Jump to content
  • Sign Up

The Hidden Part of Ranked Matchmaking


Ryo.5824

Recommended Posts

Why am I always getting bad teammates? Why are my opponents strong? Am I really just the unlucky one? For those of you who may or may not have these questions, heres an explanation that may potentially clear your doubt.It is pretty obvious that the matchmaking system makes the teams by trying to balance out the effective rating (aka ladder ranking) points between the two teams. However, as you might have noticed, if the system was only like that, then it will not make very good matchups. Guild Wars 2 has many gamemodes to play in such as Fractals, Open Worlds, WvW and more. People don't play all the time, some players can have a much lower effective rating than their skills enable them to simply because they haven't played in a while or they haven't played enough games or maybe even because they got unlucky facing strong opposing teams for a few games in a row. Better rating doesn't necessarily indicate better skills, and matchmaking using only the rating will be a disaster.I have researched and compared some other competitive games with similar laddering systems, and I came to a system called the hidden point system that exists in most other popular games, which I feel is most likely the system that exists in gw2. The logic behind is simple: you have a hidden point for your account that you cannot check based on your performance. How this system works is that both your ladder ranking and your hidden points are use when matchmaking. For example, if you have a streak of winning, your hidden point increases. Vice versa, if you are on a losing streak your hidden point decreases. But what is going to also contribute a lot to your hidden point will be your statistics. In the gw2 terminology, you got any top stats (especially like kills or healing), you are most likely going to receive a boost in hidden points. On the other hand, if you are not dealing a lot of damage compared to other players playing a very similar build as you are in your rank in a game, your hidden point is probably going to drop.With the hidden point system in mind, the matchmaking system is most likely going to pair you up with players that are similar in rank, and it is also going to try to keep the average hidden points of one team to be similar to the average hidden points of the other. This is a very efficient way to balance smurfs and alts and in general, players who are stronger or weaker than an average player in their rank.Here's an example: Imagine the hidden point system assigns hidden points to each player with a value between 0 and 100. You are a very strong player in gold 1, and your past performance is taken into account so your hidden point is 80. Now the matchmaking system is going to restrict the search of other players inside this game to be somewhere between silver 3 and gold 2, and the average hidden points of a team should be similar to the average hidden points of the other teams. So imagine you got four teammates paired up with you, all of them are in gold 1 just like you, but they aren't really good so they have 30 hidden points each. Your team will have an average hidden point of around (80+304)/5=40. Lets say your opposing team has an extremely bad player who goes afk a lot, so he has 0 points. In order for the matchmaking system to find a good game, the remaining four players will probably be at around 405/4=50 hidden points each in order to balanced out the average hidden points between your team and theirs. This means, during the game you are going to be the best player but your four teammates with 30 hidden points are going to face people with 50 hidden points, which are much more skillful players. You will then need to perform really well and hard carry the game in order to win.Hence for people already having high hidden points, you are most likely going to be put into a match where you will need to carry. If you win the match because you carried, you are going to have an even higher hidden point not only because you won a match, but also because you most likely will have had great statistics while carrying. If you unfortunately lost the game because you couldn't carry the team, your hidden point will not drop as much because you are still the one that performed the best during the fight, hence adding bonus to your hidden point. The matches will only get easier when your hidden point drops because you are not performing as well in your latest matches or you went on a losing streak so that your hidden points rapidly drops.

Now at the end of this, I do need to repeat that this is just a likelihood of the matchmaking system and is not a confirmed mechanic of the game. The best way to go up the ladder is still improving your skills, and with practice I hope all of you can reach the rank you want :)

Link to comment
Share on other sites

I could see it, but there's really nothing that can't be explained by "sometimes people get lucky or unlucky streaks, especially when mostly playing solo, and this can push people well outside the bracket they should be in".

Personally, I think that if you're soloqueueing, the rank points you gain or lose should be based on the average of your team versus the average of the enemy team, rather than the current system where it's based on your personal rating versus the average rating of the enemy team. The current system means you can get disproportionately punished for games where your team was at a disadvantage (if you were the highest-rating player of a team that had an average rating below the enemy team) and vice versa, which is probably a large part of how people can end up well outside their proper rating. I understand that there needs to be a system to prevent high-ranking players from farming rating points off weaker opponents, but maybe that could be something that starts kicking in at the higher ratings rather than among the people who normally sit somewhere in the middle.

