Matchmaking Systems Explained: What NetEase's Research Tells Us

[Shnoofn]

After seeing many discussions about matchmaking in this forum, I want to share some documented facts about different matchmaking approaches and what we know from NetEase's own published research. This post aims to provide context for the experiences many players have reported.

Two Competing Matchmaking Philosophies

Modern competitive games generally use one of two approaches to matchmaking:

1. Skill-Based Matchmaking (SBMM)

The traditional approach focused on creating fair, balanced matches.

Key features:

• Primary goal is competitive fairness
  
• Matches players of similar skill levels
  
• Uses objective performance metrics (MMR, ELO, K/D, etc.)
  
• Aims to give all players roughly equal chances of winning based on skill

2. Engagement Optimized Matchmaking (EOMM)

A newer approach that prioritizes player retention and engagement.

Key features:

• Primary goal is maximizing player engagement and playtime
  
• May create deliberate patterns of wins and losses
  
• Can prioritize "player satisfaction" over strict skill parity
  
• Uses AI and data analysis to predict what keeps players playing longer

The fundamental difference? SBMM focuses on matching equal skills, while EOMM focuses on creating engaging experiences - even if that sometimes means uneven matches.

NetEase's Published Research on Engagement Optimization

NetEase (the developer of Marvel Rivals) has published significant research on engagement-optimized matchmaking systems:

OptMatch (2020)

In their paper "OptMatch: Optimized Matchmaking via Modeling the High-Order Interactions on the Arena"[nos.netease.com] (KDD 2020), NetEase researchers explicitly state:

"The goal of matchmaking is to find the optimal assignments for players to maximize the total utility (i.e., the gross players' satisfaction)." (p. 2)

They define "player utility" as "a quantitative measure of a player's satisfaction gained from a match, for example, the reverse of churn risk for this player after a certain match."

In other words, their system aims to keep players from quitting (churning), rather than focusing exclusively on fair matches.

EnMatch (2024)

In their most recent paper "EnMatch: Matchmaking for Better Player Engagement via Neural Combinatorial Optimization"[ojs.aaai.org] (AAAI 2024), NetEase directly challenges the assumption that fair matches are always best:

"Previous methods focus on creating fair games at all times... Though this strategy can ensure fair matchmaking, it is not always good for player engagement." (Abstract)

They state this even more clearly:

"Is game fairness the only critical factor for player engagement? In most matchmaking scenes, the answer is no." (p. 1)

Their research found that "diverse win-loss experiences" (not just fair matches) lead to higher engagement, and they explicitly develop systems to create these varied experiences rather than consistently fair matches.

What This Could Mean For Players

If a game uses an engagement-optimized system like those described in NetEase's research, players might experience:

• Intentional win and loss streaks designed to keep you playing longer
  
• Matches that feel unbalanced by design, not just by chance
  
• Teams composed to create specific engagement patterns, not just based on skill
  
• The feeling that your performance isn't the main factor in your match outcomes

Many players in this forum have reported exactly these experiences.

Why This Matters: The Integrity of Competitive Gaming

This issue goes beyond just game mechanics - it touches on fundamental aspects of competitive gaming as a legitimate form of sport:

1. E-Sports Legitimacy: If we want competitive gaming to be recognized as a legitimate sport (e-sport), it must adhere to the same principles of fair competition that govern traditional sports. No traditional sport would remain credible if outcomes were manipulated behind the scenes to maximize audience engagement.
   
   
2. Clear and Transparent Rules: Every legitimate sport operates with clear, transparent rules that all participants understand. In basketball, football, or chess, the rules are explicit and equally applied. Hidden algorithms that manipulate matchmaking to create specific patterns of outcomes run counter to this fundamental principle of sporting competition.
   
   
3. Product Representation: When you purchase a game marketed as "competitive," you're entering into a consumer agreement with certain expectations. If you're sold a competitive game but given what amounts to a "competition simulator" with pre-determined patterns of outcomes, that's a significant discrepancy between what was advertised and what was delivered.
   
   
4. Competitive Integrity: True competition requires that outcomes are determined by skill, strategy, and execution - not by behind-the-scenes algorithms designed to manipulate player engagement. This is especially important in a game that positions itself as having a competitive esports future.
   
   
5. Player Agency: Athletes improve through practice, strategy, and skill development - with the expectation that these improvements will be reflected in their results. Systems that artificially manipulate outcomes undermine this core principle of player agency and the value of improvement.

Think of it this way: We would never accept a basketball league where referees were secretly instructed to make calls that ensure certain teams win or lose in patterns designed to maximize viewer engagement. Why should we accept this in competitive gaming?

What We Don't Know For Certain

While NetEase has clearly invested in developing engagement-optimized matchmaking, there are things we don't know:

• Is this specific system used in Marvel Rivals? There's no official confirmation, though the published research shows it's technology NetEase has developed.
  
