Jump to content
  • Sign Up

Another reason to get rid of pvp until its fixed


Txwel.2764

Recommended Posts

When this game calculates rating change at the end of the match, it basicly compares your own rating to the average of the match. If your rating is higher, you played against "weaker" players. Beating players below you is worth less, and getting beaten by them hurts more.
Basicly in this game it does matter who you beat/lost to. If you think farming low-ranks should net you the same gains... good luck convincing the devs by demanding they delete the gamemode lol.

  • Like 4
Link to comment
Share on other sites

4 hours ago, Bazsi.2734 said:

When this game calculates rating change at the end of the match, it basicly compares your own rating to the average of the match. If your rating is higher, you played against "weaker" players. Beating players below you is worth less, and getting beaten by them hurts more.
Basicly in this game it does matter who you beat/lost to. If you think farming low-ranks should net you the same gains... good luck convincing the devs by demanding they delete the gamemode lol.


Yes this is how that is supposed to work…until you realize that there are “strong” players with lower rating than their supposed to have, and “weak” players with rating higher than they are supposed to have.

 

You have veteran Plat 2 level players with 1400 rating, playing in the same games as greenhorns that just started PvP last week also I n 1400.

 

Because it’s all mixed together in these random team games, when the match maker hands you a match you are supposed to win you get huge divergences in point gain when you win or lose, because in reality the opposing team has the same skill level as you, but on paper they have higher(or lower) rating.

The problem is the assumption that rating = skill which is a false assumption…especially in a game where your rating is determined by 9 completely random variables in every game. 

Edited by JusticeRetroHunter.7684
  • Like 1
  • Thanks 4
Link to comment
Share on other sites

Problem as usual is population. If you care about climbing you never want to play outside prime time hours (7PM EST to 2AM EST, give or take).

Otherwise you get completely messy matchmaking where you will lose huge chunks of rating anytime you lose despite how unbalanced said matchmaking was and wins you will scrape by won't give you much.

That's not theorical, i tested it out. I lost 200 rating queueing in the middle of the afternoon cause that's the only time i could play for the past few weeks but i have 0 issues climbing back when playing in the evening.

That trend also apply to match quality (difference in score at the end of a match, the smaller the difference the higher quality the match was). Match quality DRASTICALLY increases during prime time.

Edited by Ashgar.3024
  • Like 3
Link to comment
Share on other sites

3 hours ago, JusticeRetroHunter.7684 said:


Yes this is how that is supposed to work…until you realize that there are “strong” players with lower rating than their supposed to have, and “weak” players with rating higher than they are supposed to have.

 

You have veteran Plat 2 level players with 1400 rating, playing in the same games as greenhorns that just started PvP last week also I n 1400.

 

Because it’s all mixed together in these random team games, when the match maker hands you a match you are supposed to win you get huge divergences in point gain when you win or lose, because in reality the opposing team has the same skill level as you, but on paper they have higher(or lower) rating.

The problem is the assumption that rating = skill which is a false assumption…especially in a game where your rating is determined by 9 completely random variables in every game. 

I wanted to comment on this because lately as of 2022, the glicko algorithm with the way they have it setup, is completely failing in every mentionable way.

Win trading aside, the actual algorithm functions no longer work at all in a low population community concerning any accurate representation of player value or any level of match quality.

This is mainly due to one seriously outdated algorithm function that desperately needs to be removed. 

  1. People don't realize that MMR and Seasonal Rating are two different things, and that match making in Ranked uses both of these metrics when creating matches. Here is how it works: 

    The first phase, called filtering, gathers players based on their current MMR. The primary purpose of this phase is to both reduce the number of players being considered for a match, and to ensure that the match is appropriate given each player's skill level. Over time, padding is added to your player rating. While this may decrease match quality, it helps ensure that outliers still receive matches.