Link to comment
Share on other sites

@phokus.8934 said:This is how conspiracy theorists operate. They want you to prove them wrong and even when you do they’ll still say otherwise.

@Megametzler.5729 said:Also, why would they implement this, requiring new code and all? To cater... whom? To hide... what?

Conspiracy theories are wild.

It's a proven system implemented in some ways by most dominant competitive video games in order to achieves better balance for the team, not just something random. That's probably gonna be my only explanation text about conspiracy: its not. If such system would exist its only better for everyone, not worse. Having different opinion is your freedom but putting conspiracy theorist as a tag before you actually pay attention to what it's saying is quite arrogant, don't you think? @phokus.8934And last time I repeat myself: This is a system that explains why some people are feeling "unfortunate" with their matchmaking because of its attempt to balance the teams, not a system that intentionally tries to screw people up.

@anduriell.6280 said:some spacing and punctuation would do miracles for that literally wall of text.

yeah thats mb, I didn't expect the outcome to be that bad, but I guess I can't fix it now anymore, so it's just gonna go to people who actually are interested in this

@"draxynnic.3719" said:I could see it, but there's really nothing that can't be explained by "sometimes people get lucky or unlucky streaks, especially when mostly playing solo, and this can push people well outside the bracket they should be in".

Personally, I think that if you're soloqueueing, the rank points you gain or lose should be based on the average of your team versus the average of the enemy team, rather than the current system where it's based on your personal rating versus the average rating of the enemy team. The current system means you can get disproportionately punished for games where your team was at a disadvantage (if you were the highest-rating player of a team that had an average rating below the enemy team) and vice versa, which is probably a large part of how people can end up well outside their proper rating. I understand that there needs to be a system to prevent high-ranking players from farming rating points off weaker opponents, but maybe that could be something that starts kicking in at the higher ratings rather than among the people who normally sit somewhere in the middle.

They probably have the same way to calculate ranked points for both solo players and duo players. And if it is based on team's average rating against opponents' average rating, people could duo with someone with a much lower rank and farm points off much weaker opponents.Also about the ELO system, I think it only really works if the proportion of smurfs are so small so its mostly irrelevant. This is definitely not and probably never will be the case for gw2 especially for lower ladder, as there's so many other gamemodes to play and get good in.

Link to comment
Share on other sites

@"Ryo.5824"

I did a lot of research on the matchmaker a while back. I will explain how it works, and then i will explain the consequences of how the matchmaker works leads to statistcal anomalies that describe behavior that you experience in matches.

How The Matchmaker Works

The first thing to understand is that there are a number of individual component parts that make up the matchmaking system. In the code they are described a bit differently, but the best way to think about it is imagining that each component is like a little machine that organizes information. These three machines are called the following :The ServerThe MatchmakerThe Pool

When a person hits the que button, a player enters the que. This que i will be referring to as "The pool" in which all players that are currently in the que are inside. The matchmaker is filing through the players in this pool, organizing potential sets of teams for matches.

Every 30 seconds, the server pings the matchmaker, to initiate a match. If the matchmaker is not ready, by not having assorted a team of players it deems suitable to create a match, it will return an answer to the server (The answer in this case being that the matchmaker wasn't ready.) So the server will continue to wait every 30 seconds, reminding the matchmaker that it has a job to do.

Meanwhile, the Matchmaker is searching in the pool for 10 players to create a match. This search is done by rating and is called "Phase 1."

After the matchmaker finds 10 players within an acceptable range of rating for a match, it will enter "Phase 2" which organizes those 10 players into teams based on a number of parameters. These parameters include rating, rank, games played, party size, profession, and *dishonor.

Once the Matchmaker has organized the teams in what it believes to be satisfactory (in which the matchmaker believes that each side has a fairly balanced distribution from the parameters above in both phase 1 and phase 2), The matchmaker will keep those players in a proverbial "basket", for when the server is ready to ping the matchmaker. The matchmaker can continue the above two phases for the players in this basket if it finds more suitable players in the pool.

Anyway, the server pings the matchmaker, and the matchmaker gives the green light on all the players that are in the "basket," Anyone still in the pool and not in a "basket" will remain in que until those players find a suitable match by the matchmaker.

How Matchmaking Creates Statistical Anomalies

So perhaps you've already noticed how rank (rabbit-dragon), the number of games you've played in your account history, and dishonor can effect how the matchmaker places you on teams. This is enough to describe at least some of the statistical anomalies in the matchmaker, and it makes sense as to why we observe behavior where it would seem that player skill is drastically different from player to player in the same match. For now let's just ignore rating, and just focus on these parameters.

