DQ Treatise :D

SergeantScuttler··44 votes·2,951 views
2016 Season Update

A pure solo player with rank X and a pure 5s player of the same rank don't represent the same skillset. They're vastly different experiences and don't belong on the same ladder in a competitive game. It would be similarly misguided to average out doubles and singles rankings for tennis in order to provide a unified ranking. Just like in DQ's case, this would only create an ambiguous ranking which fails to be a good indicator for either format.

For a ladder to be valid, it has to operate under controlled conditions which are consistent for all its members. DQ adds a myriad of new variables which greatly increase the difficulty in isolating individual skill while also dramatically increasing variance and leading to a feeling of reduced individual control in outcomes (for both directions). Worst of all, many of the variables are not only unequal but also unknowable (voice communication, smurfs boosting friends, well-practiced pre-game plans and strategies etc.). Unknowable variables are especially problematic because it's impossible to perfectly compensate for their absence/presence and their influence can be wildly different from game to game.

When you can't accurately compare players of the same ranking due to potentially huge queue configuration differences, you know the ranking system is broken at its core. These are not just matchmaking or queue time issues. This is about the very legitimacy of ranked and as long as such inconsistent experiences are aggregated on the same ladder, it will always fail in this dimension. Even the best voodoo math cannot change the fact that the fundamentals directly oppose key premises of a viable ladder and such attempts will always lack transparency for laymen anyway.

As ranked is further devalued, motivation to pursue the game seriously will dampen and the primary engine behind the game's growth will eventually stall. When people no longer respect and trust the ranked system, they'll be less driven by self-improvement through streams, guides, as well as watching LCS or absorbing other LoL content.

Forsaking ranked integrity in the hope of attracting more casuals is akin to plunging a dagger through your heart to cure high blood pressure. As the blood pressure first drops, it may seem that cardiovascular health is better but the cruel truth eventually manifests itself. Most of what makes LoL a vibrant game comes from its hardcore fans who make art, create websites, stream, preach to their friends, post on forums, etc. The casuals don't care all that much whether it's normals or ranked anyway and tend to follow the trendsetters from game to game. Even if ranked is more accessible to them, they may never become that "ideal ranked player" who buys a lot of skins and champions. It’s not worth losing the goodwill of the community and frustrating those who actually do care about ranked.

While SQ would be ideal, Riot seems unwilling to let their customers choose which queue will live so I’ve come up with a few compromises. The first would be to simply not allow anything beyond duo queue in Platinum and above. The most vociferous opposition to DQ comes from dedicated players in these ranks and larger premades also lengthen queue times. Since it’s generally accepted that anyone can reach Gold with some effort, allowing DQ and its distortions upon the ranked system only below Platinum is an easier pill to swallow.

Another idea would be to keep DQ but segregate out solo performance almost as if it was a separate queue. This would still be higher variance than a true SQ due to premades but at least all solo players would be exposed to the same variables over sufficient games. Unlike the current system, it would have internal validity and it also wouldn’t kill 4s like a solo ladder probably would.

49 Comments

SergeantScuttler11 votes

I realize emblems are a way to address some of the issues I posted but they're way too small of a step. They're gimmicks rather than an objective system where you can see your solo performance separately.

Besides not being retroactive, a player could simply DQ with his friends till he gets Platinum 5 or whatever desired rank and play solo from that point onwards to get the emblem. Since you can't see true MMR, it's open to abuse. They'd have to perform at ~Gold 5 level to drop out of Platinum. To others, it will appear that they reached Platinum through their own efforts even if their skill is unchanged. Moreover, it's too easy to get it. You only have to play the majority of your last 25 games solo to get it (which means you can even DQ 13 of them). It doesn't matter what total percentage of your ranked history those 25 games represent.

It also doesn't seem consistent with Riot's vision. Such an emblem actually discourages people from grouping up so that they don't lose it. Wouldn't it be better to just separate the performance and not put any additional strain on someone to maintain some arbitrary emblem?

Elikain1 votes

Emblems are there to identify you and others as a certain type of a player. The problem most people here faced is that they have no idea if someone got in a certain tier on their own or "they god boosted" by their premades.

And to quote one of the Rioters who responded to their news post on Rethinking Ranked Fives and Tuning Dynamic Queue:

##If you hold your own in a certain tier for 25 games, isn't it safe to assume you deserve to be in it?

As far as "the majority" problem goes, it's not 13/12 games and you get an emblem. It will probably be 80% of the games. That's 20 games. For the sole reason of 13/12 problem being 51/49% and that is open to abuse. So if the emblem system is tuned high, you'd actually have to play most of the games like this (and only a handful of others) to accurately represent your play style with the least bit of overreaching in others.

ofart2 votes

{quoted}

Emblems are there to identify you and others as a certain type of a player. The problem most people here faced is that they have no idea if someone got in a certain tier on their own or "they god boosted" by their premades.

And to quote one of the Rioters who responded to their news post on Rethinking Ranked Fives and Tuning Dynamic Queue:

As far as "the majority" problem goes, it's not 13/12 games and you get an emblem. It will probably be 80% of the games. That's 20 games. For the sole reason of 13/12 problem being 51/49% and that is open to abuse. So if the emblem system is tuned high, you'd actually have to play most of the games like this (and only a handful of others) to accurately represent your play style with the least bit of overreaching in others.

Even if it was 100%, it wouldn't change a thing due to the way the bottom tiers work. Emblems have basically no meaning if it's only the last 25 games rather than a total percentage. I think it's a misguided idea anyway. You shouldn't have to think of such silly things when deciding how to queue.

Elikain1 votes

How come emblems don't have any meaning? 25 games is quite a number of games to play. That's like saying that 10 promotional games don't have any meaning and are not accurate to place you in a certain Tier and Division. Sure, statistical analysis and accuracy is higher the more games you play but it has to be fair and reachable for everyone than working your ass off to change your player status.