    The second phase of the algorithm is the scoring phase. During this phase each player is scored against every other player being considered for matchmaking. The metrics used during this phase include: rating, rank, games played, party size, profession, and dishonor. With each metric the system is looking for players that are as close as possible to the average of those already selected. The system also attempts to keep the number of duplicate professions to a minimum

    What this means is that total games played & old bottom left hand of screen glory rank may not be adding to your account's base MMR directly, but they are creating a hidden segregation or a hidden "divisional split" that is not reflected in the leaderboards at all. Please allow me to explain this.
  2. Are you familiar with Mario Kart? How the game is set with difficulties where you can play on 50cc or 100cc or 150cc? On 50cc the karts drive slower and the NPCs are less aggressive. On 150cc the karts drive much faster and are harder to steer through the courses as well as the NPCs being much more aggressive. The GW2 algorithm in Ranked is essentially creating an invisible divisional split, setting an invisible difference in match difficulty between what new accounts get compared to what veteran accounts get, and it is not unlike the difference between 50cc and 150cc in Mario Kart. Let's say that the  algorithm is currently creating a match between RED 1600 duo and BLUE 1600 duo. Both of those duos have around 20,000 games played. The algorithm needs to fill 1 more slot to start the game. Due to how the match making actually works, if I was to push enter queue at the same time as someone else and we were both 1600 rated but I had a total of 20,000 games played and his account had only 2000 played, I will get selected to be thrown into that match every time, and the other guy gets a chance to remain sitting in the match making to be placed into an easier game full of accounts with a more similar total games played.
  3. The algorithm is definitely attempting to sort new players in with & against new players, and veteran players in with & against veteran players. The longer GW2 is around, the stronger this effect gets, regardless of population actually. Back in 2012-2015, the discrepancy of difference in total games played between new & veteran accounts probably looked something like 700 games played compared to 2000 or maybe a few 5000 guys out there. But in 2022 the difference is like 1000 games played or less for new guys, 5000 games played for casuals, 10,000 games played for guys who have followed the game for many years, and 20,000+ games played for the hardcore pvp community. This is a large discrepancy in difference of match making placement, with how the algorithm actually works. With being a decade into GW2, this is creating a very real invisible divisional segregation that is not reflected in the leaderboards. When an old veteran account with 20,000+ games played achieves bottom plat 1, he had to do that while being repeatedly placed into games with other guys who have 20,000+ games played whenever the algorithm could do that to him. But when a new account achieves bottom plat 1, the algorithm was hand picking him to be in games with other accounts that had less than 1000 games played whenever it could do that for him. And likewise, if the algorithm was making an easy game of RED duo 1500 rated vs. BLUE duo 1500 rated and they have 1000 games played, if the algorithm was filling 1 more slot to start that game, even if the veteran and the new guy have the same seasonal rating, the algorithm will select the new guy for that match before the veteran every time. In fact, there would be a significant line of players ahead of the veteran who would be selected to etner that match before the guy with 20,000 games played, even if he was only rated 1500. Both the 20,000 games played veteran and the new guy with 500 games played could both be ranked at 1500, but the veteran had to do it while playing in the 150cc division, whilst the new guy achieved it while being placed into 50cc division games. Do you see what I mean here? But none of this is reflected in the leaderboards. As of 2022, this is truly creating an effect where one man's plat 2 is worth another man's gold 3. This is exactly why we see so many pve guys showing up and getting baron titles in their first season of play, because they are repeatedly being selected to enter 50cc matches, any time the algorithm can do that for them.

These problems become exasperated with how many people smurf around on alt accounts nowadays. It just all equates to a situation where the match quality doesn't even make sense anymore, and veteran accounts get the worst of it. Arenanet needs to remove the total games played/glory rank sorting phase with Ranked match making. Doing so would increase match quality a great deal, as it would eliminate most of the benefits of smurfing, as well as remove a bit of the incentive to win trade. If matches made more sense again, I think less people would be willing to match manipulate. This is real simple, the match making in Ranked seasons should only function off of current seasonal rating alone. It would reflect a much truer representation in the leaderboards of actual player value.

And upon that, come on man, if we are all supposed to be on one single leaderboard, a bunch of guys who are rated at 1500 should be all getting the same match quality, not being segregated into easier or more difficult matches based on their total games played. That's just not fair and it's a great way to kitten off your veteran player and make them not want to play the game.