• How strongly does it prioritize engagement over fairness? The balance could vary by game.
  
• What specific factors does it consider? The papers mention various potential inputs but don't detail exact implementations.

How Other Games Approach Matchmaking

For context, other competitive games have been more transparent about their matchmaking:

Riot Games (Valorant) has published detailed blog posts explaining their matchmaking goals, emphasizing fair, skill-based matches.

Blizzard (Overwatch 2) has released developer blogs explaining their matchmaking system, specifically stating they don't use "winner/loser queues."

Both companies provided meaningful transparency without revealing their proprietary algorithms.

A Direct Request to NetEase Developers

If NetEase developers are reading this post (and I hope you are), I respectfully request an official statement addressing these questions:

• Does Marvel Rivals use engagement-optimized matchmaking systems similar to those described in your published research papers?
  
  
• If so, how are these systems balanced with skill-based matchmaking, particularly in competitive/ranked modes?
  
  
• Would you consider providing the level of transparency about matchmaking that other competitive game developers like Riot and Blizzard have offered to their communities?
  
  
• How do you ensure competitive integrity in a game that may use engagement optimization?

This isn't about criticism - it's about clarity. Your research in this area is impressive and innovative, but as players investing our time in what we hope is a fair competitive experience, we deserve to understand the fundamental systems that determine our matches.

What We As Players Can Do

As a community that values competitive gaming as a legitimate sport, we could:

• Continue to respectfully request transparency about matchmaking philosophies
  
• Share experiences methodically to identify patterns (with replay IDs where possible)
  
• Advocate for clear separation between casual modes (where engagement optimization might be appropriate) and competitive/ranked modes (where skill-based fairness should be paramount)
  
• Establish community standards for what constitutes fair competition in e-sports

Conclusion

NetEase has demonstrably invested in developing engagement-optimized matchmaking systems that prioritize player retention over consistently fair matches. Their own research explicitly states that fair matches "are not always good for player engagement."

Without an official statement, we can't know exactly how Marvel Rivals implements matchmaking. However, the published research provides valuable context for understanding the matchmaking philosophies that could be at play.

As we push for gaming to be recognized as a legitimate form of competitive sport, we should insist on the same standards of fairness, transparency, and competitive integrity that we expect in traditional sports. A truly competitive e-sport requires not just balanced characters and good maps, but matchmaking systems that create genuinely fair competitions based on skill - not algorithms designed to manipulate engagement.

Note: All information in this post comes from publicly available sources. The quoted papers were published by NetEase engineers themselves and are accessible through scientific databases.

[JG2thend]

https://steamcommunity.com/sharedfiles/filedetails/?id=3422002988

[BoomsticK]

not going to happen buddy, these devs are seeking maxium $$ from skins

[Steve]

Good, you're finally asking the devs. Let us know how that goes.

[Terotrous]

The OptMatch paper also states that they calculate the utility of a match as the chance of that match ending in a tie, so if they're sticking to that then there's not much issue, but it's fair to note that the EnMatch paper is much newer and no longer makes that statement.

[Sheiriff Woody]

You know, you could just ask Google what MM marval Rivals uses, and Google says...

"According to NetEase, Marvel Rivals uses skill-based matchmaking (SBMM) to create balanced games. This system looks at several stats to determine your skill level. In Competitive mode, your visible rank (from Bronze to One Above All) determines your matchmaking."

That could have saved you alot of time Editing this College style essays that wasn't needed. You could have cured Cancer with that time...Just saying.

Every Time somebody says that EOMM did this, that and this, can be tracked back to the habbits of SBMM, TSBMM, RBMM and LBMM. All with the Same flaws, uneven matches and Win/Loss Streaks. its lit identical like Twins.

[Terotrous]

I do think it's quite likely that most of the stuff they're talking about in terms of matchmaking only applies to QP. For Ranked, the matchmaking seems quite straightforward - it just matches you based on your visible rank, seemingly without many adjustments (I've definitely had matches where I've had certain players on my team one game, then later played against them).

Of course, we do know that the system performs certain adjustments to rank gain / loss. Chrono Shields, performance-based ranking adjustments (ie, you gain more points if you're MVP), generally gaining more points when you lose in low ranks, etc. This is likely how they implemented the "engagement" systems when it comes to rank rather than through matchmaking.

[Karnarax]

TL;DR any1?

[ColdWonder]

Originally posted by Karnarax:

TL;DR any1?

MM sucks because it's more interested in keeping you engaged rather than fair matches .

[Karnarax]

Originally posted by ColdWonder:

Originally posted by Karnarax:

TL;DR any1?

MM sucks because it's more interested in keeping you engaged rather than fair matches .

