Why no one likes playing ranked for rating.

[SlippyCheeze.5483]

> @"Chuck.2864" said:

> > @"Gwaihir.1745" said:

> > Actually I like people who ruin my placement games. Climbing to plat 1 again is easy, fast rewards gain before the fight to climb begins. Bronze - plat 1 is very easy to climb.

>

> I call kitten. The more I play PvP over the seasons, the more I begin to realize the importance of the placement matches, they are literally the 10 most important games you will play all season because getting significantly higher from where you were placed is incredibly difficult unless you were literally boosted and ended up with like 200 rating more than you really should have, then you will probably fall quickly down the divisions.

>

> This is probably because with the small population of PvP whether you're high Gold 2 or Plat 1 doesn't really matter, you'll end up in the same matchmaking pool. So as far as the MM system is concerned, a G2 player playing with Plat 1 players is playing within his skill level, ergo he will maintain ~50% winrate which you cannot climb with. For reference I placed in Plat 1-2 and stayed there the past two seasons with a ~50% winrate (positive). This season I had terrible placements including DC's, afkers, some meme builds, went 5-5 and was placed in Gold 2. Guess what? It's a 50% winrate again with around 65-70 matches already played this season and it's basically impossible to climb back to Plat. Every time I make gains I get hit with a 5 game loss streak.

 

A 50 percent **expected** win and loss rate will keep your MMR stable, not cause it to move, as long as you actually deliver the results. Even a streak of five **expected** losses in a row will fairly quickly bring you back up to where you should be once you get back to winning.

 

The only case you need a substantially higher win rate is where you frequently lose matches that you should have been able to win. If your experience is that you can't climb to platinum if you dip too low, the problem is more likely that you can't play at platinum level, not that the initial placement is the most significant thing overall.

 

PS: placement matches use the same MMR adjustment method as normal games.

[Gwaihir.1745]

> @"Chuck.2864" said:

> > @"Gwaihir.1745" said:

> > Actually I like people who ruin my placement games. Climbing to plat 1 again is easy, fast rewards gain before the fight to climb begins. Bronze - plat 1 is very easy to climb.

>

> I call kitten. The more I play PvP over the seasons, the more I begin to realize the importance of the placement matches, they are literally the 10 most important games you will play all season because getting significantly higher from where you were placed is incredibly difficult unless you were literally boosted and ended up with like 200 rating more than you really should have, then you will probably fall quickly down the divisions.

>

> This is probably because with the small population of PvP whether you're high Gold 2 or Plat 1 doesn't really matter, you'll end up in the same matchmaking pool. So as far as the MM system is concerned, a G2 player playing with Plat 1 players is playing within his skill level, ergo he will maintain ~50% winrate which you cannot climb with. For reference I placed in Plat 1-2 and stayed there the past two seasons with a ~50% winrate (positive). This season I had terrible placements including DC's, afkers, some meme builds, went 5-5 and was placed in Gold 2. Guess what? It's a 50% winrate again with around 65-70 matches already played this season and it's basically impossible to climb back to Plat. Every time I make gains I get hit with a 5 game loss streak.

 

You can't call bullshit on my experience on the sole grounds that your experience is different. I usually skip a season and play every other season which effectively removes my mmr history. So I usually place silver 2-3 and have climbed to gold 3 in under 100 games every season I play, peaking at ~1500-50sr when I care to play well.

[Kathul.1280]

> @"SlippyCheeze.5483" said:

> > @"Chuck.2864" said:

> > > @"Gwaihir.1745" said:

> > > Actually I like people who ruin my placement games. Climbing to plat 1 again is easy, fast rewards gain before the fight to climb begins. Bronze - plat 1 is very easy to climb.

> >

> > I call kitten. The more I play PvP over the seasons, the more I begin to realize the importance of the placement matches, they are literally the 10 most important games you will play all season because getting significantly higher from where you were placed is incredibly difficult unless you were literally boosted and ended up with like 200 rating more than you really should have, then you will probably fall quickly down the divisions.

> >

> > This is probably because with the small population of PvP whether you're high Gold 2 or Plat 1 doesn't really matter, you'll end up in the same matchmaking pool. So as far as the MM system is concerned, a G2 player playing with Plat 1 players is playing within his skill level, ergo he will maintain ~50% winrate which you cannot climb with. For reference I placed in Plat 1-2 and stayed there the past two seasons with a ~50% winrate (positive). This season I had terrible placements including DC's, afkers, some meme builds, went 5-5 and was placed in Gold 2. Guess what? It's a 50% winrate again with around 65-70 matches already played this season and it's basically impossible to climb back to Plat. Every time I make gains I get hit with a 5 game loss streak.