These emblems don't affect matchmaking in any way so you don't have to pay them any heed when "deciding how to queue". It's simply a representation of your last 25 games which determines what type of a player you are.

Is it really going to matter if you come across a DQ emblem player and you're a SQ emblem player and now you can identify each other? Of course it won't.

##EDIT:

As for the sample size of 25, that's how Riot does their data. You need about 25-30 games to reach lvl 5 mastery and for champion data, they consider players with 25+ games on a specific champion across ELOs when they are making changes.

ofart2 votes

{quoted}

How come emblems don't have any meaning? 25 games is quite a number of games to play. That's like saying that 10 promotional games don't have any meaning and are not accurate to place you in a certain Tier and Division. Sure, statistical analysis and accuracy is higher the more games you play but it has to be fair and reachable for everyone than working your ass off to change your player status.

These emblems don't affect matchmaking in any way so you don't have to pay them any heed when "deciding how to queue". It's simply a representation of your last 25 games which determines what type of a player you are.

Is it really going to matter if you come across a DQ emblem player and you're a SQ emblem player and now you can identify each other? Of course it won't.

25 games really isn't that much for this game and considering many people have hundreds of ranked games by the end of the season, it's not all that representative. I assume your "10 promotional games" are a reference to placement games and yes, they absolutely aren't accurate for placing you in a specific tier. Even Challenger players can't get placed higher than Plat 1. 10 games is an insanely small sample for a game like this. They're just a starting point.

In any case, the biggest issue is that once you reach the bottom tier like S1 -> G5, you can perform far below it without dropping. You also don't have to play those games sequentially anyway. No joke, if you just got G5 (meaning normal MMR) and lose 70% of the following 25 games, you'll still be G5 after.

I also don't think they'd require more than 13/25 games. That could actually discourage players from using DQ so they don't lose their emblem. Considering they offer no SQ IP rewards but are almost bribing people with IP rewards for DQ, it seems like they would never do something which could take away from their new baby.

BloodyTompon4 votes

For the most part I agree and I can certainly see where you are coming from. However, me being anal, I have to pull you up on a few things.

For a ladder to be valid, it has to operate under controlled conditions which are consistent for all its members

This is incorrect. No ranking ladder operates under perfectly controlled conditions that are consistent and no ladder has a 1:1 correspondence to skill. Some of the reasons are pragmatics, some are philosophical. It's better to think of the rankings as an optimisation towards a particular goal. If your goal is to measure individual skill, then the SoloQ experience does it better than the DynamicQ, but it is incorrect to say that DynamicQ does not do it and SoloQ does.

Forsaking ranked integrity in the hope of attracting more casuals....

You've made a fair few assumptions about casuals and hardcore players here. What makes you think it's mostly those that play ranked (and of that small subset the even small subset that hate dynamicQ) that make art etc? That only hardcore players do ranked etc? What exactly do you consider hardcore and casual players?

Vociferous

Nice word btw.

Another idea would be to keep DQ but segregate out solo performance almost as if it was a separate queue.

Can you elaborate on this please?

PrometheS4 votes

{quoted}

For the most part I agree and I can certainly see where you are coming from. However, me being anal, I have to pull you up on a few things.

This is incorrect. No ranking ladder operates under perfectly controlled conditions that are consistent and no ladder has a 1:1 correspondence to skill. Some of the reasons are pragmatics, some are philosophical. It's better to think of the rankings as an optimisation towards a particular goal. If your goal is to measure individual skill, then the SoloQ experience does it better than the DynamicQ, but it is incorrect to say that DynamicQ does not do it and SoloQ does.

It's meaningless due to a lack of specificity because everything is clumped together. In other words, there is a lot of information but it's not possible to draw clear conclusions from it. It's not that it measures nothing; it's that what it measures isn't singled out. It's a bit like having a study with no control. You might have statistically relevant data and it might even be right but because you lack a control, you can't make any confident assertion about cause and effect.

BloodyTompon1 votes

It's meaningless due to a lack of specificity because everything is clumped together. In other words, there is a lot of information but it's not possible to draw clear conclusions from it

Aside from this being not quite true either , you've missed my point (again, I am being anal). Seat-belts on. Prepare for a terrible explanation.

The SoloQ ladder also clumped a whole lot of variables together. DynamicQ clumped a whole lot together. They both clump a whole lot together. The both are useful for ranking too. Neither is 'better' at ranking than the other and neither 'fail' at ranking. Dynamic Queue measures individual skill too, but SoloQ does better at ranking individual skill than DynamicQueue.

The premise that Dynamic cannot rank (individually or otherwise) is faulty. The reasoning behind it (consistency) is also unsound. There is not a ranking system in the world that mandates absolute consistency because every single game has variation. SoloQ doesn't force consistency more than Dynamic. It doesn't for more consistency, bar one thing: The number of people you can queue with. There is, however, still much inconsistency. The things listed in that paragraph can equally apply to SoloQ. In other words, if the reasoning being the ranking Dynamic Queue not being valid (experience consistency) was sound then SoloQ wouldn't be valid either.

You might have statistically relevant data and it might even be right but because you lack a control, you can't make any confident assertion about cause and effect.

Do you know what time it is Mr Wolf*? Lunch time Fun Fact Tangent Time! : You can actually do studies without controls. E.g., Smoking causing heart disease has no true control group. The tobacco companies used this fact in their defence in court : Instead of tobacco causing heart troubles it might be those that are prone to heart disease are more likely to smoke. To support this tobacco companies used two personality types: A laid back one with little stress that does not smoke or get heart disease and an over-achiever with high stress that smokes that does. Enter Type A personality (and Type B, but nobody ever gives a rat's about Type B :/ )

*Do you play that game in the US btw?

