Overwatch 2 Developer Blog: Explaining matchmaker goals and plans, part 2

Welcome back, heroes, for part 2 of our matchmaking series! In part 1, we talked about how matchmaking works: It’s an incredibly complex system that needs to consider many different inputs to make every single game as fair as possible. For part 2, we’re going to address some community concerns regarding the matchmaker from the perspective of the competitive game mode. Let’s get into it! 

TL/DR

• Your ranked games are formed based on your internal matchmaking rating (MMR), regardless of your displayed skill tier. MMR changes based on the result of each match, with the amount of MMR you gain or lose depending on several factors like how highly rated your opponents were or how recently you last played. 
• Our team is aware of some community pain points with competitive and matchmaking. We have a set of improvements coming to matchmaking, and we’re also working on updates to the competitive system.
• For competitive play in Season 3, we’ll reduce the number of wins between competitive updates from 7 to 5 wins or 20 to 15 losses. For our matchmaker, we’re implementing role delta changes that will match similarly ranked players within each role.
• Season 4 will include additional information about your current wins and losses on the competitive update screens. Our team has begun work on longer-term features and updates for both the matchmaker and competitive game mode, and we look forward to sharing more on those plans in the future!

Here’s what we’re currently seeing in ranked

We’ve seen your feedback on matches with wide skill variation, and we want to talk about a few reasons that could be happening as well as our plans to address your concerns.

We allow players to group together even when there’s a difference in the MMR between party members, and this is a major source of wide skill disparities in matches. We’re working on some changes that will match parties with similar MMR disparities together more frequently, which we expect to noticeably reduce how often we make wide matches. This will make it much less likely for a solo player or a party with a narrow skill disparity to end up in a match with a wide skill disparity.

Often matches with a wide displayed skill tier difference still look very close when looking at the difference in MMR between the two teams. The partial rank reset at the beginning of the season may be exaggerating this by making it look like someone is lower than their actual rank. However, regardless of the seasonal reset, both skill tier and MMR can decay over time for players who are inactive for a considerable amount of time. Players who return may see dramatic shifts to their skill tiers as they resume playing regularly, and the game becomes more certain of that player’s current skill.

We’ve also seen the community talk about inconsistent games or how some games feel incredibly one-sided, and that feedback has been helpful. Understanding what causes one-sided matches and how to reduce their frequency has become a major area of focus for our team. This is a tricky problem because one-sided matches can happen in Overwatch, even between balanced teams. So, our first step has been to study the problem and understand the various factors that can cause one-sided matches, and we’re planning to share what we learn with you in a developer blog down the road.

Competitive updates are another topic we’ve been exploring. We want you to feel a sense of progression without needing to focus on a super granular number like a skill rating, but we feel, and agree with the community, that competitive updates are a bit too infrequent currently.

Here’s the direction we’re taking moving forward

What we’re doing in Season 3

Based on community feedback, we’re implementing some updates to the matchmaker that try to place pairs of players with similar MMR on each role on either team. This means that opposing tanks should be more similarly matched than before and likewise for the other respective roles. The goal with this change is to make the average MMR between each role more evenly matched to each other instead of looking more broadly across the entire team to balance things out. This is a major change to the matchmaker, so we’ll be actively monitoring for any unintended side effects.

We’ve seen feedback that having to play as many as 26 matches to get an update has been discouraging for players. Starting with Season 3, you’ll now get a competitive update with every 5 wins and 15 losses. In the mid-season patch for Season 3, we’re also updating the UI, so information about your progress toward a competitive update will always be viewable.

Also, beginning with our mid-season patch, Top 500 players will see their Top 500 leaderboard rank updated after every match rather than in competitive updates.

What we’re doing in the near future

We’ve heard your feedback and confusion around seasonal rank decay and rank resets. Starting with Season 4, we’re planning to remove seasonal competitive rank resets and all current and past seasonal rank decay.

We’ll continue to build out competitive updates in Season 4 when we add information about your current wins and losses to the competitive update screens. We’re hoping this additional context will help you better interpret changes to your skill tier and division in the update.

We’re also looking into possible ways to provide you with more information about the matchmaking quality of your game.

Long-term plans and goals

In the long term, we’re planning to provide new ways for players to share a challenging and competitive experience in Overwatch 2—we’ll share more on this down the road. That said, matchmaking improvements are an evergreen area of focus for the team, and we’ll stay focused on testing and deploying continuous improvements to our matchmaking algorithm.

Finally, we’ll continue to watch for your feedback and improve upon the overall ranked experience over time. Our goal continues to be delivering matchmaking that feels fair and balanced for everyone.