Let's give an example that highlights one of these anomalies.

Let's say you que into a pool of 9 other players. Out of the 9 players in que, you have the most rank (level 300 dragon) and the most games played in your account history (6000 games). Because the matchmaker first organizes players by RATING, you will automatically fall into this group of 9 players. So that means that if the matchmaker decides you belong on Team A because of those parameters in phase 2, then there will be almost no other configuration of players in which you are on a different team with different players, because the matchmaker will strive to make a balanced match. That means if you have the most, it WILL put you with a player that also has the least, because as someone that has the most, you are an outlier and outliers are balanced by this system by just giving you players with lowest value parameters.

Notice how profession also plays into Phase 2. You could have THE LEAST rank and the LEAST games played on your account, and by simply chosing a certain profession before que, you might end up on the team in which you are grouped with players with the least value parameters, because the outlier that was supposed to balance out your team, is playing the same profession as you, and therefor they are placed on the other team.

Now, this is just the anomolies that are caused on a micro-scopic level based on phase 2 of the matchmaker. This is a problem, but there is a more macroscopic anomoly that is much more sinister. It deals with how rating is essentially worthless in the context of the system, because of how anomolies end up shifting the real values of rating into more and more random values.

To explain, let's propose a hypothetical matchmaking scenario.

Player Alice, Bob, Charlie and David exist and for this example lets say that...

  1. Alice is a good player (As good as Charlie)
  2. Bob is a bad player
  3. Charlie is a good player (As good as Alice)
  4. David is a bad player

Now, they all end up in a match together. Alice is on Bob's team (A), and Charlie is on David's team (B). At the end of the game, Team A wins, and so Alice and Bob are rewarded. where as Charlie and David are punished. In this case, we can say that the Matchmaker treats Alice and Bob as being better players then Charlie and David.

The 2nd match comes along, and this time, Alice and David are on team A, while Bob and Charlie are on team B. Team A wins, and so Alice and David are rewarded, while Bob and Charlie are punished and we can say that for this game, the matchmaker treats Alice and David as better players then Bob and Charlie. If we take a look at how these two matches add together we can deduce the following based on the matchmakers decision.

Alice is the best playerBob is an average playerDavid is an average playerCharlie is the worst player

So do you see the problem with the above deduction? Charlie is judged by the matchmaker as being the worst player, while Bob and David, who are supposed to be worse than Charlie, are labeled as better players than Charlie.

So if you drag out this exercise on the order of many iterations, you will find that, the more games that are played, the more randomly distributed the allocation of good and bad players become, where players like David and Bob who are bad players, are lifted into this zone where they are considered average. Thus the most random assignment of rating happens to be the top of a bell curve. In other words, The most common rating range, is statistically the most unknown and random place in the distribution of players rating. In essence, rating means nothing, because in a setting where the win condition is based on the performance of others, your rating is essentially as good as the average rating of the team that you are on. Remember, the matchmaker is BASED on rating. If rating is meaningless, then how can a match possibly be balanced, if the number it's based on can't accurately be indicative of player skill?

In conclusion, Ryo, you are correct that there are hidden variables that determine your place in the matchmaker. Though they aren't exactly how you describe, they are indeed similar, but with no malicious intent. It's just the result of statistical anomalies that end up cascading into large scale, and long term issues with the matchmaker.

Link to comment
Share on other sites

@JusticeRetroHunter.7684 said:@"Ryo.5824"

I did a lot of research on the matchmaker a while back. I will explain how it works, and then i will explain the consequences of how the matchmaker works leads to statistcal anomalies that describe behavior that you experience in matches.

How The Matchmaker Works

The first thing to understand is that there are a number of individual component parts that make up the matchmaking system. In the code they are described a bit differently, but the best way to think about it is imagining that each component is like a little machine that organizes information. These three machines are called the following :The ServerThe MatchmakerThe Pool

When a person hits the que button, a player enters the que. This que i will be referring to as "The pool" in which all players that are currently in the que are inside. The matchmaker is filing through the players in this pool, organizing potential sets of teams for matches.

Every 30 seconds, the server pings the matchmaker, to initiate a match. If the matchmaker is not ready, by not having assorted a team of players it deems suitable to create a match, it will return an answer to the server (The answer in this case being that the matchmaker wasn't ready.) So the server will continue to wait every 30 seconds, reminding the matchmaker that it has a job to do.