6KEMBE4ORBA3 votes

I see that you want to tickle your tongue...

Ultimately, premade groups have inherent advantage over solo players even with a voice-comm.

The SoloQ ladder also clumped a whole lot of variables together. DynamicQ clumped a whole lot together. They both clump a whole lot together. The both are useful for ranking too. Neither is 'better' at ranking than the other and neither 'fail' at ranking. Dynamic Queue measures individual skill too, but SoloQ does better at ranking individual skill than DynamicQueue.

if you're aiming for more inaccuracy, you can be pretty accurate at achieving that ---> DynamicQ in a nutshell.

Why should we introduce more variables that no one can keep track of/measure so as a result the system will need more iterations and make it harder to find/place you closer to your true ELO?

For the kicks out of it?

You want to play 20; 100; 200+ more games (no one can say how much more) just cause of this change?

When you have 2 systems - you choose the one that is more accurate, fair and dosn't render your ranked ladder meaningless.

Do you agree?

BloodyTompon1 votes

{quoted}

I see that you want to tickle your tongue...

How else does one procrastinate? :p

Ultimately groups have inherent advantage over solo players even with a voice-comm.

I'm not saying it does not have an inherent advantage. I'm not even saying that number of people using voice com in Dynamic and Solo are the same. I'm saying that the same reasoning to dismiss Dynamic Queue in the OP can be used to dismiss SoloQ.

if you're aiming for more inaccuracy, you can be pretty accurate at achieving that ---> DynamicQ in a nutshell.

Why should we introduce more variables that no one (can) keep track of or measure so as result the system will need more iterations to find/place you closer to your true ELO?

No, I'm not aiming for inaccuracy. All I'm saying is that the statement that DynamicQ is invalid is not correct. The only added variable from Solo to Dynamic can and is tracked: That variable being how many people can queue together. There are others that have been altered, but they had an equivalent in the SoloQ experience. E.g., SoloQ had a 'teamwork' variance the same way DynamicQueue does. The distribution and what have you are different in the DynamicQ system, but the variable itself is the same.

Why should we introduce more variables that no one (can) keep track of or measure so as result the system will need more iterations to find/place you closer to your true ELO?

This last sentence is interesting to me. True Elo? What do you mean by that? I think what you are referring to as true Elo is what I am referring to as individual skill? In which case, I'm sceptical that DynamicQueue needs more iterations to find that. Since Dynamic Queue is worse at measuring it, what seems more likely to me is the confidence interval is larger. You could argue that then the number of iterations needed increases to reduce the confidence interval. That's an entirely defensible position. I'd counter that by arguing that ones Elo rank tends to be a range rather than a point in real life (i.e., the Elo fluctuations would probably be larger => larger CI for the larger range). Disclaimer: It's been a few years since I've done higher level stats so I could be talking out of my arse for the last bit.

I do feel like we are essentially saying the same thing though. Both queues work, one is better for the particular job of ranking individual skill.

6KEMBE4ORBA2 votes

I wasn't talking about you, I was talking about the DynamicQ.

the only added variable from Solo to Dynamic can and is tracked: That variable being how many people can queue together.

That's wrong and it seems you're confusing some things/definitions. The variables (which are countless) are appearing/disappearing into existence (which itself leads to even more variability) the moment anyone goes with groups or another one leaves a group to play solo.

Now take into account all the possible queue configurations by DQ and imagine all the variables over the span of, say, 1 year.

It certainly isn't tracked/measured and actually it can not be tracked/measured by anyone.

  • SoloQ and DQ have a totally different "teamwork variance".

  • true ELO (which is always in fluctuation) means you're in or very close to your current skill level.


You're saying that since both teams have 5 players all is kinky and good ---> Wrong cause you're making unfair match ups by putting solos with/against various sized groups.

Essentially your whole argument is: "DQ works in some way cause the game ultimately is designed to reward the better players"

I've never said that it dosn't - but compared to the older system it's way worse in achieving the goals both systems aim at + it's unfair.

You agreed to the unfair part and that is a reason alone DQ should be removed from the ranked gamemode till it's heavily reworked, VOIP is introduced along with separate ranked ladders of SOLO/PARTY MMR and then maybe we can talk again.

Until then, as I've said, you're tickling your tongue and wasting both of our times while Riot's "solution" is giving out badges nearly 7 months into the ranked season...

BloodyTompon1 votes

That's wrong and it seems you're confusing some things/definitions. The variables (which are countless) are appearing/disappearing into existence (which itself leads to even more variability) the moment anyone goes with groups or another one leaves a group to play solo.

Yeah, that was worded horribly, sorry about that. The only parameter that has changed between the systems would have been a better way of phrasing it.

I still struggle to think of many variables with a non trivial impact that apply only to the solo case that do not or could not have a similar equivalent in dynamic. I mean we could go down the infinite arbitrary variables path, but that applies to both SoloQ and DynamicQ and they're trivial for the most part.

true ELO (which is always in fluctuation) means you're in or very close to your current skill level.

True elo means you're very close to you're skill ? What is skill? If go by this, DQ and SQ have different 'definitions' of that in a way. The true Elo for each respective system is likely to be different. In the previous post I operated under the assumption that the true Elo for SQ and DQ were the same, but that's unlikely to be the case in reality.

Essentially your whole argument is: "DQ works in some way cause the game ultimately is designed to reward the better players"

It does work. I still find it difficult to understand how people say it doesn't. It doesn't work as well as soloQ for ranking individual skill.

I've never said that it dosn't - but compared to the older system it's way worse in achieving the goals both systems aim at + it's unfair.