Closing thoughts

The goal of our matchmaker is to make each match as fair as possible, meaning your team has an equal chance of winning or losing each individual match. We firmly believe the most fun games of Overwatch are when they are fair, and our team is committed to discovering and implementing new changes and updates that align our game with that goal.

Overwatch 2 is a continually evolving experience with new game modes, new heroes, and new maps to explore. This also means we need to provide a constantly updated experience that can deliver the fairest possible matches through effective changes to how our matchmaker works and through clear communication with all of you.

We look forward to seeing all of you on the battlefield!

Frequently Asked Questions

Q: Do you deliberately place players into winner queues and loser queues?
A: There are no winner or loser queues in Overwatch. Your current MMR is the only thing the matchmaker takes into consideration when forming your matches. The matchmaker doesn’t force a 50% win rate on anyone, nor do we favor certain players over others.

At its core, our matchmaker is a mathematical formula used to place similarly ranked players together in a match with the goal of creating as fair of games as possible.

Q: Why do I sometimes have a large win or loss streak?
A: Sometimes, if a player goes on a very long win/loss streak, it’s indicative that their internal rating is not well-calibrated. The best way to calibrate your rank is to continue playing competitively. The more data we have, the closer you’ll get to a rank that best represents your skill. However, there are times when players are going to get lucky with their win streaks or the opposite with loss streaks.

We’re investigating this and will be testing some changes to see if we can reduce this kind of streaking in the future.

Q: Does my competitive rank affect matchmaking?
A: We only create matches based on players’ MMR, not the visible competitive rank. A player’s visible rank will move towards their rating over time as they continue to play during a season. When we decay ranks at the beginning of a season, this has no effect on players’ underlying ratings.

Q: Why does the quality of my matches go down at the start of a season?
A: When Season 2 went live, a lot of players who stopped playing during Season 1 came back, so the population of players playing ranked changed dramatically overnight. Ratings are a measure of your skill relative to the rest of the players you interact with. Events that cause the population to change a lot will result in some turbulence. We’re exploring some ideas for how to tackle this in the future!

Q: Can we do a full rating reset?
A: A full rating reset wouldn't create a great experience since it would mean throwing out all the knowledge we have about players. This would cause new players to be matched against OWL pros, which is fun for about 30 seconds (we’ve experienced this ourselves in internal playtests).

To be clear, ratings aren’t a 100% accurate representation of every player’s skill. Some players aren't in the right spot, but the system is a vast store of knowledge about the relationship in skill between millions of players, most of whom have never even directly played against one another.

Q: Why am I hard stuck at a low rating that doesn’t reflect my skill?
A: Ultimately, the only way you can increase your rating is to win more matches than you lose. Since we’re putting any given player on a random team with 9 random other players, with enough matches, that player’s contributions are the only constant factor throughout, so their rating should end up reflecting their skill.

There are some interesting phenomena here, though, that make this more complicated. For example, we have data showing that matches become less predictable the lower the average rating of a match. Predictability is a measure of how likely the favored team is to win because two teams never have identical ratings. One team is always going to be slightly favorite for a variety of reasons: team coordination, gameplay mistakes, and new players who are trying out different roles and 30+ heroes. The result is that we get a little bit less signal to use to calibrate ratings. We’re working on some ways to use the signal we get more clearly, so stay tuned!

Q: So you don’t take the number of eliminations, damage dealt, healing provided, or any other scoreboard stats to adjust my MMR after each match?
A: In Overwatch 2, your MMR adjustment after every match is not impacted by your performance in each match (regardless of your skill tier). This is for a few reasons. We don’t want players to be focused on doing things other than trying to focus on the objectives and win the match. Dealing the most damage or getting kills won’t help your team if your actions don’t help them push the payload or capture a control point. Also, for some heroes, especially those in the support role, it can be challenging to determine if the numbers they produce reflect their skill.

Q: Why do streamers and pro players often have such long queue times and often end up in poor-quality matches?
A: Perhaps the largest factor that influences match quality is just the population of players at that skill level. Top 500 is a small population compared to the millions of players at lower ranks.

Towards the center of the rating curve, there are tons of players in the matchmaking queue for us to choose from, all of whom would allow for a good match, so it’s much easier for us to ensure a good quality match.

Contrast this with OWL pros, who make up something like 0.00001% of the Overwatch player population. For players at the very top, it’s very hard to guarantee a good experience because there are just so few players. Fundamentally it often comes down to a hard tradeoff between making worse matches or dramatically increasing queue times (and they’re unfortunately already very long for these players).

We have some changes coming that should improve match quality for high MMR players, and we’re talking about even more sophisticated things we could do in the future.