This a problem that very seriously needs to be sorted out before the steam release.

Edited by Trevor Boyer.6524
  • Like 2
  • Thanks 2
  • Confused 1
Link to comment
Share on other sites

2 hours ago, Trevor Boyer.6524 said:

I wanted to comment on this because lately as of 2022, the glicko algorithm with the way they have it setup, is completely failing in every mentionable way.

Win trading aside, the actual algorithm functions no longer work at all in a low population community concerning any accurate representation of player value or any level of match quality.

This is mainly due to one seriously outdated algorithm function that desperately needs to be removed. 

  1. People don't realize that MMR and Seasonal Rating are two different things, and that match making in Ranked uses both of these metrics when creating matches. Here is how it works: 

    The first phase, called filtering, gathers players based on their current MMR. The primary purpose of this phase is to both reduce the number of players being considered for a match, and to ensure that the match is appropriate given each player's skill level. Over time, padding is added to your player rating. While this may decrease match quality, it helps ensure that outliers still receive matches.

    The second phase of the algorithm is the scoring phase. During this phase each player is scored against every other player being considered for matchmaking. The metrics used during this phase include: rating, rank, games played, party size, profession, and dishonor. With each metric the system is looking for players that are as close as possible to the average of those already selected. The system also attempts to keep the number of duplicate professions to a minimum

    What this means is that total games played & old bottom left hand of screen glory rank may not be adding to your account's base MMR directly, but they are creating a hidden segregation or a hidden "divisional split" that is not reflected in the leaderboards at all. Please allow me to explain this.
  2. Are you familiar with Mario Kart? How the game is set with difficulties where you can play on 50cc or 100cc or 150cc? On 50cc the karts drive slower and the NPCs are less aggressive. On 150cc the karts drive much faster and are harder to steer through the courses as well as the NPCs being much more aggressive. The GW2 algorithm in Ranked is essentially creating an invisible divisional split, setting an invisible difference in match difficulty between what new accounts get compared to what veteran accounts get, and it is not unlike the difference between 50cc and 150cc in Mario Kart. Let's say that the  algorithm is currently creating a match between RED 1600 duo and BLUE 1600 duo. Both of those duos have around 20,000 games played. The algorithm needs to fill 1 more slot to start the game. Due to how the match making actually works, if I was to push enter queue at the same time as someone else and we were both 1600 rated but I had a total of 20,000 games played and his account had only 2000 played, I will get selected to be thrown into that match every time, and the other guy gets a chance to remain sitting in the match making to be placed into an easier game full of accounts with a more similar total games played.
  3. The algorithm is definitely attempting to sort new players in with & against new players, and veteran players in with & against veteran players. The longer GW2 is around, the stronger this effect gets, regardless of population actually. Back in 2012-2015, the discrepancy of difference in total games played between new & veteran accounts probably looked something like 700 games played compared to 2000 or maybe a few 5000 guys out there. But in 2022 the difference is like 1000 games played or less for new guys, 5000 games played for casuals, 10,000 games played for guys who have followed the game for many years, and 20,000+ games played for the hardcore pvp community. This is a large discrepancy in difference of match making placement, with how the algorithm actually works. With being a decade into GW2, this is creating a very real invisible divisional segregation that is not reflected in the leaderboards. When an old veteran account with 20,000+ games played achieves bottom plat 1, he had to do that while being repeatedly placed into games with other guys who have 20,000+ games played whenever the algorithm could do that to him. But when a new account achieves bottom plat 1, the algorithm was hand picking him to be in games with other accounts that had less than 1000 games played whenever it could do that for him. And likewise, if the algorithm was making an easy game of RED duo 1500 rated vs. BLUE duo 1500 rated and they have 1000 games played, if the algorithm was filling 1 more slot to start that game, even if the veteran and the new guy have the same seasonal rating, the algorithm will select the new guy for that match before the veteran every time. In fact, there would be a significant line of players ahead of the veteran who would be selected to etner that match before the guy with 20,000 games played, even if he was only rated 1500. Both the 20,000 games played veteran and the new guy with 500 games played could both be ranked at 1500, but the veteran had to do it while playing in the 150cc division, whilst the new guy achieved it while being placed into 50cc division games. Do you see what I mean here? But none of this is reflected in the leaderboards. As of 2022, this is truly creating an effect where one man's plat 2 is worth another man's gold 3. This is exactly why we see so many pve guys showing up and getting baron titles in their first season of play, because they are repeatedly being selected to enter 50cc matches, any time the algorithm can do that for them.