What are the goals of the system? I can guess you answer will be to rank people by individual skill, in which case we're in agreement. (I haven't said anything about fairness though. That's another can of worms)

Until then, as I've said, you're tickling your tongue and wasting both of our times while Riot's "solution" is giving out badges nearly 7 months into the ranked season...

For what it's worth I think that's horrible. While I'm not angry at the DynamicQ change myself, if I were I'd take that as a slap in the face. It's just so shallow.

Blanchimont2 votes

Dynamic Queue and Solo Queue teamwork is very different. In Dynamic Queue, you can have a large group of players that can play together on a regular basis. Solo Queue was limited to a maximum of two whereas Dynamic Queue does not have that limit. With just two, your impact on the overall team is quite low. Once you hit three, you now constitute a majority and things start to change. The more people you have on the team, the greater the more pointless the overall team's MMR is because the current system does not account for potential teamwork. A group that plays together often has better communication, timing, and can coordinate their objectives a lot better than a solo team of players. It's the same as taking 5 random Diamond tier players and asking them to play against CLG. Their MMR might be similar, but their teamwork is not.

BloodyTompon1 votes

{quoted}

The more people you have on the team, the greater the more pointless the overall team's MMR is because the current system does not account for potential teamwork.

It does in a way. The MMR of a team is adjusted based on the number of pre-mades. Boards were up in arms over it a while ago.

PrometheS4 votes

{quoted}

Aside from this being not quite true either , you've missed my point (again, I am being anal). Seat-belts on. Prepare for a terrible explanation.

The SoloQ ladder also clumped a whole lot of variables together. DynamicQ clumped a whole lot together. They both clump a whole lot together. The both are useful for ranking too. Neither is 'better' at ranking than the other and neither 'fail' at ranking. Dynamic Queue measures individual skill too, but SoloQ does better at ranking individual skill than DynamicQueue.

The premise that Dynamic cannot rank (individually or otherwise) is faulty. The reasoning behind it (consistency) is also unsound. There is not a ranking system in the world that mandates absolute consistency because every single game has variation. SoloQ doesn't force consistency more than Dynamic. It doesn't for more consistency, bar one thing: The number of people you can queue with. There is, however, still much inconsistency. The things listed in that paragraph can equally apply to SoloQ. In other words, if the reasoning being the ranking Dynamic Queue not being valid (experience consistency) was sound then SoloQ wouldn't be valid either.

Do you know what time it is Mr Wolf*? Lunch time Fun Fact Tangent Time! : You can actually do studies without controls. E.g., Smoking causing heart disease has no true control group. The tobacco companies used this fact in their defence in court : Instead of tobacco causing heart troubles it might be those that are prone to heart disease are more likely to smoke. To support this tobacco companies used two personality types: A laid back one with little stress that does not smoke or get heart disease and an over-achiever with high stress that smokes that does. Enter Type A personality (and Type B, but nobody ever gives a rat's about Type B :/ )

*Do you play that game in the US btw?

I understood your point perfectly, I just don't agree with it. The Solo/Duo system wasn't perfect but it was closer to a good reflection of individual skill. It WAS better as far as a valid ranking system because what it measured was more CONSISTENT and REPRODUCIBLE (key for a good ranking system). You'd also reach your true ranking with fewer games. Many of my friends who played primarily DQ could not reproduce their success on alternate accounts.

DQ isn't a good measurement of either solo or premade performance because it averages everything as explained by the OP. It's a very generic measurement with such wide ranges that it loses its predictive power. Let's say you want to find out how well someone performed in HS Biology and instead of being given that grade, you're given his total GPA from 1st grade to 12th. You may be able to make a slightly more educated guess but you still don't have your answer. Saying that neither DQ nor SQ are ideal is no refutation or argument.

And yes, a true SQ actually would force consistency and be very accurate with the caveat that you must play enough games to ride out the variance (just like in poker luck has a large impact for small samples but progressively smaller as you play more). In a true SQ, most of the consistency issues become less and less meaningful over time because all the other solo players are exposed to the same factors eventually (throws, akfs, trolls, bad comps. poor synergy etc.). xyzt=xyzt and x=x are both true statements. You can't judge a ranked system's ability to accurately assess skills over a small sample of games. First you have to think of a system that you know would, under ideal circumstances, be fair and reflective with sufficient games and then you work from there trying to reduce the number of games needed for a good approximation of skill. There will always be pragmatic elements which force you to sacrifice some accuracy. However, in a well designed ranking system, all that normally means is that you have to play more games to get there. See A.

All solo players would be in the same environment and if they could play an infinite number of games, their ranking would be a near perfect reflection of individual skill. The same cannot be said for DQ, which even with an infinite number of games would continue to be way too generalized in what it reflects to be particularly useful and would continue to be potentially misleading. For example, if a guy who plays 99.9% of his games premade in trios+ suddenly goes solo, his DQ ranking becomes an unreliable indicator despite an enormous number of games. You don't really know what you're getting. Again though, to say SQ has some issues too is no argument in favor of DQ. SQ at least has the foundation of a proper ranked system. I think everything you've brought up has directly or indirectly been addressed by the OP.

You can actually do studies without controls.

At no point did I claim that you couldn't and it's certainly easier. However, as you've explained, it's impossible to draw clear conclusions from them (just like DQ rankings). "Instead of tobacco causing heart troubles it might be those that are prone to heart disease are more likely to smoke." Studies without controls serve more as a beacon towards something worth further investigation than as actual proof.

..... I can't believe I'm using references but it was just too long. A: For example, let's say a game like LoL with SQ could provide you a ranking with 90% confidence over 25 games but it would come at the cost of hour long matchmaking. This isn't workable so they opt to give you a ranking with 60% confidence after 25 games but only having to wait 3 minutes for matchmaking. Although, as you get to 100 games, the confidence is within 10% for each system. In this hypothetical case, the second system would still be best despite being less accurate.