>

> A 50 percent **expected** win and loss rate will keep your MMR stable, not cause it to move, as long as you actually deliver the results. Even a streak of five **expected** losses in a row will fairly quickly bring you back up to where you should be once you get back to winning.

>

> The only case you need a substantially higher win rate is where you frequently lose matches that you should have been able to win. If your experience is that you can't climb to platinum if you dip too low, the problem is more likely that you can't play at platinum level, not that the initial placement is the most significant thing overall.

>

> PS: placement matches use the same MMR adjustment method as normal games.

 

 

[Kathul.1280]

IF wins and losses had the same positive or negative point loss then more wins would result in a higher rating.

 

YET losing typically results in more points lost than a win can gain overall with this glicko algorithm. I have never gained 33 points for a win. A Static point system for wins or losses would make for a more competitive and easier to understand system of rating. 1 win = 5 points. 1 loss= -5 points. More wins = more points, etc.

 

The NHL has a TEAM, and SOLO point system for an entire season that they can rate the TEAMS and Individual players on. Using a STATIC point system for both. Which by the way is collated by humans, and not and old Chess tournament system setup for 1v1 stats. (Elo, btw that is a persons last name) .

 

The math and systems for a competitive 5v5 system are in place, WORLDWIDE and used in the OLYMPICS !!

 

No individual player loses more than 2 points per game , win or loss. Based off of individual performance, which the game keeps track of ( i.e. most kills, most defense, most offense etc)

 

You want competition and ratings in a "competitive " environment well, we ALL have several examples. Anet needs to have a pint and a discussion.

[Jalal.6783]

the ranking system balance is a joke, plat 3 gets put with golds and once you cross plat 2, your rating earned gets reduced by 20%-50%and your points lost increase by 50%. Aint nobody got time for that

[Marxx.5021]

Placement matches are a joke. Playing plat 1 last season I end up in low gold. Basically all matches have been decided by afk player in the end.

[Jalal.6783]

 

 

> @"Poelala.2830" said:

> Balance aside, because that's a whole different post, the fact of rating loss and gain discrepancy being so egregious is directly discouraging players at best, and enforcing a gambler's fallacy at worse (and most common). When the scenario of a loss's absolute value being greater than or equal to the value of two wins, this means that a 66% win rate is *discouraged* and the only way to move up the ladder is with at LEAST a 75% win rate. These numbers are arrived by 2 wins (which can have the value of 8, in my case) do just as much good as one loss (equaling -16 in my case) does. This is 2 wins out of 3, or a 66% win rate. If you win 3 times out of four (75%), this is the ONLY way to rise up on the board.

>

> This isn't impossible, but highly improbable. Out of the three days this season has been out, I've seen at least 20 players in the top 5 lose rating as they play more. One of my friends were #1 and TWO LOSSES brought him down 70 rating. A popular twitch streamer lost 44 points in one game, and has a less than 75% win rate in the past 9 games and has lost rating because of this. When an unrealistic standard is enforced for god knows what reason, where it seems like playing LESS games is the most beneficial for you, guess what the player base will do? Play less games. The most awful fact of this is the fact that a 50% win rate will result in you being in bronze.

>

> How can you ban and dishonor win-traders when you are creating an environment where their existence is almost necessitated?

 

I'm in the same boat, i constantly hover between plat 1 and 2 because of this punishing system. Getting to plat 2 is easy enough and wins and losses and pretty much even rating gained/lost but as soon as 1600 comes by and with each successive win, the matchmaking system severely punishes me for a loss and barely rewards me with a win. It feels as though once you reach plat 2, the system puts you in a position that says,"Hey, you're plat 2 right? We expect to hard carry players much lower than you and if you cant, then you just aren't cut out for this division." It's rage inducing and has caused me to limit my playtime much more than i would like. I love this game and would love to play 15-20 matches a day but when it feels like i'm being punished for it, i usually end up playing around half as much before i get tilted off by the system.

[Chuck.2864]

> @"SlippyCheeze.5483" said:

> > @"Chuck.2864" said:

> > > @"Gwaihir.1745" said:

> > > Actually I like people who ruin my placement games. Climbing to plat 1 again is easy, fast rewards gain before the fight to climb begins. Bronze - plat 1 is very easy to climb.