Heard that one before... Thought, maybe, there is something new now. Thank you though!

[Steve]

Originally posted by Karnarax:

Originally posted by ColdWonder:

MM sucks because it's more interested in keeping you engaged rather than fair matches .

Heard that one before... Thought, maybe, there is something new now. Thank you though!

No, this is absolutely nothing new. This is literally all rehash from several threads.

The only really notable part is that none of it's ChatGPT.

[VillianByNecessity]

The problem is most of these people complaining about fair ratings would be hard stuck in silver/gold with 20% win rate at best losing their minds on the forums about that instead.

[AvidExpert]

american gvt did lots of R&D into mind controlling it's population, tons of papers published on it too.

america is mind controlling it people, wake up sheeple

[﷽﷽﷽﷽﷽﷽﷽﷽﷽﷽]

Originally posted by Shnoofn:

After seeing many discussions about matchmaking in this forum, I want to share some documented facts about different matchmaking approaches and what we know from NetEase's own published research. This post aims to provide context for the experiences many players have reported.

Two Competing Matchmaking Philosophies

Modern competitive games generally use one of two approaches to matchmaking:

1. Skill-♥♥♥♥♥ Matchmaking (SBMM)

The traditional approach focused on creating fair, balanced matches.

Key features:

• Primary goal is competitive fairness
  
• Matches players of similar skill levels
  
• Uses objective performance metrics (MMR, ELO, K/D, etc.)
  
• Aims to give all players roughly equal chances of winning ♥♥♥♥♥ on skill

2. Engagement Optimized Matchmaking (EOMM)

A newer approach that prioritizes player retention and engagement.

Key features:

• Primary goal is maximizing player engagement and playtime
  
• May create deliberate patterns of wins and losses
  
• Can prioritize "player satisfaction" over strict skill parity
  
• Uses AI and data analysis to predict what keeps players playing longer

The fundamental difference? SBMM focuses on matching equal skills, while EOMM focuses on creating engaging experiences - even if that sometimes means uneven matches.

NetEase's Published Research on Engagement Optimization

NetEase (the developer of Marvel Rivals) has published significant research on engagement-optimized matchmaking systems:

OptMatch (2020)

In their paper "OptMatch: Optimized Matchmaking via Modeling the High-Order Interactions on the Arena"[nos.netease.com] (KDD 2020), NetEase researchers explicitly state:

"The goal of matchmaking is to find the optimal assignments for players to maximize the total utility (i.e., the gross players' satisfaction)." (p. 2)

They define "player utility" as "a quantitative measure of a player's satisfaction gained from a match, for example, the reverse of churn risk for this player after a certain match."

In other words, their system aims to keep players from quitting (churning), rather than focusing exclusively on fair matches.

EnMatch (2024)

In their most recent paper "EnMatch: Matchmaking for Better Player Engagement via Neural Combinatorial Optimization"[ojs.aaai.org] (AAAI 2024), NetEase directly challenges the assumption that fair matches are always best:

"Previous methods focus on creating fair games at all times... Though this strategy can ensure fair matchmaking, it is not always good for player engagement." (Abstract)

They state this even more clearly:

"Is game fairness the only critical factor for player engagement? In most matchmaking scenes, the answer is no." (p. 1)

Their research found that "diverse win-loss experiences" (not just fair matches) lead to higher engagement, and they explicitly develop systems to create these varied experiences rather than consistently fair matches.

What This Could Mean For Players

If a game uses an engagement-optimized system like those described in NetEase's research, players might experience:

• Intentional win and loss streaks designed to keep you playing longer
  
• Matches that feel unbalanced by design, not just by chance
  
• Teams composed to create specific engagement patterns, not just ♥♥♥♥♥ on skill
  
• The feeling that your performance isn't the main factor in your match outcomes

Many players in this forum have reported exactly these experiences.

Why This Matters: The Integrity of Competitive Gaming

This issue goes beyond just game mechanics - it touches on fundamental aspects of competitive gaming as a legitimate form of sport:

1. E-Sports Legitimacy: If we want competitive gaming to be recognized as a legitimate sport (e-sport), it must adhere to the same principles of fair competition that govern traditional sports. No traditional sport would remain credible if outcomes were manipulated behind the scenes to maximize audience engagement.
   
   
2. Clear and Transparent Rules: Every legitimate sport operates with clear, transparent rules that all participants understand. In basketball, football, or chess, the rules are explicit and equally applied. Hidden algorithms that manipulate matchmaking to create specific patterns of outcomes run counter to this fundamental principle of sporting competition.
   
   
3. Product Representation: When you purchase a game marketed as "competitive," you're entering into a consumer agreement with certain expectations. If you're sold a competitive game but given what amounts to a "competition simulator" with pre-determined patterns of outcomes, that's a significant discrepancy between what was advertised and what was delivered.
   