BloodyTompon1 votes

The Solo/Duo system wasn't perfect but it was closer to a good reflection of individual skill. It WAS better as far as a valid ranking system

Reading this sentence, I actually think we are saying the same thing for almost everything haha I'm not actually defending Dynamic queue, I'm pulling up the OP up for saying something that strictly speaking wasn't true: Dynamic Queue is valid, even if it's not as good at measuring individual skill. As I said, I'm being anal. (Relevant: https://xkcd.com/386/ )

You'd also reach your true ranking with fewer games

I'm not convinced that the true rank of solo and dynamic would be the same. In truth, I'm highly doubtful. I'm not arguing with this btw. If we assume the systems both have the same rank I think it an entirely defensible position.

DQ isn't a good measurement of either solo or premade performance because it averages everything as explained by the OP

It depends entirely on what you want to measure. I agree, DQ does worse than SoloQ in individual skill or groups of 5 than 5 man. You'll hear no argument from me on that point.

And yes, a true SQ actually would force consistency and be very accurate with the caveat that you must play enough games to ride out the variance

Yeah, I completely agree. I think we'd disagree that Dynamic does the same though? It's one extra variable being the number of people you're queued with. In the idealised infinite case, the variance in both systems are basically eliminated.

References

Did you link references? I can't see any? :( That makes me a sad panda. Or is the reference the example with confidence interval?

At no point did I claim that you couldn't and it's certainly easier

Oh, I didn't mean to imply such. I'm procrastinating today and thought it was a fun fact. Also you missed the most important part: Does NA play the child's game What's the time Mr Wolf?

PrometheS5 votes

I disagree that DQ is valid because it fails in specificity (terribly so), reproducibility and consistency.

I never claimed that the rank in DQ and SQ would be the same. When I wrote that you'd reach your true rank faster in SQ than DQ that's exclusive to the SQ ladder and what it's meant to measure.

DQ doesn't do the same thing even with infinite games and it was explained why many times in this thread. The extra variables in the form of queue configurations do NOT cancel out primarily because it's all on one ladder and it continues to have a constantly shifting definition from player to player. If you play an infinite number of games only solo while also playing an infinite number of games only as 4s on a separate account, those will still not be identical for most people yet they are averaged into a single ranking without this artificial separation.

Moreover, others aren't forced to match your queue configurations so that further clouds any meaning to be derived from the ranking and makes it hard to compare players on the ladder even with a massive number games. For example, Mark can play 35% solo, 30% trio 10% duo, 5% 4s and 20% 5s. Bob has the same ranking as Mark but plays 10% solo, 10% duo, 30% trio, 0% 4s and 50% 5s. All of these are allowed to exist on the same ladder and more games won't make these queue configurations equivalent between Mark and Bob. This is an inconsistency that isn't fixed with time, as opposed to most of SQ's problems.

If there was a ladder for each queue configuration (which isn't really workable), then you'd be right that it would be equally valid with infinite games.

BloodyTompon1 votes

I disagree that DQ is valid because it fails in specificity (terribly so), reproducibility and consistency.

Unless your suggesting that DQ has no (or negligible) correlation with skill, I don't see how you can defend that. Regardless of the problems it has (and I'll be the first to concede it has many), it still correlates to a non trivial degree.

When I wrote that you'd reach your true rank faster in SQ than DQ that's exclusive to the SQ ladder and what it's meant to measure.

Just to clarify, you're saying that you'll reach the true rank of SQ in SQ faster than in DQ, while also saying that they don't measure the same thing? You've operationalised individual skill as the true rank in SQ. By definition you're going to reach that faster in SQ. Unless I've misunderstood, this looks like a tautology.

DQ doesn't do the same thing even with infinite games and it was explained why many times in this thread. The extra variables in the form of queue configurations do NOT cancel out primarily because it's all on one ladder and it continues to have a constantly shifting definition from player to player. If you play an infinite number of games only solo while also playing an infinite number of games only as 4s on a separate account, those will still not be identical for most people yet they are averaged into a single ranking without this artificial separation.

In an infinite number of games, every configuration of DynamicQ is played an infinite number of times. The instance of one playing no solo, exactly 2 solo etc. have a probability of 0 in the theoretical case. It's unlikely to converge to the same value as in SQ, which is what it sounds like you're desiring for this.

Moreover, others aren't forced to match your queue configurations so that further clouds any meaning to be derived from the ranking and makes it hard to compare players on the ladder even with a massive number games. For example, Mark can play 35% solo, 30% trio 10% duo, 5% 4s and 20% 5s. Bob has the same ranking as Mark but plays 10% solo, 10% duo, 30% trio, 0% 4s and 50% 5s. All of these are allowed to exist on the same ladder and more games won't make these queue configurations equivalent between Mark and Bob. This is an inconsistency that isn't fixed with time, as opposed to most of SQ's problems.

That inconsistency isn't fixed because you've essentially defined to be such. You are only accepting individual skill as having any meaning and seemingly define individual skill as the output of soloQ. Is the issue you have that it's discrete? You can have discrete variables in the SoloQ. Mark plays 20% with voice comms. Bod plays 80% with voice comms. Still allowed to exist on the same queue with that, even though that is a discretely different experience. Similar to how soloQ is still able to output a correlation with individual skill despite this discrete variable, so to can DynamicQ with discrete cases.

You're basically a measure isn't valid because because a subset of it can be measured more accurately. (Aside from being wrong because there is still correlative power) More importantly they're not actually measuring the same thing. The test is what the test tests.

ofart3 votes

{quoted}

Unless your suggesting that DQ has no (or negligible) correlation with skill, I don't see how you can defend that. Regardless of the problems it has (and I'll be the first to concede it has many), it still correlates to a non trivial degree.