> >

> > I call kitten. The more I play PvP over the seasons, the more I begin to realize the importance of the placement matches, they are literally the 10 most important games you will play all season because getting significantly higher from where you were placed is incredibly difficult unless you were literally boosted and ended up with like 200 rating more than you really should have, then you will probably fall quickly down the divisions.

> >

> > This is probably because with the small population of PvP whether you're high Gold 2 or Plat 1 doesn't really matter, you'll end up in the same matchmaking pool. So as far as the MM system is concerned, a G2 player playing with Plat 1 players is playing within his skill level, ergo he will maintain ~50% winrate which you cannot climb with. For reference I placed in Plat 1-2 and stayed there the past two seasons with a ~50% winrate (positive). This season I had terrible placements including DC's, afkers, some meme builds, went 5-5 and was placed in Gold 2. Guess what? It's a 50% winrate again with around 65-70 matches already played this season and it's basically impossible to climb back to Plat. Every time I make gains I get hit with a 5 game loss streak.

>

> A 50 percent **expected** win and loss rate will keep your MMR stable, not cause it to move, as long as you actually deliver the results. Even a streak of five **expected** losses in a row will fairly quickly bring you back up to where you should be once you get back to winning.

>

> The only case you need a substantially higher win rate is where you frequently lose matches that you should have been able to win. If your experience is that you can't climb to platinum if you dip too low, the problem is more likely that you can't play at platinum level, not that the initial placement is the most significant thing overall.

>

> PS: placement matches use the same MMR adjustment method as normal games.

 

What exactly is "Platinum level" though? That's exactly my point. In 2 seasons I was able to stay in Plat 1 without any issue (where according to you I can't play at "that level", and so if the MM is perfectly fine I should have fell down), but suddenly 2 weeks later I'm a Gold 2 player rather than a "Plat player"? Maybe I'm not even a G3 player anymore because I sometimes dip back into G2 when on a loss streak?

 

I mean this is what I'm saying. "Platinum level" is meaningless and arbitrary because I'm playing with people in Plat 1 anyway because this is what MM can do with the low population, so if Plat 1 is indeed my ceiling I have no way to actually get back to it because I happened to have DC"s and otherwise bad games in my placements, so gg. Wherever you place is mostly where you will stay unless you were either incredibly lucky\boosted or incredibly unlucky and ended up in Silver when you should be Plat. If it's fine margins though, i.e G3-Plat 1, you will find it super difficult to make that climb. It's time ANet recognize these divisions are arbitrary and almost meaningless with the current population and cut it down to maybe 3 divisions.

 

Answer me this - do you actually consider someone with 1530 rating to be playing at a different level to someone with 1470 rating because one is in Gold and the other in Plat? Even the MM system doesn't believe that, but the latter player might never break that 1500 barrier whilst consistently being matchmade with the former player

[Arheundel.6451]

> @"SlippyCheeze.5483" said:

> > @"Arheundel.6451" said:

> > > @"SlippyCheeze.5483" said:

> > > > @"Poelala.2830" said:

> > > > Balance aside, because that's a whole different post, the fact of rating loss and gain discrepancy being so egregious is directly discouraging players

> > >

> > > Losses are going to result in a large MMR drop if, and only if, you were predicted to win. A predicted loss will give a small MMR drop.

> > >

> > > The scenario you describe: mostly winning matches you were predicted to win (small gain) and occasionally losing matches you were predicted to win (big loss) suggests you are at around the appropriate MMR, nothing more. (Though, ideally, it should be small in both directions, but ... it's early in the season, MMR hasn't settled yet, so you probably encounter better players working up to their own appropriate MMR more often than you will next week.)

> > >

> > > > This isn't impossible, but highly improbable. Out of the three days this season has been out, I've seen at least 20 players in the top 5 lose rating as they play more. One of my friends were #1 and TWO LOSSES brought him down 70 rating. A popular twitch streamer lost 44 points in one game, and has a less than 75% win rate in the past 9 games and has lost rating because of this.

> > >

> > > In both cases, these players were expected to win that fight. They didn't, so they lost big, and their opponents gained big. The system adjusted expectations of them appropriately for the skill they, and their opponents, demonstrated.

> > >

> > > > When an unrealistic standard is enforced for god knows what reason, where it seems like playing LESS games is the most beneficial for you, guess what the player base will do? Play less games. The most awful fact of this is the fact that a 50% win rate will result in you being in bronze.