   
4. Competitive Integrity: True competition requires that outcomes are determined by skill, strategy, and execution - not by behind-the-scenes algorithms designed to manipulate player engagement. This is especially important in a game that positions itself as having a competitive esports future.
   
   
5. Player Agency: Athletes improve through practice, strategy, and skill development - with the expectation that these improvements will be reflected in their results. Systems that artificially manipulate outcomes undermine this core principle of player agency and the value of improvement.

Think of it this way: We would never accept a basketball league where referees were secretly instructed to make calls that ensure certain teams win or lose in patterns designed to maximize viewer engagement. Why should we accept this in competitive gaming?

What We Don't Know For Certain

While NetEase has clearly invested in developing engagement-optimized matchmaking, there are things we don't know:

• Is this specific system used in Marvel Rivals? There's no official confirmation, though the published research shows it's technology NetEase has developed.
  
• How strongly does it prioritize engagement over fairness? The balance could vary by game.
  
• What specific factors does it consider? The papers mention various potential inputs but don't detail exact implementations.

How Other Games Approach Matchmaking

For context, other competitive games have been more transparent about their matchmaking:

Riot Games (Valorant) has published detailed blog posts explaining their matchmaking goals, emphasizing fair, skill-♥♥♥♥♥ matches.

Blizzard (Overwatch 2) has released developer blogs explaining their matchmaking system, specifically stating they don't use "winner/loser queues."

Both companies provided meaningful transparency without revealing their proprietary algorithms.

A Direct Request to NetEase Developers

If NetEase developers are reading this post (and I hope you are), I respectfully request an official statement addressing these questions:

• Does Marvel Rivals use engagement-optimized matchmaking systems similar to those described in your published research papers?
  
  
• If so, how are these systems balanced with skill-♥♥♥♥♥ matchmaking, particularly in competitive/ranked modes?
  
  
• Would you consider providing the level of transparency about matchmaking that other competitive game developers like Riot and Blizzard have offered to their communities?
  
  
• How do you ensure competitive integrity in a game that may use engagement optimization?

This isn't about criticism - it's about clarity. Your research in this area is impressive and innovative, but as players investing our time in what we hope is a fair competitive experience, we deserve to understand the fundamental systems that determine our matches.

What We As Players Can Do

As a community that values competitive gaming as a legitimate sport, we could:

• Continue to respectfully request transparency about matchmaking philosophies
  
• Share experiences methodically to identify patterns (with replay IDs where possible)
  
• Advocate for clear separation between casual modes (where engagement optimization might be appropriate) and competitive/ranked modes (where skill-♥♥♥♥♥ fairness should be paramount)
  
• Establish community standards for what constitutes fair competition in e-sports

Conclusion

NetEase has demonstrably invested in developing engagement-optimized matchmaking systems that prioritize player retention over consistently fair matches. Their own research explicitly states that fair matches "are not always good for player engagement."

Without an official statement, we can't know exactly how Marvel Rivals implements matchmaking. However, the published research provides valuable context for understanding the matchmaking philosophies that could be at play.

As we push for gaming to be recognized as a legitimate form of competitive sport, we should insist on the same standards of fairness, transparency, and competitive integrity that we expect in traditional sports. A truly competitive e-sport requires not just balanced characters and good maps, but matchmaking systems that create genuinely fair competitions ♥♥♥♥♥ on skill - not algorithms designed to manipulate engagement.

Note: All information in this post comes from publicly available sources. The quoted papers were published by NetEase engineers themselves and are accessible through scientific databases.

The easiest way to see Marvel Rivals using EOMM is win and losing streak
Check your own games people

In your winning streak games y
Your teammates will be better than enemies in all terms
Your Teammates will be on winning streak or in start of winning streak
Your Teammates will have better Total Average WinRate or Hero WinRate

In your losing streak games
Your teammates will be worse off than enemies in all terms
Your Teammates will be on losing streak or in start of losing streak
Your Teammates will have worse Total Average WinRate or Hero WinRate

This rule is even valid for pro players account in Celestial and even work for other games that uses EngagementOptimizedMatchMaking such as Dota 2 which I used to analyse and my tool got banned by the Dota 2 devs :)

You can make an matchmaking fairness detector using the same logic above like I did its accurate but lets see if Marvel Rivals devs will ban mine :)

[Vanilla Trance]

I'm not reading all that, and you would be better off actually playing the game and getting better instead of typing this non sense lmao

[GatlingCombo]

People saying "I won't read" and "this looks like a college paper". Holy crap humanity sure has come a long way. This is a paper WRITTEN for NetEase. If this were Ubisoft or Activision Blizzard, or Capcom (maybe), people would care a lot more. But not their sweet netease.