Meanwhile, the Matchmaker is searching in the pool for 10 players to create a match. This search is done by rating and is called "Phase 1."

After the matchmaker finds 10 players within an acceptable range of rating for a match, it will enter "Phase 2" which organizes those 10 players into teams based on a number of parameters. These parameters include rating, rank, games played, party size, profession, and *dishonor.

Once the Matchmaker has organized the teams in what it believes to be satisfactory (in which the matchmaker believes that each side has a fairly balanced distribution from the parameters above in both phase 1 and phase 2), The matchmaker will keep those players in a proverbial "basket", for when the server is ready to ping the matchmaker. The matchmaker can continue the above two phases for the players in this basket if it finds more suitable players in the pool.

Anyway, the server pings the matchmaker, and the matchmaker gives the green light on all the players that are in the "basket," Anyone still in the pool and not in a "basket" will remain in que until those players find a suitable match by the matchmaker.

How Matchmaking Creates Statistical Anomalies

So perhaps you've already noticed how rank (rabbit-dragon), the number of games you've played in your account history, and dishonor can effect how the matchmaker places you on teams. This is enough to describe at least some of the statistical anomalies in the matchmaker, and it makes sense as to why we observe behavior where it would seem that player skill is drastically different from player to player in the same match. For now let's just ignore rating, and just focus on these parameters.

Let's give an example that highlights one of these anomalies.

Let's say you que into a pool of 9 other players. Out of the 9 players in que, you have the most rank (level 300 dragon) and the most games played in your account history (6000 games). Because the matchmaker first organizes players by RATING, you will automatically fall into this group of 9 players. So that means that if the matchmaker decides you belong on Team A because of those parameters in phase 2, then there will be almost no other configuration of players in which you are on a different team with different players, because the matchmaker will strive to make a balanced match. That means if you have the most, it WILL put you with a player that also has the least, because as someone that has the most, you are an outlier and outliers are balanced by this system by just giving you players with lowest value parameters.

Notice how profession also plays into Phase 2. You could have THE LEAST rank and the LEAST games played on your account, and by simply chosing a certain profession before que, you might end up on the team in which you are grouped with players with the least value parameters, because the outlier that was supposed to balance out your team, is playing the same profession as you, and therefor they are placed on the other team.

Now, this is just the anomolies that are caused on a micro-scopic level based on phase 2 of the matchmaker. This is a problem, but there is a more macroscopic anomoly that is much more sinister. It deals with how rating is essentially worthless in the context of the system, because of how anomolies end up shifting the real values of rating into more and more random values.

To explain, let's propose a hypothetical matchmaking scenario.

Player Alice, Bob, Charlie and David exist and for this example lets say that...

  1. Alice is a good player (As good as Charlie)
  2. Bob is a bad player
  3. Charlie is a good player (As good as Alice)
  4. David is a bad player

Now, they all end up in a match together. Alice is on Bob's team (A), and Charlie is on David's team (B). At the end of the game, Team A wins, and so Alice and Bob are rewarded. where as Charlie and David are punished. In this case, we can say that the Matchmaker treats Alice and Bob as being better players then Charlie and David.

The 2nd match comes along, and this time, Alice and David are on team A, while Bob and Charlie are on team B. Team A wins, and so Alice and David are rewarded, while Bob and Charlie are punished and we can say that for this game, the matchmaker treats Alice and David as better players then Bob and Charlie. If we take a look at how these two matches add together we can deduce the following based on the matchmakers decision.

Alice is the best playerBob is an average playerDavid is an average playerCharlie is the worst player

So do you see the problem with the above deduction? Charlie is judged by the matchmaker as being the worst player, while Bob and David, who are supposed to be worse than Charlie, are labeled as better players than Charlie.

So if you drag out this exercise on the order of many iterations, you will find that, the more games that are played, the more randomly distributed the allocation of good and bad players become, where players like David and Bob who are bad players, are lifted into this zone where they are considered average. Thus the most random assignment of rating happens to be the top of a bell curve. In other words, The most common rating range, is statistically the most unknown and random place in the distribution of players rating. In essence, rating means nothing, because in a setting where the win condition is based on the performance of others, your rating is essentially as good as the average rating of the team that you are on. Remember, the matchmaker is BASED on rating. If rating is meaningless, then how can a match possibly be balanced, if the number it's based on can't accurately be indicative of player skill?