> > >

> > > It really doesn't. It ends up sitting you at your MMR. The real one, not the one you imagine you deserve.

> >

> > The expectations vastly outstrip reality in most cases, you're placed in a system with 4 other players of untested skill level

>

> Their MMR should be quite close to yours; it is tested the same way that yours is. Would you say that your being placed on the team represented someone of "untested skill level" being added, forcing the other four players to cover for your mistakes?

>

> If not, and I bet you wouldn't, then it seems a little unfair to characterize only other players that way since, I'd guess, you are not the one-and-only competent player. :)

 

Unless we all agree that real MMR hidden than your statement is wrong at 200%, I normally get people doing their placement matches..they have no visible rating

[Trevor Boyer.6524]

> @"SlippyCheeze.5483" said:

> > @"Poelala.2830" said:

> > @"SlippyCheeze.5483" I have screenshots of games where I lost 500-77 and lost 17 points right next to games I lost 500-470 and lost 17 points. I clearly had the better team in the second one. Why did this not affect how much I lost?

>

> Because GLICKO2 doesn't take into account how close the game was, just the expectation (eg: are you predicted to win or lose) and the boolean "win/lose" result. That, in turn, comes from the ELO heritage of the system, which comes from real world work on the systems by people who are experts in the area, who I guess determined that the degree of win or loss wasn't ultimately necessary to account for in order to get good results.

>

> I am not arguing that GLICKO2 is the best possible system, even though it is close to the best publicly documented system available at this time, but rather, that it is not as bad as some of the claims made here indicate. Microsoft have their system which is apparently better at handling random group matches, though I'm unable to find much work published studying how effective it actually is.

>

> I am absolutely confident that there is plenty of fame, and glory, and a PhD thesis waiting out there for someone who studies in the area. Probably plenty of high pay jobs at game studios shortly to follow, because I'm sure ANet would love to do better ... as long as there is as rigorous a definition of what better is, and how it works, as their current system. :)

 

Elo, Glicko, Trueskill, all of that is actually widely and highly documented as completely ineffective for team based games. They work fantastic in 1v1 rankings but the system just doesn't make any sense for something like a solo/duo only GW2 ranked season.

 

As far as what can be done better? Well, something like a solo/duo only season with an algorithm judging individual skill based on raw win/loss rate within a 5 vs. 5 setting is something that just doesn't work. Not only does it not work, but it highly encourages gaming the system, which just ends up making the system feel even worse than it did when players were actually playing legitimately. All argument aside, the Glicko system or anything like this that was based off the 1v1 Elo, will completely fall apart in a game like 5 vs. 5 conquest. The answer is real simple, regardless of whatever Anet's reasoning is as to why they won't do this -> 5 man leagues, sign your team in, always play with the same people = block wintrading and get accurate positions on the ladder rankings. I'm sure there are 101 reasons concerning low population and player votes "yada yada" as to why this isn't happening, but if we want an actual functional ladder rating system, this is the only way and it literally solves every single problem within what is making Glicko not work.

 

I mean, just put back solo que and team que, even during ranked season. This way the solo ques can solo que and the people who want to avoid the bullshit can just play in teams. People always bring up the wait times on ques as an argument against this or how it will split player population, which I don't believe to be true at all. Most players will play both ques anyway and wait times really aren't a big issue when you can go into wvw or go work on a map complete while waiting for your que.

 

I'd just love to have a team league again like during S1 and S2. Then I can look at the final standings and say "Yup, we did that with the same 5 guys all season and that is an accurate placing for our team." <- Notice the idea here is that it is a TEAM ranking, not an individual ranking.

 

 

 

Just gonna drop this here again:

 

[Abazigal.3679]

> @"Trevor Boyer.6524" said:

> > @"SlippyCheeze.5483" said:

> > > @"Poelala.2830" said:

> > > @"SlippyCheeze.5483" I have screenshots of games where I lost 500-77 and lost 17 points right next to games I lost 500-470 and lost 17 points. I clearly had the better team in the second one. Why did this not affect how much I lost?

> >

> > Because GLICKO2 doesn't take into account how close the game was, just the expectation (eg: are you predicted to win or lose) and the boolean "win/lose" result. That, in turn, comes from the ELO heritage of the system, which comes from real world work on the systems by people who are experts in the area, who I guess determined that the degree of win or loss wasn't ultimately necessary to account for in order to get good results.

