Guild Wars 2

OK, here was my original rough analysis of the flaws in the MMR system:

https://forum-en.gw2archive.eu/forum/game/pvp/A-serious-examination-of-S2-MMR-Math-Inside

But since a majority of the forum base thinks I’m wrong, and swears there is no MMR hell………I offer you a challenge:

Prove me wrong and I’ll buy you SHINIES! (Some random stuff from the gem store)(NA Server)
If a dev wants to play, I’ll chip in cash to a charity of their choosing.

Challenge 1 (Small Shiny): Prove a player with system MMR in the bottom 5% but actual skill of 75% for his matchmaking area can obtain a system MMR that reflects his skill. Assume he’s at 0 pips in Ruby.

Challenge 2 (Big Shiny): Prove a player with system MMR in the bottom 5% but actual skill of 50% for his matchmaking area can obtain a system MMR that reflects his skill. Assume he’s at 0 pips in Ruby.

Challenge 3 (Win Both Shinies and a bonus Shiny): Create a simulation (sim1) using the available information for this season, create a simulation (sim2) using purely GLICKO for matchmaking. Run both. Find that sim1 produces MMR for players with at most twice the error generated by sim2. Feel free to use any statistically valid method for calculating error

Best to assume a normal distribution of players, but I’ll probably pay out for a uniform distribution solution.

Requirements:
Math and example cases, no anecdotal evidence.
You must assume MMR is an accurate predictor of game win loss.
You must assume all players other than the player trapped in hell have accurate MMR. You must use the current S2 matchmaking sytem.
Other than that, no solutions that rely on a small percentage chance event to occur or require an inhumanly possible number of games.
No solutions that require everyone (or most or a lot or etc) above that player to rank out of that area.
No solutions relying on disconnections or team play.
No solutions requiring the player to sit out playing a portion of the season.

You’re proving that this players skill will get him where he belongs.

You have 24 hours…..enjoy…..

-Tene

Edit:To clarify this is for a SOLO QUEUE PLAYER
Edit2: First solution to each challenge, I’m not loaded
Edit3: Challenge 3 Posted to make it easier to see

Nice incentive.

Seriously, there are ppl who simply are not interested to hear what they do not like to hear is the problem and if a fact or a truth is a liability to how they like to think the world function they just dismiss it.

It is just so sad to witness so little good faith on that topic so far. Some ppl are obviously trolls tho…

+1 for the analysis

i think you can just trust Anet and take their word that the system is based for skill and unworthy players wont be able to grind anymore unless they abuse MMR in some way (like many did in s1)

Lexander.4579:

i think you can just trust Anet and take their word that the system is based for skill and unworthy players wont be able to grind anymore unless they abuse MMR in some way (like many did in s1)

Or…..I could look at it and as far as I can tell it’s flawed…..and then I could ask people if they agreed with the analysis…..and then I could challenge people to offer proof that it’s not flawed….then I could offer to basically pay for proof it’s not broken…..

Tenebria.7239:

Lexander.4579:

i think you can just trust Anet and take their word that the system is based for skill and unworthy players wont be able to grind anymore unless they abuse MMR in some way (like many did in s1)

Or…..I could look at it and as far as I can tell it’s flawed…..and then I could ask people if they agreed with the analysis…..and then I could challenge people to offer proof that it’s not flawed….then I could offer to basically pay for proof it’s not broken…..

the majority of players are not good, as a huge part (and flaw) of human nature they will naturaly seek excuses instead of admitting that they should improve and thus they will agree with anything that allows them to stick with their excuses and cling into any sort of vague proof which makes them think they are not bad

considering all these arguments… i doubt people will manage to get to the facts when the facts are against them

This is pretty easy to do. You simply need to do a simulation to see the result.
The simulation should be relatively simple but will need some information support from Anet.

Lexander.4579:

the majority of players are not good, as a huge part (and flaw) of human nature they will naturaly seek excuses instead of admitting that they should improve and thus they will agree with anything that allows them to stick with their excuses and cling into any sort of vague proof which makes them think they are not bad

No you’re just making assumptions you don’t have any actual proof that it is the case. You apparently rode to legend on the back of high mmr so anybody should be able to do it.

Lexander.4579:

the majority of players are not good, as a huge part (and flaw) of human nature they will naturaly seek excuses instead of admitting that they should improve and thus they will agree with anything that allows them to stick with their excuses and cling into any sort of vague proof which makes them think they are not bad

considering all these arguments… i doubt people will manage to get to the facts when the facts are against them

This is true, but…

Just because players can play better doesn’t mean the system isn’t unfair and broken. There can be unwarranted complaints, Dunning Kruger stuff, and all sorts of delusions, but the fundamental is that players are grouped with like MM ratings and matched against random groups. Those ratings are based on win/lose. No matter how bad or delusional people are, this should still logically form an unfair playing field where some bads are moving up and some goods are stuck at the bottom. Even if a terrible player ruins a match, his MMR is so high that the next one will probably give him 4 good players and the enemy team will have 4 bad ones. It will undo his loss and he will snowball towards victory.