In conclusion, Ryo, you are correct that there are hidden variables that determine your place in the matchmaker. Though they aren't exactly how you describe, they are indeed similar, but with no malicious intent. It's just the result of statistical anomalies that end up cascading into large scale, and long term issues with the matchmaker.

I wont read all of this but number of matches played and pvp level doesnt matter for matchmaking.

Link to comment
Share on other sites

@"Filip.7463" said:I wont read all of this but number of matches played and pvp level doesnt matter for matchmaking.

Wow. The delusion is real.

This is how the matchmaker works, ranking level and number of matches are involved when assorting the teams in the matchmaking process. Confirmed by the PVP Matchmaking Systems Dev himself. He was the one who also wrote the wiki page.

Source -Justin O'Dell Dev Posthttps://www.guildwars2.com/en/news/finding-the-perfect-match/

https://wiki.guildwars2.com/wiki/PvP_Matchmaking_Algorithm

Link to comment
Share on other sites

@Ryo.5824 said:

@phokus.8934 said:This is how conspiracy theorists operate. They want you to prove them wrong and even when you do they’ll still say otherwise.

@Megametzler.5729 said:Also, why would they implement this, requiring new code and all? To cater... whom? To hide... what?

Conspiracy theories are wild.

It's a proven system implemented in some ways by
most
dominant competitive video games in order to achieves better balance for the team, not just something random. That's probably gonna be my only explanation text about conspiracy: its not. If such system would exist its only better for everyone, not worse. Having different opinion is your freedom but putting conspiracy theorist as a tag before you actually pay attention to what it's saying is quite arrogant, don't you think? @phokus.8934And last time I repeat myself: This is a system that explains why some people are feeling "unfortunate" with their matchmaking because of its attempt to balance the teams, not a system that intentionally tries to screw people up.

@anduriell.6280 said:some spacing and punctuation would do miracles for that literally wall of text.

yeah thats mb, I didn't expect the outcome to be that bad, but I guess I can't fix it now anymore, so it's just gonna go to people who actually are interested in this

@"draxynnic.3719" said:I could see it, but there's really nothing that can't be explained by "sometimes people get lucky or unlucky streaks, especially when mostly playing solo, and this can push people well outside the bracket they should be in".

Personally, I think that if you're soloqueueing, the rank points you gain or lose should be based on the average of your team versus the average of the enemy team, rather than the current system where it's based on your personal rating versus the average rating of the enemy team. The current system means you can get disproportionately punished for games where your team was at a disadvantage (if you were the highest-rating player of a team that had an average rating below the enemy team) and vice versa, which is probably a large part of how people can end up well outside their proper rating. I understand that there needs to be a system to prevent high-ranking players from farming rating points off weaker opponents, but maybe that could be something that starts kicking in at the higher ratings rather than among the people who normally sit somewhere in the middle.

They probably have the same way to calculate ranked points for both solo players and duo players. And if it is based on team's average rating against opponents' average rating, people could duo with someone with a much lower rank and farm points off much weaker opponents.Also about the ELO system, I think it only really works if the proportion of smurfs are so small so its mostly irrelevant. This is definitely not and probably never will be the case for gw2 especially for lower ladder, as there's so many other gamemodes to play and get good in.Simply put you don’t know how Glicko-2 scores nor how the GW2 matchmaker works. You want there to be something hidden behind it because you and countless others get terrible games. You’re trying to find something that isn’t there. And refer to the MMR wiki as that explains everything in detail.

Link to comment
Share on other sites

@phokus.8934 said:Simply put you don’t know how Glicko-2 scores nor how the GW2 matchmaker works. You want there to be something hidden behind it because you and countless others get terrible games. You’re trying to find something that isn’t there. And refer to the MMR wiki as that explains everything in detail.

You mean refer people to the source where it it explains that what Ryo said is essentially true and that you are wrong? lmao Did you even read or understand the code in the wiki?

The code in the wiki literally explains how it uses a hidden scoring process to evaluate what team people end up on in the matchmaker.

Link to comment
Share on other sites

Nothing of that, population is simply low.

If wanna rank consistently and you are decent/good player you need to duo.

SoloQ is tough and the twilight zone is gold1/2/3 where the majority of bots are.

You might be a platinum skill level guy, but you keep being matches with bots which drop you down inevitably to lower ranks.

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...