Director's Take: Balancing Heroes and Matchmaking
A few weeks ago, we published the first in this series of updates on the game. It’s crazy how much has happened since then! We launched Season 3, announced a collab with One-Punch Man, and I’m now dating Genji! I feel so lucky…
I wanted to dive into two topics today. Season 3 balance and the matchmaker.
We feel that the start of Season 3 is in a much better state of balance than the start of Season 2. We hear from players that balance is the best it’s been since the launch of OW2, while the game feels fresh based on all the changes made. It’s been great watching that discussion, and to foster it, I’d like to give a bit more details and a bit less commentary.
Let’s look at the support role first. On the positive side, most supports are viable in nearly every skill tier. Brigitte has really popped this season and has the highest win rate for nearly all skill tiers except for Top 500, where Zen takes the lead, with both averaging out to a nearly 55% win rate. On the other end of the spectrum are Kiriko and Moira at around 45%. This goes to show, that Heroes with the highest win rates aren’t always the heroes that are played the most. When we look at who is actually being played, the top supports are Ana, Kiriko, and Mercy for nearly all skill tiers until Silver and Bronze, where Moira becomes picked quite a bit. It’s counterintuitive that Kiriko has such a high pick rate at the highest levels of play, but such a low win rate, and Brigitte has such a high win rate but a mediocre play rate. While it would be fun to pick on this group... we tend to think that highly skilled players are continuing to pick certain heroes for good reasons, and we’re getting additional metrics for maps, team comps, and team fights to try and understand this. If it’s fruitful, expect to hear more about it.
Tanks! Tanks are in a great spot right now. It’s almost hard to pick a bad tank. At the top end, Reinhardt has the highest win rate of tanks for most tiers, reaching about 58% in Bronze through Platinum. At Masters and above, Sigma edges him out and sits at 55% in Top 500. Hog is struggling a bit, even though there are still situations where he can be successful. Winston and Zarya can perform well at certain ranks, and the rest of the roster is solid across all skill tiers. Wrecking Ball is probably the biggest tank topic in the community right now, with takes varying. Some think he’s overpowered, others feel that he is in a good spot, and still, others think both… and flip flop, back and forth… His win rate is among the top 3-4 tanks at most tiers and bounces between 51-55%. His pick rate is lower in the metal ranks, increasing as player skill rises. By Top 500, he is far and away the most picked Tank. We think that this has enabled other heroes to stand out in the current meta along with Wrecking Ball, but we are keeping an eye on him and want to make sure that players don’t get fatigued by seeing too much of the little guy.
There have been some big swings in the Damage role. There are two stories here. From a number’s perspective, DPS feels fairly balanced across the board, and Season 3 has been called the “season of hitscan” by many of you – or possibly the season of Cassidy! Changes to Sojourn, as well as buffs to a few other characters, have opened a new set of heroes for people to play. Cassidy, Pharah, Widowmaker, and Tracer have all risen in playtime. In fact, Cassidy is the most picked hero by a huge margin at all ranks except for Top 500 where he’s tied with Tracer, and he continues to move up. Around 30% of matches have a Cassidy on BOTH sides! Besides Symmetra, it’s harder to talk about the most powerful heroes for DPS at middle to higher ranks. The band that most damage heroes fall within is between a 47%-53% win rate. Pharah, Widow, Tracer, and Sym hover around the top, but things shift around almost daily. At low ranks, the top heroes by win rate are Symmetra and Torbjörn. It’s always Symmetra and Torbjörn! In fact, Symmetra typically has the highest win rate in the game for all roles at all ranks except Masters and above. Sombra, at the low end of power, did have a significant lift this season and has a win rate of around 46% at Diamond and above.
With the current state of the game, we’re not looking at doing balance changes until our midseason patch.
We made some big changes to the matchmaker this season. The primary goal was to optimize the pairing of players to minimize the MMR difference within a specific role. For example, when making matches, we put more emphasis on the DPS players on each team being closer in skill than we do for the DPS players on one team being evenly matched with tank players on the other team. This has worked, and matches have a much tighter skill delta between players of the same role.
There have been some other positive effects to these changes. In Ranked, queue times have dropped, and for high-skill matches, they’ve dropped significantly. Average match quality has also risen across the board. Also, matches with longer queue times are typically of a higher quality now than they were previously. Previously the longer a player waited, the looser our restrictions on the skill difference could potentially be. This is still true, but that range should be more condensed.
However, there have been some downsides to these changes, some of them unforeseen. Queue times have risen in Unranked a bit, and for groups with large skill differentials, they’ve risen quite a bit. We’ve made changes here, and queue times for Unranked are already moving lower. We’ve also heard from players that they are experiencing large win/loss streaks in game. To be clear, we don’t try to cause or prevent streaks based on an ideal win rate for players. We’re studying these reports and will share more information when we get it.
The issue that we’re hearing most from our players is encountering people from vastly different skill levels in their Ranked games. Prior to Season 3, the worst 5% of Grand Master level matches would have players 7 divisions apart, the equivalent of Gold 3 to Silver 5. When we initially shipped 2.3 this leaped to 12 divisions apart or greater. We have made numerous changes since then, and are making changes almost daily, and this has now been reduced to 9 divisions. We have additional changes planned for the midseason patch that should improve this further.
Ok. That’s it! Talk to you all soon, and I’ll see you in game!