There’s a good reason why some people swear by soloq and others swear by 5 man teams…

Lexander.4579:

Tenebria.7239:

Lexander.4579:

i think you can just trust Anet and take their word that the system is based for skill and unworthy players wont be able to grind anymore unless they abuse MMR in some way (like many did in s1)

Or…..I could look at it and as far as I can tell it’s flawed…..and then I could ask people if they agreed with the analysis…..and then I could challenge people to offer proof that it’s not flawed….then I could offer to basically pay for proof it’s not broken…..

the majority of players are not good, as a huge part (and flaw) of human nature they will naturaly seek excuses instead of admitting that they should improve and thus they will agree with anything that allows them to stick with their excuses and cling into any sort of vague proof which makes them think they are not bad

considering all these arguments… i doubt people will manage to get to the facts when the facts are against them

Oh please, if you want to use “human nature” you might add that those who are in the good favor of any given system will also seek excuses to avoid loosing their status and privilege.

Now if we are done with this pathetic joke of an non-argument we might fall back on facts and logic to find out what we say we want to know? The OP has done some very legit work, let’s see what you offer that hold any value.

what facts have been provided?

the burden of truth should be on those that make the claims, people can post screenshots of game scores but they probably dont tell the whole story.

as much as alot of you want to believe the people winning are deluded into defending a flawed system one could argue the same for the other group, but what group is more likely to go to the forums to voice their disapproval?

people on these forums dont understand those complaining are most likely the vocal minority.

NeXeD.3042:

what facts have been provided?

the burden of truth should be on those that make the claims, people can post screenshots of game scores but they probably dont tell the whole story.

as much as alot of you want to believe the people winning are deluded into defending a flawed system one could argue the same for the other group, but what group is more likely to go to the forums to voice their disapproval?

people on these forums dont understand those complaining are most likely the vocal minority.

I did post facts, and then what I believe the results of those facts are. The idea is if someone has better facts or more accurate results than what I had that disprove mine: I PAY THEM.

No insulting people necessary for either side of this discussion.

so considering MMR is hidden from the player, isn’t this impossible to actually test?

choovanski.5462:

so considering MMR is hidden from the player, isn’t this impossible to actually test?

Sure, it’s impossible to test. But not impossible to simulate or theorycraft since we know the general formula’s.

Unless you know the standard deviation of the MMR this is impossible to prove either way. With that piece of knowledge and assuming the average MMR is 50% then I could calculate exactly your probabilities assuming a normal distribution.

@OP

Long post incoming on MMR hell

So MMR hell is a place where better players get stuck with division titles that are beneath them. Why is Ruby an unacceptable place, a place that doesn’t reflect the entitled progress of someone with 50th percentile skill? Do you think Diamond/Legend are better places? The leagues are moving towards being skill based, and Ruby is around the middle of the pack.

I would say that a 75th percentile skill player who managed to get a bottom 5th percentile MMR is probably confused and disillusioned as to how valuable his playstyle is to winning matches.

I’ll use an example, many people think that AFKing on a point after capping is carrying: “I’m holding a point, but my team lost everything else.” When their team was 4v5, and then the defending player gets snowballed on, and in his eyes, he’s 1v3 all game, and has bad teammates.

MMR hell is more explained by the fact that in the depths of this hell, the players who do very, very bad plays are unpredictably showing up, MMR can’t really predict them early on.

OP’s assumptions that paint the picture of an inescapable MMR hell are not fairly assumed:

MMR hell is all about being underrated. OP assumed you can’t encounter other underrated players with the same MMR, in the same boat, (all other teammates assumed to have accurate skill captured by MMR). Sometimes the MMR hell inmates, ie, underrated players, luckily get together on a streak, snowballing their MMR.

MMR can’t be assumed to be the accurate end-all of match results. MMR is really volatile and inaccurate, dude. That’s one of the fundamental points of MMR hell – your MMR doesn’t predict your better ability.

Volatile MMR is one thing, but, there’s even a bunch of other dice rolled. Class makeup, partied people even if you’re solo, it’s really not able to be assumed volatile MMR is the match results indicator.

You could fail the MMR roll, and then win the roll for comp makeup, premade players, other underrated players, unpredictably clueless players, or any of other factors. Since random, it’s possible to be even carried, let alone claw your way out.

So that’s all the reasons why I say your assumptions aren’t good things to assume.

The fact that MMR grouping will favor one side, it’s only one factor of many.

Now we haven’t even touched on the fact that an underrated player in MMR hell will always have the ability to carry, as a matter of fact. One single player having carry ability above that division such as superior mechanical skill, ability to multiclass many meta builds and recognize the counter picks, these players can for sure generate value.

A lot of players I’ve met who I’ve coached through ‘MMR hell’ have just been holding their own, while enemies were carrying by farming the ‘MMR hell inmates’s’ team. If you deserve to move up, you gotta be the one farming value, learning how to carry like your enemies are. A lot of it is seemingly luck until you recognize the strategies.

Remember, if you get a super bad roll, you can still carry and lose, or have no opportunities to carry, more accurately. This is the real feeling of being in MMR hell, but it’s not permanent.

Just remember, if you’re holding your own, not carrying, you’re not underrated, or in MMR hell. So learn things, or wait to get lucky if you want to progress.

And good luck to any MMR hell inmates

Also, have them prove that Hell does not exist while they are at it.