> >

> > I am not arguing that GLICKO2 is the best possible system, even though it is close to the best publicly documented system available at this time, but rather, that it is not as bad as some of the claims made here indicate. Microsoft have their system which is apparently better at handling random group matches, though I'm unable to find much work published studying how effective it actually is.

> >

> > I am absolutely confident that there is plenty of fame, and glory, and a PhD thesis waiting out there for someone who studies in the area. Probably plenty of high pay jobs at game studios shortly to follow, because I'm sure ANet would love to do better ... as long as there is as rigorous a definition of what better is, and how it works, as their current system. :)

>

> Elo, Glicko, Trueskill, all of that is actually widely and highly documented as completely ineffective for team based games. They work fantastic in 1v1 rankings but the system just doesn't make any sense for something like a solo/duo only GW2 ranked season.

 

This. Fact is every single game nowadays uses glicko, but every single game has the same complaints. Coupling it with a season system is good at giving seasonal fun, but keeps splitting the population between " average " and " top players " . Let's be honest, most plat 2 pugs do get facerolled 500-100 by top teams. From this observation, you can believe the ranked system doesnt really hold value..

 

The game is theorically good if everyone plays as much during each season. Experience showed however than you have anyone at any rank between 1100 and 1600 because players do minimal games requirement in order to avoid usual crap happening lately.

 

 

[SlippyCheeze.5483]

> @"Chuck.2864" said:

> > @"SlippyCheeze.5483" said:

> > > @"Chuck.2864" said:

> > > > @"Gwaihir.1745" said:

> > > > Actually I like people who ruin my placement games. Climbing to plat 1 again is easy, fast rewards gain before the fight to climb begins. Bronze - plat 1 is very easy to climb.

> > >

> > > I call kitten. The more I play PvP over the seasons, the more I begin to realize the importance of the placement matches, they are literally the 10 most important games you will play all season because getting significantly higher from where you were placed is incredibly difficult unless you were literally boosted and ended up with like 200 rating more than you really should have, then you will probably fall quickly down the divisions.

> > >

> > > This is probably because with the small population of PvP whether you're high Gold 2 or Plat 1 doesn't really matter, you'll end up in the same matchmaking pool. So as far as the MM system is concerned, a G2 player playing with Plat 1 players is playing within his skill level, ergo he will maintain ~50% winrate which you cannot climb with. For reference I placed in Plat 1-2 and stayed there the past two seasons with a ~50% winrate (positive). This season I had terrible placements including DC's, afkers, some meme builds, went 5-5 and was placed in Gold 2. Guess what? It's a 50% winrate again with around 65-70 matches already played this season and it's basically impossible to climb back to Plat. Every time I make gains I get hit with a 5 game loss streak.

> >

> > A 50 percent **expected** win and loss rate will keep your MMR stable, not cause it to move, as long as you actually deliver the results. Even a streak of five **expected** losses in a row will fairly quickly bring you back up to where you should be once you get back to winning.

> >

> > The only case you need a substantially higher win rate is where you frequently lose matches that you should have been able to win. If your experience is that you can't climb to platinum if you dip too low, the problem is more likely that you can't play at platinum level, not that the initial placement is the most significant thing overall.

> >

> > PS: placement matches use the same MMR adjustment method as normal games.

>

> What exactly is "Platinum level" though?

 

An arbitrary point in the range of MMR that ANet picked and assigned the label to.

 

> That's exactly my point. In 2 seasons I was able to stay in Plat 1 without any issue (where according to you I can't play at "that level", and so if the MM is perfectly fine I should have fell down), but suddenly 2 weeks later I'm a Gold 2 player rather than a "Plat player"? Maybe I'm not even a G3 player anymore because I sometimes dip back into G2 when on a loss streak?

 

So .... uh, yes? MMR is an estimate of your skill **compared to all other players**, so it can change even if your actual skill doesn't change in that time.

 

> Answer me this - do you actually consider someone with 1530 rating to be playing at a different level to someone with 1470 rating because one is in Gold and the other in Plat? Even the MM system doesn't believe that, but the latter player might never break that 1500 barrier whilst consistently being matchmade with the former player

 

Nope. I think that the MMR system is generally correct -- at least to the same level as other competitive games that use it, which is to say, widely complained about, considered "broken" on the forums, and delivering the expected results, which are appropriately challenging matches, by the developers who closely watch the results.

 

The fairly arbitrary ranking system slapped onto MMR on top of that is just that: arbitrary. The reason I used it in the comments above is because it is a commonly used trope in these discussions, and it's good enough. The reality is, as you say, that those two players are pretty close in skill, and so are expected to perform roughly equivalently, so yeah, they form fair and challenging matches when grouped together because of that similarity.

