forums and write about their experiences, but the people who got perfect rolls may very well do that because they had a rare event happen to them. That is why conspiracy threads like this one become so common and seem plausible, because out of all of the millions of players, only those who experience these rare events end up telling others about it.
I agree very much that it is essential to disregard personal and anecdotal experience in evaluating this, for all the reasons you describe. Unless someone like Vicious Syndicate or hsreplay is going to do a mass data analysis to identify whether something unusual is going on, we're stuck with parsing Blizzard comments and not really knowing unless they decide to share.
It also allows the player to not feel as much pressure to play an archetype they may not want to play ... if the local meta is primarily fast decks, and I want to play that C'thun deck I haven't played in a while, it would make sense for the system to match me with slower decks, where I might stand a chance.
To me, it just seems too obvious. Coincidence can only be assumed so many times. Much like what Sinti described, I have swapped to some very rarely played decks as well, only to have similar decks come up in the first couple games.
I wanna glide down, over Mulholland
I wanna write her, name in the sky
I wanna free fall, out into nothin'
Gonna leave this, world for awhile
In this thread from a couple weeks ago, I posted a quote from Blizzcon that seemed to suggest that some type of meta-based matching (either by class or by deck contents) had been adopted. Here's a little data that suggests there might be more to see:
I've been playing a dragon priest deck a fair amount recently, taking me from rank 5 to 3 (and in the losing streak represented in this data set, back again.) Here's the distribution of classes I've faced:
Here's the latest Vicious Syndicate meta report win rate chart for the three major Priest archetypes. (The four darkest red boxes on the Dragon Priest row, representing low win rates, are aggro paladin, murloc paladin, Kingsbane rogue, and pirate warrior) :
Finally, here's the overall class distribution in ranks 5-1 from the most recent VS meta report:
Note that dragon priest is relatively strong against everything except aggro paladin, murloc paladin, Kingsbane rogue, and pirate warrior. Meanwhile, my number of matches against paladin on a dragon priest deck (eight out of 20) far exceeds the two or three game expectation value for a 20 game series.
This suggests that matches are selected in such a way as to push the result toward 50%. It's not absolute proof, of course. A completely random series of matches could look the same. However, it certainly seems to warrant a closer look.
Here are a couple thoughts on what the implications of such a choice would be on Hearthstone from a business standpoint, and for us as players, which might shed some light on why it might make sense to choose to do something like this:
Players like winning more than they like losing, and losing too much is probably highly correlated with disengaging from the game. This is why they've stated that they've biased matchmaking for new players toward other new players, for example. So, from a business standpoint, attempting to account for choice of class/deck somehow in trying to get each individual player as close as possible to a 50% win rate makes sense. The most players will stay engaged if the winning is spread around as evenly as possible.
Another possible argument for it is that it helps to factor out the coarse decision of what class/deck to play from measurement of skill. By adjusting rates of matching against other decks to reduce overall dominance of one deck/class choice, the remaining variation is more likely to be driven by quality of play, which is what the ranking system (particularly at ranks 5+) is meant to measure.
One potential downside to such an approach is that if (as implied by the Blizzcon quote above) matchmaking is driven by business intelligence data, it means that changes that might drive the last business intelligence data drop out of sync with the current game (such as releasing an expansion, buffs/nerfs, or major new unexpected deck-building insights by players) could allow exploitable class/deck choice advantages for a window of time until the data is again updated.
I should note that I don't consider using such matchmaking strategies "rigging," to the extent that that term connotes unfairness. There are valid fairness reasons to want to minimize the impact of deck choice on measuring player skill, such as that the availability of aggregate data to players like I've posted here makes choosing a strong deck archetype pretty trivial.
If the game (ladder mode) is supposed to be competitive, you should be matched against someone of your same rank at random, period. Anything else is completely unacceptable from a competitive standpoint. Not all players will have a win rate above 50% - it's impossible since every game produces a winner and a loser (or 2 losers).
Any 'MMR' beyond straight ladder rank is tantamount to giving someone a win they don't deserve at the expense of someone who earned it. Sure, sometimes you get a favored match, but it should be random, not manipulated to aid a player on a loss streak.
Free to try and find a game, dealing cards for sorrow, cards for pain.
Please, if you have a moment, take a look at my follow-up post. Sounds like you disagree but I wonder if you'd like to offer any thoughts about why my arguments don't hold water for you.
Edit: My TL;DR about why it is possibly a reasonable thing to do is that deck archetypes will always have varying strength and NOT to do something like that rewards people who slavishly follow widely-available data (from sites like Vicious Syndicate or HSReplay.net) in choosing what to play, as opposed to people who master whatever deck they choose and play with care and forethought (which are usually what people think of when they talk about "skill" at Hearthstone.)
Edit 2: From a practical standpoint, as a player, it would seem to me that one benefit is that it makes it more viable to play a wider variety of decks. Something that's not as viable against all decks in the meta averaged together can still be worth mastering and playing well, and that means more different decks that can perform well on ladder.