@chaith

Problem:

When you hold your own against people who carry and thats fine. But in system eyes you decrease to someone who can’t hold your own. Which leads you to be paired with someone who can’t hold their own. Which leads to same unbalanced match up when “hold their own” people arrives. System where you are stuck and pray on rng to get out of it’s not a good system. Every MMR based system in another games is based on people who hold their own playing against exact same people where skill progress is a key factor to progress, but it doesn’t diminish you because you ve faced way stronger opponents and doesn’t involve lesser teammates.

Better another example can underrated Zoose with diamonds (being on the bottom of devision) carry against full abjured roster and everybody else full roster x100 times?

I would really prefer if people didn’t actually talk about common mistakes or how to play better etc in this thread. That isn’t really the point, it’s not a whine about how I’m misranked or about the backpack or whatever. It’s about whether the system is sound and consistent and produces the appropriate output from the input it is given

@Chaith

I wasn’t saying anything about how a player got there or about their playing ability.
IMO, a fair matchmaking system would take a player at 5%/75% to 75%/75% in a reasonable amount of time. Also in that system this would be an extreme abberation.

I’m not saying people can’t play out of it by being even better than they are now, I’m just saying it produces inaccurate results. I.E. two players with the same skill will not have comparable MMR in this system because it encourages those on the outlying areas to remain there.

The question is: Is this system accurate? Prove that when presented with an edge case it correctly compensates, because to me it looks like it fails those cases.

That’s the point. And you have to assume that it’s accurate for everyone else or if you don’t then your proof would fail because it relies on the system failing for player two in order to work for player one.

When you make Ruby, you don’t have bottom MMR. You have to get a positive win rate in any division in order to progress. The more you lose, you’re falling into a lower and easier PIP range, in reality people are trading wins and not tanking MMR. There’s only a problem when your backwards PIP progression is stopped, and you manage to not be able to carry for so many games that your MMR gets tanked, and the disadvantage becomes greater for that unchanging PIP range.

A simple fix ArenaNet could implement would be no MMR drops when you’re zero Pips in a tier or division (still lose MMR for losses in Amber)

You guys are aware that in glicko, MMR based games, etc, you experience reduced MMR loss when you lose vs people with much higher MMR?

Chaith.8256:

When you make Ruby, you don’t have bottom MMR. You have to get a positive win rate in any division in order to progress. The more you lose, you’re falling into a lower and easier PIP range, in reality people are trading wins and not tanking MMR. There’s only a problem when your backwards PIP progression is stopped, and you manage to not be able to carry for so many games that your MMR gets tanked, and the disadvantage becomes greater for that unchanging PIP range.

A simple fix ArenaNet could implement would be no MMR drops when you’re zero Pips in a tier or division (still lose MMR for losses in Amber)

Not disagreeing with you at all. Indeed if the rules are changed to allow division drops, the system will eventually correct itself. There are many small changes that would allow this.

I should also add the disclaimer here: many people who feel they are in MMR Hell may or may not be

My point was that it is a possibility in the current system, and since many people insisted that it was not and the system was fair and consistent (i.e. ALL players wind up where they deserve based solely on skill)

So I’m willing to pay someone a prize for proving that a player will always wind up where skill determines. (The two challenge cases are just simple examples where it was easiest for me to locate a potential system failure.)

Give me an account that is in “MMR hell”, and I will be happy to stream how the games go and prove it is a result of your own lack of skill.

Other than that, there is no “mathematical” argument anybody can make since we cant prove our own MMR.

Celeras.4980:

Give me an account that is in “MMR hell”, and I will be happy to stream how the games go and prove it is a result of your own lack of skill.

Other than that, there is no “mathematical” argument anybody can make since we cant prove our own MMR.

Of course higher skill takes you out of it. I just feel like you’re missing the point. You shouldn’t need that higher skill if the system was accurate and fair.

And you don’t need to know your mmr. You know the rules of the system. So it’s easy to simulate. There are some parameters you don’t know, but I don’t think you can find any value for those parameters where your simulation doesn’t fail to be accurate. If you can-Grats-You’ll win the challenge.

I was almost diamond t5 then lost a game vs lower MMR team and it kittened my MMR. now at t2 and cant seem to win

I had a warrior on my team who went far and died 1v1 the whole game. How are you supposed to win in that case. I think once your MMR lowers from an unlucky losss or lose streak vs lower MMR team you are screwed.

Celeras.4980:

Give me an account that is in “MMR hell”, and I will be happy to stream how the games go and prove it is a result of your own lack of skill.

Other than that, there is no “mathematical” argument anybody can make since we cant prove our own MMR.

“your own” do you even read ?

So you agree with me that backwards PIP loss being stopped while losing dozens of games, thus tanking MMR is the only way, if at all, to increase difficulty past what you can reasonably carry.

This is a temporary rut, MMR hell is not forever. If you match vs. far superiorly rated enemies, you lose little, and gain lots on a win. That’s how Glicko/MMR games work. Due to this Eventually you’ll crawl out, and get where you belong, after fluking a win based on good rolls on the other factors I mentioned before.

The only way you’ll not get where you deserve is if you give up there.

Chaith.8256:

So you agree with me that backwards PIP loss being stopped while losing dozens of games, thus tanking MMR is the only way, if at all, to increase difficulty past what you can reasonably carry.