 

I could have just said "the problem is more likely that you can't play at a 1501 MMR level, not that the initial placement is the most significant thing overall", and the statement would be identical. Exactly as you say. :)

[SlippyCheeze.5483]

> @"Arheundel.6451" said:

> > @"SlippyCheeze.5483" said:

> > > @"Arheundel.6451" said:

> > > > @"SlippyCheeze.5483" said:

> > > > > @"Poelala.2830" said:

> > > > > Balance aside, because that's a whole different post, the fact of rating loss and gain discrepancy being so egregious is directly discouraging players

> > > >

> > > > Losses are going to result in a large MMR drop if, and only if, you were predicted to win. A predicted loss will give a small MMR drop.

> > > >

> > > > The scenario you describe: mostly winning matches you were predicted to win (small gain) and occasionally losing matches you were predicted to win (big loss) suggests you are at around the appropriate MMR, nothing more. (Though, ideally, it should be small in both directions, but ... it's early in the season, MMR hasn't settled yet, so you probably encounter better players working up to their own appropriate MMR more often than you will next week.)

> > > >

> > > > > This isn't impossible, but highly improbable. Out of the three days this season has been out, I've seen at least 20 players in the top 5 lose rating as they play more. One of my friends were #1 and TWO LOSSES brought him down 70 rating. A popular twitch streamer lost 44 points in one game, and has a less than 75% win rate in the past 9 games and has lost rating because of this.

> > > >

> > > > In both cases, these players were expected to win that fight. They didn't, so they lost big, and their opponents gained big. The system adjusted expectations of them appropriately for the skill they, and their opponents, demonstrated.

> > > >

> > > > > When an unrealistic standard is enforced for god knows what reason, where it seems like playing LESS games is the most beneficial for you, guess what the player base will do? Play less games. The most awful fact of this is the fact that a 50% win rate will result in you being in bronze.

> > > >

> > > > It really doesn't. It ends up sitting you at your MMR. The real one, not the one you imagine you deserve.

> > >

> > > The expectations vastly outstrip reality in most cases, you're placed in a system with 4 other players of untested skill level

> >

> > Their MMR should be quite close to yours; it is tested the same way that yours is. Would you say that your being placed on the team represented someone of "untested skill level" being added, forcing the other four players to cover for your mistakes?

> >

> > If not, and I bet you wouldn't, then it seems a little unfair to characterize only other players that way since, I'd guess, you are not the one-and-only competent player. :)

>

> Unless we all agree that real MMR hidden than your statement is wrong at 200%, I normally get people doing their placement matches..they have no visible rating

 

...it is hidden during placement?

[SlippyCheeze.5483]

> @"Trevor Boyer.6524" said:

> > @"SlippyCheeze.5483" said:

> > > @"Poelala.2830" said:

> > > @"SlippyCheeze.5483" I have screenshots of games where I lost 500-77 and lost 17 points right next to games I lost 500-470 and lost 17 points. I clearly had the better team in the second one. Why did this not affect how much I lost?

> >

> > Because GLICKO2 doesn't take into account how close the game was, just the expectation (eg: are you predicted to win or lose) and the boolean "win/lose" result. That, in turn, comes from the ELO heritage of the system, which comes from real world work on the systems by people who are experts in the area, who I guess determined that the degree of win or loss wasn't ultimately necessary to account for in order to get good results.

> >

> > I am not arguing that GLICKO2 is the best possible system, even though it is close to the best publicly documented system available at this time, but rather, that it is not as bad as some of the claims made here indicate. Microsoft have their system which is apparently better at handling random group matches, though I'm unable to find much work published studying how effective it actually is.

> >

> > I am absolutely confident that there is plenty of fame, and glory, and a PhD thesis waiting out there for someone who studies in the area. Probably plenty of high pay jobs at game studios shortly to follow, because I'm sure ANet would love to do better ... as long as there is as rigorous a definition of what better is, and how it works, as their current system. :)

>

> Elo, Glicko, Trueskill, all of that is actually widely and highly documented as completely ineffective for team based games. They work fantastic in 1v1 rankings but the system just doesn't make any sense for something like a solo/duo only GW2 ranked season.