My main issue with the 50% argument is that I am a player and not a Blizzard executive. My loyalty is for the game, not profits. If I owned stock I might be more inclined. I do understand the challenge they face - Hearthstone is a 'first' CCG for a lot of people. People unfamiliar with these types of games might be inclined to think that a 'good' player win almost every match they play. It would be easy to understand a lot of them quickly becoming demoralized and uninstalling the game. I think the ranking indications exacerbate this - I mean half of players are in 2 of the 21 ranks (ignoring 25-21). I guess on the other hand, people who think that they are doing good at the game don't buy cards either.
I do agree to some extent about the deck archetypes angle. Some will say nerf cards faster or make a ban, but neither of these is a realistic option. Then again, lopsided decks are somewhat self inflicted. They print Drakonid Operative, then publicly state it was to make keep dragon priest viable for the expansion (I'll have to find the blue post). Then when everyone plays it, they throw you a few extra bad matchups to keep things level? My issue with all of this is that well it may be well-intentioned, it is not in the spirit of fair play or legitimate competition. I would go so far as to encourage Blizzard to clean it up and hire a 3rd party to certify that the system is unbiased - like American Idol did during its run.
Free to try and find a game, dealing cards for sorrow, cards for pain.
Really well written thread, props to you guys.
Although I don't have any conclusive evidence or enough data to prove something statistically, personal experience tends to point matchmaking through specific cards (or even decks). Similar to one of the posts above, I played a C'Thun Warrior out-of-whim one day to be matched to another C'Thun deck. And on another instance, I used a buff priest similarly out-of-whim (i.e. Shadow Ascendant) only to be match with the similar archetype.
I wouldn't be surprised if matchmaking is tiered -- something like: ranking -> specific keycards (commonly used cards, legendaries, DKs) -> length of active play. By putting the community into subsets, it'd be easier for Blizz to match specific players. I brought up length of active play because recently (maybe since Oct/Nov?) I've been paired against gold players frequently in Standard matches even if I don't have a single gold hero and hardly ladder at all. Using one-off weird decks still pits me against common meta deck archetype until I start having 5 consecutive losses.
Blizz has enough data to map a model to retain a business-oriented mmr, and at the same time balance it with the competitive scene. What they do not have a direct control of are forums such as these, but given a lot of people tend to net deck or use in-fashion decks from pros, Blizz can use pro player decks as templates to match deck specific deck archetypes against each other.
P.S. Hopefully what I said made sense... I'm just having my coffee right now. But please keep up the conversation. :)
Fair enough. I would say some of how reasonable it is depends on specifics of how it works. It's possible they're just doing a coarse class vs. class adjustment, in which case clever deckbuilding details can still get one an edge. It's also possible they're trying to identify archetypes by looking for specific cards or card groupings. They might even be using some type of automated means to identify groupings of decks or archetypes, something like Vicious Syndicate's "class radar" feature.
All that said, I do personally like the idea that ranking should be a test of execution and not "what do I put in my deck." As a player, that would mean to me that the optimal strategy might be finding a difficult-to-play deck and getting really good with it, rather than just going for easy and overall powerful. (Arguably, like Spiteful Summoner Dragon Priest.)
Edit: BTW, statistics above are from my lunchtime games at work. I'll check statistics at home later today to see if they agree.
I get the impression that the MMR (assuming it exists), is mostly archetype based, with possibly a smaller emphasis on class. Since some archetypes are more prevalent to a specific class, this can explain the skewed numbers of same class matchups as well.
My games last night consisted of mostly Control Warlock, with again, the majority of my games being against Warlock by a more than 2-1 margin.
I then switched to a Big Spell Hybrid Hunter. This deck almost seemed to "confuse" the system. The deck essentially plays out like same old Mid-Hunter, with the exception of Summoner's and 2x Call of the Wild. The matchups there were nearly all aggro / midrange.
Playing Control Warlock, my matchup breakdown for 10 games was 5 Warlock, 3 Priest, and 2 Rogue.
After swapping to Hunter, my matchups were 4 Hunter, 2 Paladin, a Priest (running an aggro list ... Patches, Keleseth, Stonehill, ect ... with some dragons mixed in), and aggro druid.
In other words, 2 completely different sets of matchups, all played within a couple hours of each other, and all in 13-12 ranks. And again .. while playing Warlock, I faced 5 Warlocks, and no Hunters. When playing Hunter, I faced 4 hunters, and no Warlocks.
I wanna glide down, over Mulholland
I wanna write her, name in the sky
I wanna free fall, out into nothin'
Gonna leave this, world for awhile
Quick follow-up: I got home and looked at stats on another 45 games of Spiteful Summoner priest, played on my home computer.
Over my total of 65 games played both at lunchtime and at home, these were the numbers for the three top-represented classes:
18 games (27%) were paladin
18 games (27%) were priest
10 games (15%) were rogue.
Priest is exactly where it ought to be based on representation in the population. Paladin, at 27%, is a bit over double its expected representation, and was overrepresented in both sets of games. Rogue, at 15%, is also overrepresented by about 50%, but the number of games is low enough that that could just be random variation.
Nevertheless, if one wanted to overrepresent classes to cancel out a wide range of other good match-ups, rogue and paladin would be the ones to match me against. I feel this picture is very consistent with the hypothesis that this is how matchmaking now works.