This is a temporary rut, MMR hell is not forever. If you match vs. far superiorly rated enemies, you lose little, and gain lots on a win. That’s how Glicko/MMR games work. Due to this Eventually you’ll crawl out, and get where you belong, after fluking a win based on good rolls on the other factors I mentioned before.

The only way you’ll not get where you deserve is if you give up there.

Mostly true, there are other ways to get there, tanked mmr in sapphire/emerald since you only need a 3 win streak to hit the next save point. Or doing profession achieves and losing games the system expects you to win causing faster mmr drop than pip loss. And when you remove division save points you’d have to remove all save points or it would just become T5 or 6 sapphire hell.

And every other glicko/mmr based game I know of gives you a roughly balanced match to allow your MMR to correct itself. In this system you might be relying on a low probability event and have to play an inhuman number of games to have that event occur consistently enough for the system to correct itself. (As far as I can tell)

Tenebria.7239:

Chaith.8256:

So you agree with me that backwards PIP loss being stopped while losing dozens of games, thus tanking MMR is the only way, if at all, to increase difficulty past what you can reasonably carry.

This is a temporary rut, MMR hell is not forever. If you match vs. far superiorly rated enemies, you lose little, and gain lots on a win. That’s how Glicko/MMR games work. Due to this Eventually you’ll crawl out, and get where you belong, after fluking a win based on good rolls on the other factors I mentioned before.

The only way you’ll not get where you deserve is if you give up there.

Mostly true, there are other ways to get there, tanked mmr in sapphire/emerald since you only need a 3 win streak to hit the next save point. Or doing profession achieves and losing games the system expects you to win causing faster mmr drop than pip loss. And when you remove division save points you’d have to remove all save points or it would just become T5 or 6 sapphire hell.

And every other glicko/mmr based game I know of gives you a roughly balanced match to allow your MMR to correct itself. In this system you might be relying on a low probability event and have to play an inhuman number of games to have that event occur consistently enough for the system to correct itself. (As far as I can tell)

No, not remove safe pips. That would be very unpopular. Removing MMR drops on safe pip spots, basically freeze your progress like a save point, that would be a much smarter decision.

I’d be fine if ArenaNet also tweaked down the MMR grouping matchmaking, meaning you could have a SLIGHTLY bigger spread in your teammate’s skill, just so it would clump better vs. worse players less often. The feature is much appreciated, it seems to be too strict.

But I digress, I don’t quite agree with an ‘inhuman’ assessment of games required to reach your max. For the vast majority of players who instead of insanely massive lose streaks, are more likely to trade wins in Ruby or early Diamond, I don’t quite think MMR hell exists for them.

Also, tanking MMR to gain profession achievements without actually knowing how to play those professions, I think that’s completely self inflicted m8

Sorry about that, had two thoughts in my head, and you are probably right that this doesn’t apply to a lot of the people who think they might be in mmr hell. I also probably should have mentioned in the OP the same thing can occur for average players (especially new to pvp accounts) on win streaks pushing them higher than they should be and trapping them past a save point that they can’t compete at.

And most of this damage was probably self inflicted for people at this point. I’m really trying to avoid making it about player decisions and just about the system not handling edge cases correctly is all.

Chaith.8256:

No, not remove safe pips. That would be very unpopular. Removing MMR drops on safe pip spots, basically freeze your progress like a save point, that would be a much smarter decision.

Basing opponents on mmr instead of pip range would be even smarter.

Chaith.8256:

When you make Ruby, you don’t have bottom MMR. You have to get a positive win rate in any division in order to progress…

@Chaith

That’s not taking into account the comeback pip and win streak, I can lose 4 games straight and win 3 games straight and I’ll still get legend over time:
4 games lost = -4 pips
3 games won = +5 pips (2 comeback pips, 1 pip, 2 pips from winstreak)

Now that is a horrible win ratio of 3/7 I.e. ~43% win rate that could potentially get to legend.

PadreNike.1803:

Chaith.8256:

When you make Ruby, you don’t have bottom MMR. You have to get a positive win rate in any division in order to progress…

@Chaith

That’s not taking into account the comeback pip and win streak, I can lose 4 games straight and win 3 games straight and I’ll still get legend over time:
4 games lost = -4 pips
3 games won = +5 pips (2 comeback pips, 1 pip, 2 pips from winstreak)

Now that is a horrible win ratio of 3/7 I.e. ~43% win rate that could potentially get to legend.

Yeah fun facts, but who really loses 4 and wins 3 every time without fail? If you’re progressing, let’s be honest, probably not a 43% win rate, it’s 50%+

Chaith.8256:

PadreNike.1803:

Chaith.8256:

When you make Ruby, you don’t have bottom MMR. You have to get a positive win rate in any division in order to progress…

@Chaith

That’s not taking into account the comeback pip and win streak, I can lose 4 games straight and win 3 games straight and I’ll still get legend over time:
4 games lost = -4 pips
3 games won = +5 pips (2 comeback pips, 1 pip, 2 pips from winstreak)

Now that is a horrible win ratio of 3/7 I.e. ~43% win rate that could potentially get to legend.

Yeah fun facts, but who really loses 4 and wins 3 every time without fail? If you’re progressing, let’s be honest, probably not a 43% win rate, it’s 50%+

It’s the way it went for below average tier. That’s why there is a lot whine which doesn’t necessary means “git gud”, they just can’t instantly become so much better in so short time frame.