Just to clarify, you're saying that you'll reach the true rank of SQ in SQ faster than in DQ, while also saying that they don't measure the same thing? You've operationalised individual skill as the true rank in SQ. By definition you're going to reach that faster in SQ. Unless I've misunderstood, this looks like a tautology.

In an infinite number of games, every configuration of DynamicQ is played an infinite number of times. The instance of one playing no solo, exactly 2 solo etc. have a probability of 0 in the theoretical case. It's unlikely to converge to the same value as in SQ, which is what it sounds like you're desiring for this.

That inconsistency isn't fixed because you've essentially defined to be such. You are only accepting individual skill as having any meaning and seemingly define individual skill as the output of soloQ. Is the issue you have that it's discrete? You can have discrete variables in the SoloQ. Mark plays 20% with voice comms. Bod plays 80% with voice comms. Still allowed to exist on the same queue with that, even though that is a discretely different experience. Similar to how soloQ is still able to output a correlation with individual skill despite this discrete variable, so to can DynamicQ with discrete cases.

You're basically a measure isn't valid because because a subset of it can be measured more accurately. (Aside from being wrong because there is still correlative power) More importantly they're not actually measuring the same thing. The test is what the test tests.

It legit feels like you're just arguing for the sake of arguing. You're ignoring or constantly misunderstanding key points posted and then misrepresent what others reply. For instance, no one said that individual skill is the only thing having meaning but DQ doesn't accurately represent anything on the spectrum. It doesn't even represent premade team skill well because SQ is included.

Let's say I play half my games SQ and perform absolutely awful. The rest of my games are played with my 3 other friends and I perform amazingly due to our great synergy. We win almost every game and perform as well as Diamonds. Unfortunately, since my SQ is awful, I'm still low Gold. This ranking isn't a good representation of anything and people wouldn't know what I played premade without digging through all my game history manually - even then you wouldn't know how much of my total ranked that is. This ranking would be a big overestimation of my SQ skill and a big underestimation of my premade performance. Just because it still has a correlation to skill doesn't make it a valid ranking system. Validity is a way more rigorous concept than that.

An infinite number of games wouldn't suddenly make all the queue configurations for every player on the ladder identical in distribution like 20% for each queue configuration. The proportions can still vary drastically so even as a general ranking, it wouldn't converge to bring consistency for everyone. It's like having 5 hoses with neverending streams but different water pressures. The practical scenarios are even more ridiculous. Some players may never play 5s or 4s while others may never play solo. The only way DQ makes theoretical sense and would be internally consistent for a ladder system is if everyone had identical queue variation proportions for everything. That's not realistic in the slightest.

BTW: I suggest you check the OP's link if you don't understand why variance is higher in a system with more variables which we can't even fully understand. It's like how playing large poker tournaments will require a bigger sample size to reach your true ROI for them compared to those with only 6 players.

Also, I've got to say that your usage of "you're" and "your" tilts me. So for my revenge, I will give you the equally tilting xD.

BloodyTompon1 votes

{quoted}

It legit feels like you're just arguing for the sake of arguing.

Lol. I feel like the same applies from my perspective. I've put it down to this not being the best forum for this kind of thing. If you feel i'm ignoring parts, by all means call me up on it.

For instance, no one said that individual skill is the only thing having meaning but DQ doesn't accurately represent anything on the spectrum. It doesn't even represent premade team skill well because SQ is included. . Just because it still has a correlation to skill doesn't make it a valid ranking system. Validity is a way more rigorous concept than that.

if it doesn't represent anything on the spectrum, does't that mean it has no correlation?

Honing in on your second part. Yeah, it kind of does in this context. In this case it is of degree and not type - That is validity a continuum not a binary yes/no. Validity is much murkier than people might expect. Depends on peoples background because it does vary somewhat between domains and context.

Consider it another way. You have a system that correlates with skill. It can place them in an order that roughly corresponds with that skill. You place people roughly in order of skill. (The previous three sentences are equivalent statements) You've ranked them roughly in terms of skill. You could conceptualise correlation as the degree to which something can be ordered or ranked.

An infinite number of games wouldn't suddenly make all the queue configurations for every player on the ladder identical in distribution like 20% for each queue configuration. The proportions can still vary drastically so even as a general ranking, it wouldn't converge to bring consistency for everyone.

The infinite number of games only applies to one person and refers to a single person converging to their true rank. The particular distribution of the discrete configurations won't impinge on that. Having said that, it could be extended to everybody. Similarly for the SQ with voice comms situation.

... internally consistent for a ladder system is if everyone had identical queue variation proportions for everything. That's not realistic in the slightest.

Internally consistent with what? To what degree?

Does it even matter? I'd say not really. Ordering is well defined. The difficulty, difference and most likely the disagreement is in skill. That's a fuzzy concept at best. The two skill measurements that we have are individual skill and 5 man skill. But DQ isn't trying to be solely one of those, it's a more like a skill construct of the sum so an internal consistency test is pointless: There's no existing metric to compare it to.

So now we've come the impasse. Given correlation is the ease of ranking something it comes down to one thing or two things. Does DQ correlate with skill? Is the correlation strong enough? Absolutely. It was strong enough to be useful and that's one of the best forms of validity you'll get.

BTW: I suggest you check the OP's link if you don't understand why variance is higher in a system with more variables which we can't even fully understand.

I never said there were fewer variables or that the variance is less. If I did I either misspoke, didn't explain myself well or am being misunderstood. I completely accept both that there fewer variables and less variance in SQ.

ofart2 votes

{quoted}

if it doesn't represent anything on the spectrum, does't that mean it has no correlation?

You're a sentence into the content of the response and are already misrepresenting a post which was meant to correct other mischaracterizations. The crazy thing is you actually have the quote included and still somehow managed to type all that. I've come across poor reading comprehension but this just feels completely futile. Even what you were trying to contradict isn't pertinent to the primary point. You're missing the forest for the trees.