These problems become exasperated with how many people smurf around on alt accounts nowadays. It just all equates to a situation where the match quality doesn't even make sense anymore, and veteran accounts get the worst of it. Arenanet needs to remove the total games played/glory rank sorting phase with Ranked match making. Doing so would increase match quality a great deal, as it would eliminate most of the benefits of smurfing, as well as remove a bit of the incentive to win trade. If matches made more sense again, I think less people would be willing to match manipulate. This is real simple, the match making in Ranked seasons should only function off of current seasonal rating alone. It would reflect a much truer representation in the leaderboards of actual player value.

And upon that, come on man, if we are all supposed to be on one single leaderboard, a bunch of guys who are rated at 1500 should be all getting the same match quality, not being segregated into easier or more difficult matches based on their total games played. That's just not fair and it's a great way to kitten off your veteran player and make them not want to play the game.

This a problem that very seriously needs to be sorted out before the steam release.

 

Right exactly.

 

Like your Mario Kart example, let me provide a another, more universal example, one that is heavily researched as to provide a concrete analysis on the issue of algorithm's in a general sense.

 

The YouTube Algorithm

The YouTube algorithm functions in very much the same way as the matchmaker. In theory, good video's rise to the top, bad video's stay at the bottom, through a competitive selection process. 

 

However, as many people I'm sure are aware, the behavior of how the algorithm should work in theory, over time, leads to a certain sort of bias, where catchy thumbnails and titles, bordering on clickbait, as well as video content containing mostly shock value and novelty (low quality memes, news stories, vlogs) are more-favored by the algorithm than actual high quality content, like art...or content that takes a lot more time and effort to produce like movies and animation)

 

In essence, YouTube doesn't actually put into practice, the goal that it was designed to do in theory. The metrics laid out by this algorithm tend to promote not good high quality videos, but instead bad video's that take advantage of human behaviors (the incessance to click on a suggestive thumbnail or a shocking title)

 

Paperclip Paradox

There's been a lot of folks that have gone to congress to present the dangers of algorithm's, which boil down to a fundamental problem with them, which is the paperclip paradox: You design a machine with a task, and the machine will set out to achieve this goal, in spite of and sometimes even against our human interests. In the case of the paperclip machine, one designs this machine with the simple goal of "creating paperclips" which eventually leads to slaughtering people, to make space for factories, consuming resources and the environment and eventually destroying the world (and in the end, the whole universe) in the interest of creating paperclips.

 

GW2 Matchmaker

The same story takes place with algorithm's like the matchmaker. The algorithm is working toward a simple goal "Give players a numerical evaluation of skill, and using that number, give out fair games", but in practice, what it is actually doing, is forcing everyone into the same rating, to fit an idealized bell curve, which is technically still "fair" because if everyone eventually has 1400 rating, then the games are still fair right... and this aligns with the algorithm's goal...but it doesn't actually align with our human interests, and what we would think of as "a true theory of skill."

 

Because the matchmaker will tend to group veterans with veterans and greenhorns with other greenhorns, a competitive advantage will appear where veterans are incentivized to create new account, to be matched with greenhorns, for easier games. This is one, of many algorithmically driven  incentives that you also listed (like duo-queing, win-trading, match manipulating and so on...) I even tested for myself, creating a smurf account and found the difference in match quality staggering, what a difference it makes when you have 100 games rather than 10,000 games. Many of these games spent destroying completely new players, since I was always placed on the team of veterans with 10,000 games. 