Just imagine first week of the season you getting stomped by Good/Pros, then Above average then average. On more or less fresh ranked account with not much games, even though those losses take minimal impact it’s multiplied by the amount of losses and volatility. And system forces you to further move in the tiers and devisions, because u ll eventually win some. Which will create even more unbalanced matches.

It’s still very hard to crawl back for not above average / good player. Needless to say, that there are no even close matches. That experience is not what players newly introduced to pvp should face imo.

Not proof per se, but rather a bit of fun fact.

Ive got several accounts Im playing in pvp. Atm, Ive got those two that are pretty close together pip-wise (ruby t5).

First Account

Second Account

First screenshot is an account with quite a few matches played, I started with around 52% winrate on that account into season 2.

Second screenshot is an account that just hit rank20 about a week ago or so. So mmr is still volatile I guess.

Funny thing is… if I queue off primetime (like 7am for example) on the first account, I always get put on a horrible team. Nobody assists, nobody rotates… just horrible.
If I queue immediatly after loosing on first account with the second account… I get put on the team that just won against me. Assist, support, rotation… everything fine.

That first account dropped down to t3 ruby, after hitting t6 ruby about 2 weeks ago.

That second account hasnt even yet had a loosing streak longer than 2 matches, and that was in amber. This is actually the fourth account Im pushing through leagues, and I feared I ruined that account because I pretty much started seasons on that account with 2 losses. But apparently it didnt affect my mmr that badly.

/edit:
Btw, Im pretty much always running the same build and the same profession, on first account I started season with elementalist. But carry-potential of elementalist is rather not so good, if your team just runs straight on point and melts due to not having put up lifeforce etc. So I decided to go back to my roots and play necromancer, which is the profession Im best at anyway. I usually run something really glassy, but switch to a bit more tanky (carrion or mercenary) when I see 3 revs/necros on other team.

I’ve messed around with some toy models of the rating systems of the past seasons and have one that I think passes the smell test; I’ve attached some output for this season below. I can’t realistically reproduce the match-ups that come out of the matchmaker – there are big differences in how much people play that is hard to account for – so I just pick a player at random from the population, match him with a team according to the matchmaking rules, then pick out an opposing team from the remaining population at random.

The upshot of that is that the number of matches each player participates in is random and normally distributed, which lacks the skew of the real population but I think represents what the full time players experience pretty well.

A player’s ‘true’ skill is assumed to be fixed and unknown to the matchmaker and MMR algorithm (but is used to determine actual outcomes); MMR thus bounces around randomly as players win and lose. In the attached data set I began the season with really, incredibly terrible ratings accuracy (correlation coefficient between true skill and MMR = 0.544) but after the ~200 games per player in the model ratings line up really well with underlying skill (cor = 0.944).

The big takeaways, according to this toy model:

- Players pinned to the bottom of ruby are generally below average players – the median 0 pip ruby player is in roughly the 30th percentile of player skill in this simulation.

- There really isn’t any sort of Elo-hell. The premise of the original post doesn’t fit any sort of model I’ve looked at – if you’ve reaching ruby, your MMR cannot possibly be that bad, and will have recovered significantly from even pretty ridiculous initial conditions.

- The new matchmaking does exaggerate MMR. That is, above average players will have a higher rating than their ‘true’ rating, while below average players have a lower rating than their ‘true’ rating. I’m not 100% sure about what’s driving that effect, though I’m pretty sure it has to do with team formation.

If you have some specific circumstance you want me to model, let me know and I can plug it in – I’m only avoiding running a specific sim of the original post’s questions because they don’t make a lot of sense to me (show me that someone in low ruby with a terrible MMR will eventually have their MMR recover…while staying in low ruby?)

(I’d attach raw data or code fragments, but there’s a file type restriction in place – preferences on where to host a .Rdata dump?)

What players perceive as mmr hell, is basically the exaggeration effect. Somewhere along the way you get lucky, or unlucky, as an average mmr player. You dont even really notice that until you are firmly stuck on the wrong side of average and the quality of your teammates gets worse and worse.

And due to the exaggeration effect, the climb up is more difficult.

By dividing the playerbase in teams according their mmr and then NOT prioritizing similar ratings between the teams facing each other, anet caused the matchmaking system to fall into a state in that there’s a winner playerpool and a looser playerpool.

Idk how your simulations exactly work, but could you test what would happen if the worst dont get matched against the best, but rather the matchmaker tries to find two teams with similar mmr and match those two?
My guess is that the exaggeration effect wont be as bad, if you exclude matchups that are too skewed.

it think is easy to prove….send me the shines.
you can if in your pip range you have more team with less than you mmr then team above your mmr.
in this scenario you can go up even if you are bad.(this scenario is the situation after first match of the season, if you win you go up, till the match that you face a team of all go up player, there you test your mmr, if you win again up, if you lose possible hell)
in a 50/50 scenario, if you are better then other you win 8/15 (the case that you match a team in your mmr).
I don’t have any math on how far you can carry a team, so maybe you can win 10/15 or more.
ex.
Team A you 75% other 5% vs Team B 10%, you can win? maybe.
so in a 50/50 scenario yes, you can go up, slowly but yes.