 

You know TrueSkill is designed to work better than Glicko exactly in the situation that a GW2 ranked season produces, right? Like, it is basically Glicko plus trying to improve the accuracy of estimation when you play ranked games in a "randomly" assigned group? I'm checking because your comment seems to suggest that none of these were intended for that use...

 

> Just gonna drop this here again:

>

 

Do you have a written summary of the content of that? I'm otherwise not going to spend the time to respond to a video.

[Arheundel.6451]

> @"SlippyCheeze.5483" said:

> > @"Chuck.2864" said:

> > > @"SlippyCheeze.5483" said:

> > > > @"Chuck.2864" said:

> > > > > @"Gwaihir.1745" said:

> > > > > Actually I like people who ruin my placement games. Climbing to plat 1 again is easy, fast rewards gain before the fight to climb begins. Bronze - plat 1 is very easy to climb.

> > > >

> > > > I call kitten. The more I play PvP over the seasons, the more I begin to realize the importance of the placement matches, they are literally the 10 most important games you will play all season because getting significantly higher from where you were placed is incredibly difficult unless you were literally boosted and ended up with like 200 rating more than you really should have, then you will probably fall quickly down the divisions.

> > > >

> > > > This is probably because with the small population of PvP whether you're high Gold 2 or Plat 1 doesn't really matter, you'll end up in the same matchmaking pool. So as far as the MM system is concerned, a G2 player playing with Plat 1 players is playing within his skill level, ergo he will maintain ~50% winrate which you cannot climb with. For reference I placed in Plat 1-2 and stayed there the past two seasons with a ~50% winrate (positive). This season I had terrible placements including DC's, afkers, some meme builds, went 5-5 and was placed in Gold 2. Guess what? It's a 50% winrate again with around 65-70 matches already played this season and it's basically impossible to climb back to Plat. Every time I make gains I get hit with a 5 game loss streak.

> > >

> > > A 50 percent **expected** win and loss rate will keep your MMR stable, not cause it to move, as long as you actually deliver the results. Even a streak of five **expected** losses in a row will fairly quickly bring you back up to where you should be once you get back to winning.

> > >

> > > The only case you need a substantially higher win rate is where you frequently lose matches that you should have been able to win. If your experience is that you can't climb to platinum if you dip too low, the problem is more likely that you can't play at platinum level, not that the initial placement is the most significant thing overall.

> > >

> > > PS: placement matches use the same MMR adjustment method as normal games.

> >

> > What exactly is "Platinum level" though?

>

> An arbitrary point in the range of MMR that ANet picked and assigned the label to.

>

> > That's exactly my point. In 2 seasons I was able to stay in Plat 1 without any issue (where according to you I can't play at "that level", and so if the MM is perfectly fine I should have fell down), but suddenly 2 weeks later I'm a Gold 2 player rather than a "Plat player"? Maybe I'm not even a G3 player anymore because I sometimes dip back into G2 when on a loss streak?

>

> So .... uh, yes? MMR is an estimate of your skill **compared to all other players**, so it can change even if your actual skill doesn't change in that time.

>

> > Answer me this - do you actually consider someone with 1530 rating to be playing at a different level to someone with 1470 rating because one is in Gold and the other in Plat? Even the MM system doesn't believe that, but the latter player might never break that 1500 barrier whilst consistently being matchmade with the former player

>

> Nope. I think that the MMR system is generally correct -- at least to the same level as other competitive games that use it, which is to say, widely complained about, considered "broken" on the forums, and delivering the expected results, which are appropriately challenging matches, by the developers who closely watch the results.

>

> The fairly arbitrary ranking system slapped onto MMR on top of that is just that: arbitrary. The reason I used it in the comments above is because it is a commonly used trope in these discussions, and it's good enough. The reality is, as you say, that those two players are pretty close in skill, and so are expected to perform roughly equivalently, so yeah, they form fair and challenging matches when grouped together because of that similarity.

>

> I could have just said "the problem is more likely that you can't play at a 1501 MMR level, not that the initial placement is the most significant thing overall", and the statement would be identical. Exactly as you say. :)

 

You keep saying at what level people should play...like the outcome of the match depends solely on them and the game doesn't account only for w/l..

[Faux Play.6104]

if you match players with the same rating on a team, glicko will work well. The problem is we have huge skill gaps on a team. This causes everyone to have a very volatile rating. There is no reason for doing the soft reset at the start of the season. There already is history of what someone should be rated at. All it does is create chaos at the start of the season. I had 3 placement matches where the team we beat scored less than 50 pts. None of them had 4 on their team until people afk'd because the match was way out of hand.