 

In aggregate, the problems with the matchmaker boil down to I think a simple premise. That we simply can not compartmentalize a "true theory of skill."

 

This is a video of Stephan Wolfram, one of my favorite computer scientists (Billionaire, with a 160 IQ), in front of congress explaining the nature of the issues of algorithms, and how one might go about dealing with the above issues with them.

 

 

Edited by JusticeRetroHunter.7684
Link to comment
Share on other sites

2 hours ago, JusticeRetroHunter.7684 said:

 

Right exactly.

 

Like your Mario Kart example, let me provide a another, more universal example, one that is heavily researched as to provide a concrete analysis on the issue of algorithm's in a general sense.

 

The YouTube Algorithm

The YouTube algorithm functions in very much the same way as the matchmaker. In theory, good video's rise to the top, bad video's stay at the bottom, through a competitive selection process. 

 

However, as many people I'm sure are aware, the behavior of how the algorithm should work in theory, over time, leads to a certain sort of bias, where catchy thumbnails and titles, bordering on clickbait, as well as video content containing mostly shock value and novelty (low quality memes, news stories, vlogs) are more-favored by the algorithm than actual high quality content, like art...or content that takes a lot more time and effort to produce like movies and animation)

 

In essence, YouTube doesn't actually put into practice, the goal that it was designed to do in theory. The metrics laid out by this algorithm tend to promote not good high quality videos, but instead bad video's that take advantage of human behaviors (the incessance to click on a suggestive thumbnail or a shocking title)

 

Paperclip Paradox

There's been a lot of folks that have gone to congress to present the dangers of algorithm's, which boil down to a fundamental problem with them, which is the paperclip paradox: You design a machine with a task, and the machine will set out to achieve this goal, in spite of and sometimes even against our human interests. In the case of the paperclip machine, one designs this machine with the simple goal of "creating paperclips" which eventually leads to slaughtering people, to make space for factories, consuming resources and the environment and eventually destroying the world (and in the end, the whole universe) in the interest of creating paperclips.

 

GW2 Matchmaker

The same story takes place with algorithm's like the matchmaker. The algorithm is working toward a simple goal "Give players a numerical evaluation of skill, and using that number, give out fair games", but in practice, what it is actually doing, is forcing everyone into the same rating, to fit an idealized bell curve, which is technically still "fair" because if everyone eventually has 1400 rating, then the games are still fair right... and this aligns with the algorithm's goal...but it doesn't actually align with our human interests, and what we would think of as "a true theory of skill."

 

Because the matchmaker will tend to group veterans with veterans and greenhorns with other greenhorns, a competitive advantage will appear where veterans are incentivized to create new account, to be matched with greenhorns, for easier games. This is one, of many algorithmically driven  incentives that you also listed (like duo-queing, win-trading, match manipulating and so on...) I even tested for myself, creating a smurf account and found the difference in match quality staggering, what a difference it makes when you have 100 games rather than 10,000 games. Many of these games spent destroying completely new players, since I was always placed on the team of veterans with 10,000 games. 

 

In aggregate, the problems with the matchmaker boil down to I think a simple premise. That we simply can not compartmentalize a "true theory of skill."

 

This is a video of Stephan Wolfram, one of my favorite computer scientists (Billionaire, with a 160 IQ), in front of congress explaining the nature of the issues of algorithms, and how one might go about dealing with the above issues with them.

 

 

I've found the match difficulty / quality to be about the same on a new acc vs an old acc at similar ratings. I think what you are experiencing is mostly placebo.

Edited by agrippastrilemma.8741
Link to comment
Share on other sites

33 minutes ago, agrippastrilemma.8741 said:

I've found the match difficulty / quality to be about the same on a new acc vs an old acc at similar ratings. I think what you are experiencing is mostly placebo.

We are talking about total games played not the age of your account.

You wouldn't notice much of a difference if your old account is 1000 games and your new account is 200 games.

We are talking about the difference between old accounts with like 10000+ games played vs. less than 1000 games. Then you start seeing a serious discrepancy in the quality of match making between the two accounts.