The rest of your response rambles a lot and continues in the same pattern where you just don't seem to grasp key points or mischaracterize what is written. Some of these points have been repeated many times in different ways and with different examples. Even the questions you pose have been amply examined. You need to carefully read what you're replying to or it's just impossible to have a discussion this way. It's exasperating for me so maybe someone else can help you.

Here is your quote again:

For instance, no one said that individual skill is the only thing having meaning but DQ doesn't accurately represent anything on the spectrum. It doesn't even represent premade team skill well because SQ is included. . Just because it still has a correlation to skill doesn't make it a valid ranking system. Validity is a way more rigorous concept than that.

BTW: When someone says something is meaningless in colloquial language, it doesn't actually mean that it's utterly devoid of meaning. It's the suggestion that it doesn't do a good job of representing what is expected or isn't useful in a certain context. For example, if Pobelter shows up to a running race and proclaims his 200 IQ, that would seem meaningless to most. And yes, I realize he was just joking on stream.

I'll give you a sample response with my BloodyTompon simulator version 0.1 set only at 10% mischaracterization, 50% rambling and 75% missing the point:

"IQ tests don't actually have a 1:1 correspondence with each other and become increasingly inaccurate the higher you go. Most don't even measure very high and the sample size at those levels is too small to be useful. They also cover only a tiny subset of intelligence and are primarily a reflection of speed. It's important to ask "What is intelligence?". Different cultures emphasize different things. For example, seafaring nations may place special value on navigational ability.

The new school of thought focuses on a more comprehensive understanding of a person's abilities as elucidated by Howard Gardner's "Theory of Multiple Intelligences." EQ, which reflects social intelligence, work habits, perseverance etc. seems to have a higher correlation with positive real world outcomes than just the "potential" an IQ test can hint at.

For a 200 traditional IQ to come about, that suggests DNA replication made no grave errors for cognitive and nervous system development. If that's true, it's possible that other organs had fewer replication errors than normal and it may mean his muscles might perform better than average. Since IQ isn't purely hereditary and 200 is far outside the norm, he may also have had a healthier lifestyle than the norm with more focus on studying. Good sleep, diet and exercise would all help muscle development too. Moreover, the conscientiousness developed through intense study might make it easier for him to apply willpower in other directions, such as improving his running performance and knowing how to pace oneself while applying perfect technique. It's thus possible that his 200 IQ could be a relevant factor and correlate with performance. We might even be able to order the likely place of each runner by their IQ when we don't have other data.

It's also important to note that Pobelter didn't claim his 200 IQ was from the traditional tests. Perhaps he was referring to his motor IQ. Someone with otherworldly motor coordination would be better at running on average. It's misleading to suggest that everyone who refers to IQ is talking about only one measure. We should take all data into account and Pobelter's 200 IQ is nothing to scoff at when determining the outcome of a running race. I think I've rebutted the main point which suggests that IQ and running performance have zero correlation.

In fact, Riot probably isn't using enough of the data well. In games like StarCraft, EPM (Effective Actions Per Minute) tends to correlate with ranking. This is a measure that may well apply to League too and it would transcend everything in the DQ vs SQ debate. It may be worth considering ranking players by EPM in some way if this correlation holds. "

69982196_DEL3 votes

Honestly dude I don't care about reading that wall of text. I'm tired of arguing about the subject, Riot's bs has gone too far to make wanna try. A fucking emblem, a cosmetic is suppossed to make want to play DQ?

Hello I am Bird2 votes

It's supposed to be the starting steps until they figure out wtf else to do while keeping Dynamic Queue

69982196_DEL2 votes

Of course, let's wait and see if season 7 will be fucked aswell.

dbddbs3 votes

Well said OP. You've addressed a lot of what I dislike about DQ in a much more beautiful way than I ever could.

The Yetii Rider2 votes

The solo performance is shown through the champion mastery system. No one will earn that S but you. If you have Mastery 7 on a champion, team work didn't give you that.

ofart7 votes

[{quoted}](name=The Yetii Rider,realm=NA,application-id=3ErqAdtq,discussion-id=bcGkE8t6,comment-id=0001,timestamp=2016-06-18T00:02:14.038+0000)

The solo performance is shown through the champion mastery system. No one will earn that S but you. If you have Mastery 7 on a champion, team work didn't give you that.

Mastery is mostly a matter of time invested rather than a direct representation of skill. Grades also aren't all that meaningful. I've seen Mastery 7 Lees with 50+ games in ranked and ~45% winrates. While fun, the mastery system is mostly glitz to show off your favorite champions. When I see a Mastery 7 Azir, I don't think "oh wow this guy must be a god at Azir". I just think "oh cool this guy likes Azir and has a lot of experience on him."

macromite1 votes

first line, replace mastery with rank, and it is still a valid statement.

Dextix LT4 votes

No, because many people have risen with low ammount of games into high divisions because they are good.

And many people have 1k games yet cant get out of bronze.

Your statement, is false.

macromite1 votes

no, you just don't look at the evidence properly. of course, in that its a waste of my time to continue this, just know that you wrong.

Dextix LT1 votes

What evidence? You have given none.

macromite1 votes

The evidence is apparent, once you know what you're looking for.

Dextix LT1 votes

Then give it, or point to it?

macromite1 votes

I cannot force one to open their own eyes.

BloodyTompon1 votes

You're both kind of right imo. The Mastery system does tend to measure time investment much more strongly than ranked does, but the nature of neading S does give it some correlation of individual skill. It'll be interesting to see where Riot heads to with future updates on this (and what they consider to be the problem space they want to solve with it). When they first introduced Dynamic Queue I thought they were heading towards making the mastery a replacement of individual rank. However, they could just as easily have the goal of alleviating the feeling of ranked stagnation by making Mastery akin to level progression.