[SlippyCheeze.5483]

> @"Arheundel.6451" said:

> > @"SlippyCheeze.5483" said:

> > > @"Chuck.2864" said:

> > > > @"SlippyCheeze.5483" said:

> > > > > @"Chuck.2864" said:

> > > > > > @"Gwaihir.1745" said:

> > > > > > Actually I like people who ruin my placement games. Climbing to plat 1 again is easy, fast rewards gain before the fight to climb begins. Bronze - plat 1 is very easy to climb.

> > > > >

> > > > > I call kitten. The more I play PvP over the seasons, the more I begin to realize the importance of the placement matches, they are literally the 10 most important games you will play all season because getting significantly higher from where you were placed is incredibly difficult unless you were literally boosted and ended up with like 200 rating more than you really should have, then you will probably fall quickly down the divisions.

> > > > >

> > > > > This is probably because with the small population of PvP whether you're high Gold 2 or Plat 1 doesn't really matter, you'll end up in the same matchmaking pool. So as far as the MM system is concerned, a G2 player playing with Plat 1 players is playing within his skill level, ergo he will maintain ~50% winrate which you cannot climb with. For reference I placed in Plat 1-2 and stayed there the past two seasons with a ~50% winrate (positive). This season I had terrible placements including DC's, afkers, some meme builds, went 5-5 and was placed in Gold 2. Guess what? It's a 50% winrate again with around 65-70 matches already played this season and it's basically impossible to climb back to Plat. Every time I make gains I get hit with a 5 game loss streak.

> > > >

> > > > A 50 percent **expected** win and loss rate will keep your MMR stable, not cause it to move, as long as you actually deliver the results. Even a streak of five **expected** losses in a row will fairly quickly bring you back up to where you should be once you get back to winning.

> > > >

> > > > The only case you need a substantially higher win rate is where you frequently lose matches that you should have been able to win. If your experience is that you can't climb to platinum if you dip too low, the problem is more likely that you can't play at platinum level, not that the initial placement is the most significant thing overall.

> > > >

> > > > PS: placement matches use the same MMR adjustment method as normal games.

> > >

> > > What exactly is "Platinum level" though?

> >

> > An arbitrary point in the range of MMR that ANet picked and assigned the label to.

> >

> > > That's exactly my point. In 2 seasons I was able to stay in Plat 1 without any issue (where according to you I can't play at "that level", and so if the MM is perfectly fine I should have fell down), but suddenly 2 weeks later I'm a Gold 2 player rather than a "Plat player"? Maybe I'm not even a G3 player anymore because I sometimes dip back into G2 when on a loss streak?

> >

> > So .... uh, yes? MMR is an estimate of your skill **compared to all other players**, so it can change even if your actual skill doesn't change in that time.

> >

> > > Answer me this - do you actually consider someone with 1530 rating to be playing at a different level to someone with 1470 rating because one is in Gold and the other in Plat? Even the MM system doesn't believe that, but the latter player might never break that 1500 barrier whilst consistently being matchmade with the former player

> >

> > Nope. I think that the MMR system is generally correct -- at least to the same level as other competitive games that use it, which is to say, widely complained about, considered "broken" on the forums, and delivering the expected results, which are appropriately challenging matches, by the developers who closely watch the results.

> >

> > The fairly arbitrary ranking system slapped onto MMR on top of that is just that: arbitrary. The reason I used it in the comments above is because it is a commonly used trope in these discussions, and it's good enough. The reality is, as you say, that those two players are pretty close in skill, and so are expected to perform roughly equivalently, so yeah, they form fair and challenging matches when grouped together because of that similarity.

> >

> > I could have just said "the problem is more likely that you can't play at a 1501 MMR level, not that the initial placement is the most significant thing overall", and the statement would be identical. Exactly as you say. :)

>

> You keep saying at what level people should play...like the outcome of the match depends solely on them and the game doesn't account only for w/l..

 

For any individual game, yeah, that's true, it doesn't depend solely on them. Over **many games**, statistically, GLICKO-2 and friends show the desired behaviour: people move to an approximation of their skill when grouped with different random people. Your point would be much more significant if, for example, sPvP was 5v5 teams, but team assignment was random **AND FIXED THE SEASON**.

 

Otherwise, on average, it works out, over the order of ten to twenty games. At least to the satisfaction of pretty much every large company that makes money explicitly from providing interesting matches with long term replay value, let alone all the ones like GW2 that it is a sideline in.