I have streamed many test runs many seasons in a row, where my main account with over 20,000 games played places in gold 3 after the first 15 games, but my alt with about 800 games played rides directly into 1600 margins after the first 15 games. I've actually done this live on about 4 different occasions. On one occasion, I ran an extreme test with a f2p account that barely qualified to enter ranked queue vs. my main account. I tried hard in the first 10 games on my main account and placed gold 3. On the f2p account I sarcastically AFK'd in every game in the first 10 matches. Guess what? The f2p that AFK'd placed only slightly lower than the main by less than 10 points. There is so much wrong with this, I'm not going to get into it here & now.

Link to comment
Share on other sites

True I have 10k+ games on main account and all brand new f2p accounts are so much easier to place p2-p3 it's a shame for a same player to have almost 200 points diff while the only different variable is the account on which u play.

 

I would that this phenomena is even more enlighted when u duo with 2 10k+ account and then go duo on new accounts its really relevant

 

Just boycoyt ranked and go have fun unrank with bunch of friends like good old times. We have the power but we do not use it.

Link to comment
Share on other sites

7 hours ago, Trevor Boyer.6524 said:

We are talking about total games played not the age of your account.

You wouldn't notice much of a difference if your old account is 1000 games and your new account is 200 games.

We are talking about the difference between old accounts with like 10000+ games played vs. less than 1000 games. Then you start seeing a serious discrepancy in the quality of match making between the two accounts.

I have streamed many test runs many seasons in a row, where my main account with over 20,000 games played places in gold 3 after the first 15 games, but my alt with about 800 games played rides directly into 1600 margins after the first 15 games. I've actually done this live on about 4 different occasions. On one occasion, I ran an extreme test with a f2p account that barely qualified to enter ranked queue vs. my main account. I tried hard in the first 10 games on my main account and placed gold 3. On the f2p account I sarcastically AFK'd in every game in the first 10 matches. Guess what? The f2p that AFK'd placed only slightly lower than the main by less than 10 points. There is so much wrong with this, I'm not going to get into it here & now.

My old acc is like 10k my new acc is like 150 games. Seems about the same.

But maybe ill do some tests at some point.

Edited by agrippastrilemma.8741
Link to comment
Share on other sites

  • 2 weeks later...

ELO system is great whenever the population is high.

Nowadays,

1. Population is very low

2. The game mode has been monopolized by the same few 30 top players for many years

3. Many of them are involved in cheating areas and some of them have even been banned

4. These are the ones spending money and advertising mostly the game mode through their social networks (twitch, youtube, etc.)

 

As a result, population keeps decreasing

And It makes many player run away from the game mode.

 

Edited by Yannsen.9736
  • Like 1
Link to comment
Share on other sites

On 4/1/2022 at 1:30 PM, JusticeRetroHunter.7684 said:


Yes this is how that is supposed to work…until you realize that there are “strong” players with lower rating than their supposed to have, and “weak” players with rating higher than they are supposed to have.

 

You have veteran Plat 2 level players with 1400 rating, playing in the same games as greenhorns that just started PvP last week also I n 1400.

 

Because it’s all mixed together in these random team games, when the match maker hands you a match you are supposed to win you get huge divergences in point gain when you win or lose, because in reality the opposing team has the same skill level as you, but on paper they have higher(or lower) rating.

The problem is the assumption that rating = skill which is a false assumption…especially in a game where your rating is determined by 9 completely random variables in every game. 

you gotta thanks duo q for that, where 1400 players are plat 2, and 1600 players are gods of arena, duo q + q dodge offhours is too much of a strong tool to just go for whatever ranking you aiming, if you have 24h free to play the game you could just dodge everybody that could beat you and play 80% safe matches for easy mmr. I really think a true plat 2 player that spams matches and don't try to avoid anything is now somewhere around 1450-1550 rating, while 1350 players duo qing and q dodgint are sitting at  1550+ easily

  • Like 1
Link to comment
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now
×
×
  • Create New...