the problem that this scenario is simply an utopia, you need to remove all pip safe spot, and run a season for an year or more, and have a lot of player.
other thin,g you basically have 1 fair match over 15 (less but make it simple), in that match is tested your mmr all the other match are unusefull, only a time loss.
so is quite obvious that s2 system is totally broken.

yes mathematically in s2 a single loss can put you in mmr hell.

here the example of 8 team the player base have 2 kind of skill good and bad.
the s2 matchmaking is fair and don’t voluntary match winner vs looser (i hope is fair but don’t know this).
after first match team with 3 good beat team with 2 good.
now winner pool have more good than loser, and for speed up matchmaking probably the first 2 team ready are 1 from loser pool and 1 from winner and so on.
so after 7 match 4 team are outside the range of other 4, and need to face.
here next match choose next mmr hell for these team.
in this way you have a good player that have lost all the matches, on the bottom of the ladder, and a bad player at top of the ladder that have win all the matches.

ok gw2 is little more complex with more player that don’t play the same number of matches and have a wide range of personal skill…but more or less you can be in this hell or you can be in the other side.

NeXeD.3042:

what facts have been provided?

the burden of truth should be on those that make the claims, people can post screenshots of game scores but they probably dont tell the whole story.

as much as alot of you want to believe the people winning are deluded into defending a flawed system one could argue the same for the other group, but what group is more likely to go to the forums to voice their disapproval?

people on these forums dont understand those complaining are most likely the vocal minority.

Glicko 2 is a system that was designed for world chess tournaments, counter strike and other games that have over a million players world wide. The system assesses player skill with a percentage of deviation that is considered reliable. The playing pieces in these games do not change over time. The rules of these games do not change over time.

Guild wars 2 doesn’t have a large fan base.
The playing pieces in the game often change.
The rules remain the same, but the objective change depending on which map you’re in.

If you take 2 players who have never touched gw2 and got them to play for 5 hours each and then pitted them against each other, we should see each player has a 50% chance to win or lose. If we make one player a corruption reaper with snap ground targeting and the other a sword/dagger thief, with experience and confidence we can say the necro is heavily favoured to win that 50%. We can also confidently say if the thief narrowly loses they have done well with a profession that isn’t meta and has a steepish learning curve to it. However the system will recognise the necro as better and the thief has worse.

The system itself cannot account for these details, therefore it is fair to assume match making will be imperfect. These imperfect moments are what is being called MMR hell..

The necro gets moved up and can’t hold there own anymore vs better players and the thief doesn’t progress at all and gets placed with others on low MMR which means most of their matches will be only slightly better quality than a hotjoin match.

MMR hell only disappears once you’re on a winning streak, playing meta only professions and builds. Which this unto itself defeats the entire purpose of having a league in the first place. If it were called meta build season 2, then I’d understand.

This is why tier 5 sapphire-tier 3 ruby is being referred to as MMR hell zone.

Too many builds have carried lack lustre players there, and too many ‘could be good’ players never got there because they ran builds outside the meta.

I have a legendary account, 3 ruby accounts on their way to legendary with only a handful lost games since season start, and one diamond account which was stuck in ruby for weeks until today.

I can carry, I can climb the ranks, and I can see the huge flaws with the current mechanics.

Everybody just saying “everything works, only baddies whining” obviously hasnt looked closely enough at the problems.

Open your eyes and look at the quality of a lot of players in ruby and diamond. Why are they there? How can they get to fourth division out of six, without learning not to rush on point when there’s a dh standing on it?

Ensign.2189:

I’ve messed around with some toy models of the rating systems of the past seasons and have one that I think passes the smell test; I’ve attached some output for this season below. I can’t realistically reproduce the match-ups that come out of the matchmaker – there are big differences in how much people play that is hard to account for – so I just pick a player at random from the population, match him with a team according to the matchmaking rules, then pick out an opposing team from the remaining population at random.

The upshot of that is that the number of matches each player participates in is random and normally distributed, which lacks the skew of the real population but I think represents what the full time players experience pretty well.

A player’s ‘true’ skill is assumed to be fixed and unknown to the matchmaker and MMR algorithm (but is used to determine actual outcomes); MMR thus bounces around randomly as players win and lose. In the attached data set I began the season with really, incredibly terrible ratings accuracy (correlation coefficient between true skill and MMR = 0.544) but after the ~200 games per player in the model ratings line up really well with underlying skill (cor = 0.944).

The big takeaways, according to this toy model:

- Players pinned to the bottom of ruby are generally below average players – the median 0 pip ruby player is in roughly the 30th percentile of player skill in this simulation.

- There really isn’t any sort of Elo-hell. The premise of the original post doesn’t fit any sort of model I’ve looked at – if you’ve reaching ruby, your MMR cannot possibly be that bad, and will have recovered significantly from even pretty ridiculous initial conditions.

- The new matchmaking does exaggerate MMR. That is, above average players will have a higher rating than their ‘true’ rating, while below average players have a lower rating than their ‘true’ rating. I’m not 100% sure about what’s driving that effect, though I’m pretty sure it has to do with team formation.

If you have some specific circumstance you want me to model, let me know and I can plug it in – I’m only avoiding running a specific sim of the original post’s questions because they don’t make a lot of sense to me (show me that someone in low ruby with a terrible MMR will eventually have their MMR recover…while staying in low ruby?)

