Conversations Around The Community: Matchmaking
As the Heroes of the Storm closed beta rolls on and the game continues to evolve, we’d like to provide our perspective on some of the bigger topics we’ve seen discussed in the community and forums. For this blog, we’re looking at the subject of Matchmaking. There are a number of things to chat about, so let’s dive in.
Veterans vs. Newer Players
We’ve noticed a few concerns around the community regarding veterans and newer players being matched up together. We understand the community’s frustration. It’s not a solid experience to be matched against, or on the same team as players of drastically different skill level.
With the game being young and in Closed Beta, our matchmaker is designed to weigh a player’s MMR more than their overall experience and time in the game. As Heroes of the Storm matures, and we bring more players into the game, we'll shift our matchmaker toward catering to more experienced players. This should make games feel much better for the veterans.
On the flip side, we want to take care of our new players. We’re looking into ways that we can help the new player adjust and more quickly find their correct MMR tier. It currently takes 80-100 games, on average, to accurately establish a player’s MMR. This means a newer player could be tossed around in games that could be way above their skill level, ultimately hurting their overall match experience. We’ll continue to evaluate and find solutions.
With that in mind, we do want to stress that each mode in Heroes of the Storm has a separate MMR. The matchmaker will use MMR to find an even match, rather than the visible rankings on a profile. Being a high-level player does not necessarily mean a player has a high skill ceiling (and vice versa), so consider this information when transitioning to different modes.
Team Compositions in Quick Match
A number of players have requested changes to our Quick Match system in order to force more balanced matchups by team composition. When looking at feedback, the most common requests were around the idea of forcing a Support or Warrior for each team. We took some time to check out the data and we’d like to share our findings with you.
With the findings shown above, there is a clear indicator that not having a support on a team, versus a team with a support, could cause a loss more often than we’d like. Due to the data and the community’s feedback, we’ll be adding another rule to the current set rules we released in the matchmaking blog on August 8, 2014. We’ll be adding the following in a future patch:
- A zero Support team will always match versus a zero Support team
- This rule will not affect a premade party of five members
A team with zero Warriors, versus a team with one Warrior and other matchups, seems to need less attention as the win rates are similar. We’ll continue to examine the rules and make adjustments, as needed. Our goal is to keep match quality in a solid state and queue times low. As the population grows, we’ll be able to layer on more rules without conflicting with the goal mentioned above.
Six Minute Rule
If our matchmaker can’t find you a close match before the end of a six minute queue, it instead finds the quickest possible match for you and your teammates. This rule was originally set early in the alpha as our data told us waiting longer than six minutes didn’t yield a better match experience for a majority of players.
As our population increases, we’re exploring the possibility of increasing the six minute rule to a higher time limit. We believe this change would make better matches for a number of our players, as currently they are being forced into matches against high level players. Because the six minute queue is forcing matches for high level players against the next possible group, they end up facing players substantially lower in MMR.
While the change would certainly help a majority of our players, our data suggests top-tier players wouldn’t see much improvement without other top-tier players being available for a match. The players at the highest level are the cream of the crop and eventually, regardless of how long the queue benchmark is, will be matched with players or teams of a lower skill level. We’re looking at ways to help this problem and hope to have a solid solution at a later date. Any feedback from the community is important to us, so please share your thoughts.
Solo player vs. Large Parties and Solo Queue
To begin, we’d like to mention that our matchmaker will always attempt to match party size and composition first, before attempting to match similar parties with small variances. The idea is to create matches that are as fair as possible. However, there are rare occurances where the matchmaker may deviate from that plan, as we simply need to get you into a game. Now, let’s talk about solo queue players and their potential to hit a large party.
The solo player experience is an important factor to us when considering matchmaking and we have our eye on making sure individual players aren’t forced to play against large parties. Below, you’ll find a graph that details a per game basis of what type of matches the average group of solo queue players will be matched up against.
As you can see, a team of five individual solo players have an incredibly low chance of hitting a five stack. In fact, 92.49% of players grouped by the matchmaker, from queuing solo, will queue up against another team of solo players, resulting in an even match. The other 7.51% of players will hit other types of parties, but most seem to shy away from hitting larger groups. For example, a five stack of players will hit a stack of five solo players only .02% of the time. Simply put, a team of solo queue players will rarely be matched against a large party. We believe, as the population increases, this will hold true for an even higher percentage of matches.
Now, let’s take a look at what a stack of five players will be matched up against according to our data:
When players queue as a five stack, we attempt to match the party against another five stack or large party. Nearly 98% of five stacks will be matched against another party of five, four and one, or two and three. The other 2% of data is spread out amongst the other options. We’ll continue to monitor the matchups we’re seeing as time progresses, but as more players enter the Nexus, we should be able to continue matching large parties against one another.
Overall, we feel the solo queue experience is in a manageable state. Hero League and Quick Match are flexible enough to allow players to queue up as a solo player, and be matched fairly. Currently we do not plan to add a solo league mode into the Nexus. This decision could change in the future, based on player experience, data, and our own analysis.
As a final note, we wanted to address the issue of a Surrender option. For now, we do not believe the addition of a Surrender option is needed in Heroes of the Storm.
Heroes of the Storm matches are designed to be fast-paced, action packed, and played in under twenty minutes. In our experience, the community's feedback, and internal data has shown that there are ample opportunities to mount a comeback. We believe the game is never officially over until the core is destroyed. Adding a Surrender option could tempt players to bail out at the slightest setback, removing focus from the game and potentially introduce even more toxic behavior. As Heroes of the Storm matures, it could be something we’ll add in the future. But, for now, we love to see awesome games where teams mount a comeback and win. For example:
Matchmaking is a difficult beast to tackle. We feel that the matchmaker is heading in the right direction, but agree that we have more work to do. The information above represents only a fraction of the many things we have to juggle when creating close matchups for our players. We are working to improve the system and have a few more ideas in the pipeline, such as individual hero MMR, weighted MMR based on the role being played, and more. We hope to see more of your feedback as we continue introducing more players to the community.