DemonofDota2 votes

Read the first few lines, didn't need to read any further. Good post. Couldn't agree more. +1

Karn Bishop2 votes

I am tired of trying ot figure out if i can't play certain staple champions because they do not work vs pre mades at all, I am tired of new emblem rewards that are abusused easily, 25 games? really? why not 75% of all games must be solo to keep this icon and dynamic players get no icon because we solo players don't get extra chests, keys or ip bonuses so why do they get a chance to earn our icon?

PrometheS1 votes

[{quoted}](name=Karn Bishop,realm=NA,application-id=3ErqAdtq,discussion-id=bcGkE8t6,comment-id=0005,timestamp=2016-06-18T02:30:49.338+0000)

I am tired of trying ot figure out if i can't play certain staple champions because they do not work vs pre mades at all, I am tired of new emblem rewards that are abusused easily, 25 games? really? why not 75% of all games must be solo to keep this icon and dynamic players get no icon because we solo players don't get extra chests, keys or ip bonuses so why do they get a chance to earn our icon?

I can tell you that Nocturne is pretty ineffective vs premades. When I played ranked with 3 others, it was pretty painful trying to make him work in most comps. He works fine in SQ though.

Cavemantero2 votes

its k, we got Emblems [slayer-jinx-wink]

Nimble Manta2 votes

It would be similarly misguided to average out doubles and singles rankings for tennis in order to provide a unified ranking. Just like in DQ's case, this would only create an ambiguous ranking which fails to be a good indicator for either format.

False analogy. A better analogy would be ranking a doubles team that practices together in the same ladder as a doubles team that doesn't practice together... which is what happens so yeah.

Your analogy would be appropraite if riot put twisted treelines and summoners rift on the same ladder.

PrometheS3 votes

[{quoted}](name=Nimble Manta,realm=NA,application-id=3ErqAdtq,discussion-id=bcGkE8t6,comment-id=0008,timestamp=2016-06-18T03:36:36.861+0000)

False analogy. A better analogy would be ranking a doubles team that practices together in the same ladder as a doubles team that doesn't practice together... which is what happens so yeah.

Your analogy would be appropraite if riot put twisted treelines and summoners rift on the same ladder.

I'd argue that that the difference is large enough between a premade of 4s compared to solo play for the analogy to work and you still play on the same court too. An analogy just conveys an idea though, it's not supposed to be equal in all respects. The tennis analogy is actually less extreme since you won't ever be pinning 2 tennis players vs 1 player + a random. Everyone is premade or solo and matched for equivalent configurations. Thus, even if tennis rankings were averaged between doubles and singles, it would at least not have that queue configuration distortion in matchmaking (like 5 solos vs a trio in league even if they try to adjust for it).

Nimble Manta1 votes

As a tennis player, doubles is pretty much an entirely seperate game. It's completely incomparable to dynamic que because a 5 man que and 5 solo players are still playing the same game. It's just one side as a communication advantage and the other as an mmr advantage, which can throw things off.

PrometheS2 votes

[{quoted}](name=Nimble Manta,realm=NA,application-id=3ErqAdtq,discussion-id=bcGkE8t6,comment-id=000800000000,timestamp=2016-06-18T04:29:14.066+0000)

As a tennis player, doubles is pretty much an entirely seperate game. It's completely incomparable to dynamic que because a 5 man que and 5 solo players are still playing the same game. It's just one side as a communication advantage and the other as an mmr advantage, which can throw things off.

Playing solo vs 5 isn't the same game either. Just like tennis, you may be on the same court but it's very different and beyond just elementary communication. You can't really have in depth plans as a solo player (at least if you expect team cooperation) or do many lane swaps with support invades like in LCS and fast side turret taking. It's night and day. It's literally a different starting configuration for the map and lanes. If you watched LCS, the lane swap I'm talking about with fast side-turrets taken was done universally. That's essentially where the game started since everyone was doing this due to reasons which aren't pertinent here. That's not the same game anymore.

It's hard to even come up with perfect analogies for this since it's not something that any sport would ever even consider. The teams which play against each other are all pre-selected with equivalent configurations. Whether you agree or not, the gist of the analogy is quite clear. You have many different queue configurations all on the same ladder under the same ranking and no clue what solo performance is despite allowing people to start a game solo.

6KEMBE4ORBA2 votes

The analogy can kinda work cause we still have solos alongside (against) duos in the same court.

I get what you're saying - there are 2 teams in the court but the impact that the enemy premade has and the ways they can push their inherent advantage (in LoL) over a solo player and completely nulify his performance are not to be taken lightly.

And I believe the OP was talking solely about the meaning of ranks - in the case of which - the analogy is spot on.

General Esdeath 1 votes

That's great! Too bad they're still on my team wondering how they lost as Riven vs Garen.

Dealth1 votes

is anyone reading this? we already know DQ sucks. Riot knows too.

MonkeyMadness1 votes

I have a better compromise: bring back soloQ or we Riot...that was a pun...I'm very punny... . . . . . . . . Mordekaiser hue

Blanchimont1 votes

Even reaching Gold is much more difficult under the Dynamic Queue system because of the ability to queue with players above or below your tier. It is possible for a Silver + Gold duo on your team to have to face a Gold + Platinum duo on the opposite end during a promotion series. This happened to me twice so far and the feeling is awful when you realize that in order to advance into Gold, I have to defeat a Platinum tier player.

I feel that Dynamic Queue is better off as a replacement for Ranked 5v5 than Solo Queue. It's definitely not a solution for both, but perhaps if they set up something like an in game voice system, Dynamic Queue could serve as a way to help people find a ranked team.

More from Gameplay

More by SergeantScuttler