(I’d attach raw data or code fragments, but there’s a file type restriction in place – preferences on where to host a .Rdata dump?)

Nice work but you do know that correlation doesn’t mean causation, right?? Because that’s what I am getting from your post. Here are a few things, you should add to your model.

- How is the meta impacting your data? (I.e. class imbalance and what not)
- New players started with higher MMR than folks who tried to cripple their way to legendary/diamond.
-What are your independent and dependent variables?

Well, only time i had lose streaks was when i played on classes that im not that great at. My longest being 9 lose streak, on a class i have 48% win rate on. Long ago when i played ranked more seriously on another class i had roughly 50% win rate. I stopped ranked for about a year and played other classes more casually and somehow i ended up learning another class way better than the other ones which i now have 54% win rate on. Thats also the class i flew all the way to diamond with, basically ONLY win streaks, but my mmr is indeed wrong when i play on that class, because all matches are too easy but i would have to play tons more matches to even raise my mmr since i have thousands of matches played. If i can get win streaks in a mmr level which is wrong for that particular class i play, i think people can dig their way out of mmr hell if they rotate correctly and try win the team fights. Because, if you wanna get out of mmr hell you NEED to win the team fights.

Celeras.4980:

Tenebria.7239:

Celeras.4980:

Give me an account that is in “MMR hell”, and I will be happy to stream how the games go and prove it is a result of your own lack of skill.

Other than that, there is no “mathematical” argument anybody can make since we cant prove our own MMR.

Of course higher skill takes you out of it. I just feel like you’re missing the point. You shouldn’t need that higher skill if the system was accurate and fair.

Andddd there it is. The ladder system isn’t “fair” because it rewards people with higher skill.

At least someone finally said it. It has nothing to do with matchmaking, and everything to do with people wanting to be rewarded for their own incompetence. Because its “fair”. GG.

The thread isn’t even about it, or skill, it’s about being stuck in elo/ruby hell what ever you call it and if you are even 50 % (skill level) of the ruby devision you won’t be able to climb since you are at the buttom, which will require way more “actual” skill then it would in fair circumstances.

sanctuary.1068:

Well, only time i had lose streaks was when i played on classes that im not that great at. My longest being 9 lose streak, on a class i have 48% win rate on. Long ago when i played ranked more seriously on another class i had roughly 50% win rate. I stopped ranked for about a year and played other classes more casually and somehow i ended up learning another class way better than the other ones which i now have 54% win rate on. Thats also the class i flew all the way to diamond with, basically ONLY win streaks, but my mmr is indeed wrong when i play on that class, because all matches are too easy but i would have to play tons more matches to even raise my mmr since i have thousands of matches played. If i can get win streaks in a mmr level which is wrong for that particular class i play, i think people can dig their way out of mmr hell if they rotate correctly and try win the team fights. Because, if you wanna get out of mmr hell you NEED to win the team fights.

Class MMR is no longer exist for Ranked. Only your Account MMR.

Raek.8504:

Celeras.4980:

Tenebria.7239:

Celeras.4980:

Give me an account that is in “MMR hell”, and I will be happy to stream how the games go and prove it is a result of your own lack of skill.

Other than that, there is no “mathematical” argument anybody can make since we cant prove our own MMR.

Of course higher skill takes you out of it. I just feel like you’re missing the point. You shouldn’t need that higher skill if the system was accurate and fair.

Andddd there it is. The ladder system isn’t “fair” because it rewards people with higher skill.

At least someone finally said it. It has nothing to do with matchmaking, and everything to do with people wanting to be rewarded for their own incompetence. Because its “fair”. GG.

The thread isn’t even about it, or skill, it’s about being stuck in elo/ruby hell what ever you call it and if you are even 50 % (skill level) of the ruby devision you won’t be able to climb since you are at the buttom, which will require way more “actual” skill then it would in fair circumstances.

sanctuary.1068:

Well, only time i had lose streaks was when i played on classes that im not that great at. My longest being 9 lose streak, on a class i have 48% win rate on. Long ago when i played ranked more seriously on another class i had roughly 50% win rate. I stopped ranked for about a year and played other classes more casually and somehow i ended up learning another class way better than the other ones which i now have 54% win rate on. Thats also the class i flew all the way to diamond with, basically ONLY win streaks, but my mmr is indeed wrong when i play on that class, because all matches are too easy but i would have to play tons more matches to even raise my mmr since i have thousands of matches played. If i can get win streaks in a mmr level which is wrong for that particular class i play, i think people can dig their way out of mmr hell if they rotate correctly and try win the team fights. Because, if you wanna get out of mmr hell you NEED to win the team fights.

Class MMR is no longer exist for Ranked. Only your Account MMR.

I know, thats not what i was saying. What im saying is, if you really are that great at a class you will win more than you lose despite being in mmr hell.

How can we prove any of those things you want when we don’t know anybodies MMR. Thats the problems with all these things you are making claims about x when we don’t know x. Only A-net knows X. Thats why I question things like MMR hell because we don’t know everyones X. Then you say something like someone with 50% skill. What does that even mean? When you lay out parameters that people can actually test maybe someone will take you up on it. It would actually be helpful if A-net would post about this and either give some kind of credence to idea or to totally shoot it down if they can. With my understanding how MMR works no I don’t believe there is MMR hell and why should I have to prove it to you although imo with the information we have currently available its pretty much impossible to prove either way!

brannigan.9831:

How can we prove any of those things you want when we don’t know anybodies MMR. Thats the problems with all these things you are making claims about x when we don’t know x. Thats why I question things like MMR hell because we don’t know everyones X. Is that really that hard to understand?

It’s simple simulation, you don’t need to know anybody’s MMR to simulate it.

sanctuary.1068:

Raek.8504:

Celeras.4980:

Tenebria.7239:

Celeras.4980:

Give me an account that is in “MMR hell”, and I will be happy to stream how the games go and prove it is a result of your own lack of skill.

Other than that, there is no “mathematical” argument anybody can make since we cant prove our own MMR.

Of course higher skill takes you out of it. I just feel like you’re missing the point. You shouldn’t need that higher skill if the system was accurate and fair.

Andddd there it is. The ladder system isn’t “fair” because it rewards people with higher skill.

At least someone finally said it. It has nothing to do with matchmaking, and everything to do with people wanting to be rewarded for their own incompetence. Because its “fair”. GG.

The thread isn’t even about it, or skill, it’s about being stuck in elo/ruby hell what ever you call it and if you are even 50 % (skill level) of the ruby devision you won’t be able to climb since you are at the buttom, which will require way more “actual” skill then it would in fair circumstances.

sanctuary.1068:

Well, only time i had lose streaks was when i played on classes that im not that great at. My longest being 9 lose streak, on a class i have 48% win rate on. Long ago when i played ranked more seriously on another class i had roughly 50% win rate. I stopped ranked for about a year and played other classes more casually and somehow i ended up learning another class way better than the other ones which i now have 54% win rate on. Thats also the class i flew all the way to diamond with, basically ONLY win streaks, but my mmr is indeed wrong when i play on that class, because all matches are too easy but i would have to play tons more matches to even raise my mmr since i have thousands of matches played. If i can get win streaks in a mmr level which is wrong for that particular class i play, i think people can dig their way out of mmr hell if they rotate correctly and try win the team fights. Because, if you wanna get out of mmr hell you NEED to win the team fights.

Class MMR is no longer exist for Ranked. Only your Account MMR.

I know, thats not what i was saying. What im saying is, if you really are that great at a class you will win more than you lose despite being in mmr hell.

I like how you probably never had an average or low MMR but are sure that you can amount for a majority of the outcome despite the 9 other players.

I only have over 5k hours on my ele and I can carry but never despite too many players doing their own business. I don’t think you realize the kind of conditions you can get yourself in just often enough to negate your progress and loose your appetite for playing after too many of it.

Raek.8504:

brannigan.9831:

How can we prove any of those things you want when we don’t know anybodies MMR. Thats the problems with all these things you are making claims about x when we don’t know x. Thats why I question things like MMR hell because we don’t know everyones X. Is that really that hard to understand?

It’s simple simulation, you don’t need to know anybody’s MMR to simulate it.

it is explain to me how we are going to do it. Simulate it if its so easy. Whats a skill level of 75% mean? If its so easy you do it. Maybe it is but maybe it isn’t I don’t claim to be some kind of mathematical genius. But if its so easy someone do the math to prove MMR hell. Its goes two ways buddy I never said I knew 100% there wasn’t a MMR Hell just that I feel like it probably doesn’t exist based on what A-net has told us about MMR but for you to say its easy to prove and then it exists and then you won’t prove it thats ridiculous. You will never win a debate among adults that way. Its for the people that claim 100% there is a MMR hell to prove it. I am getting sick of this nonsense. Also, saying the current system can be improved is not the same as saying there is MMR hell. Of course the system can be and should be improved.

brannigan.9831:

Raek.8504:

brannigan.9831:

How can we prove any of those things you want when we don’t know anybodies MMR. Thats the problems with all these things you are making claims about x when we don’t know x. Thats why I question things like MMR hell because we don’t know everyones X. Is that really that hard to understand?

It’s simple simulation, you don’t need to know anybody’s MMR to simulate it.

it is explain to me how we are going to do it. Simulate it if its so easy. Whats a skill level of 75% mean? If its so easy you do it. Maybe it is but maybe it isn’t I don’t claim to be some kind of mathematical genius. But if its so easy someone do the math to prove MMR hell. Its goes two ways buddy I never said I knew 100% there wasn’t a MMR Hell just that I feel like it probably doesn’t exist based on what A-net has told us about MMR but for you to say its easy to prove and then it exists and then you won’t prove it thats ridiculous. You will never win a debate among adults that way. Its for the people that claim 100% there is a MMR hell to prove it. I am getting sick of this nonsense.

Hm did you read whole OP post ? Basically MMR hell in chess, does not exist, it doesn’t exist in LoL or Dota or CS:GO. Because those games doesn’t artificially tune the MMR system.

We have grind system in conjunction with mmr system, where safe guards exist. It’s easy to assume that matches which should not happen by the nature of how real MMR system works will happen in GW2 system.

Basically the very existence of total blowout matches (30 loosing streak) is already a prove that MMR hell exists.

75 % ? It does mean that a player in best 25 % of his devision. His true level at least.

OP doesn’t want you to prove that MMR hell is real, he wants someone to prove the contery. But i m not even trying to win argument.