Sat. June 10th
This essay will defend a limited role queueing system for improving the overall experience of (exclusively) Ranked Matchmaking in Overwatch. I recognize the importance of minimizing queue times, but hold that the concessions to such a system–properly designed and well implemented–are quite small relative to the potential gains in match quality and the ability of the Skill Rating system to accurately track skill.
I don’t defend any system that enforces the specificities of the metagame. By metagame here, I mean the current team compositions and play styles found to be the strongest in a given patch by the esport side of Overwatch. However, I do think that there are some propositions so fundamental to the design of the game itself that they remain constant across all metagames, patches, and skill levels.
P1: A team composition that features at least one healer hero will be significantly more effective than any composition with zero healing heroes.
P2: A team composition that features at least one tank hero will be significantly more effective than any composition with zero tank heroes.
P1 is, in my judgement, very plausibly true at every skill level. If one was to measure the power of a team at any given moment in game, total heath pool as a percentage of maximum would be the most impactful variable in the formula. Support characters (except Symmetra who has been misclassified since her rework) hold the vast majority of the responsibility for the regeneration of this absolutely crucial resource.
P2 is slightly more controversial. It is slightly harder to see the truth of P2 because tanks primarily contribute to the relative aggregated resource pool of their team by stymieing opposing attempts to diminish it rather than by increasing it directly or reducing that of opponents. Nonetheless, show me a team composition with no tanks and I’ll show you a composition that has a directly superior counterpart.
I hear and respect the concerns voiced by the Overwatch developers themselves and by the community at large. I would hate to see any system implemented that would hinder the ability of players to be creative. That’s why my implementation is simple, unobtrusive, and sharply limited in scope.
The User Interface requirement of the role queue system I imagine would be two check boxes. These boxes would be marked ‘Healer’ and ‘Tank’ respectively. Players could check neither, one, or both of these options depending on their predilection for different roles. The matchmaker would then ensure that any potential match includes at least one player who has checked ‘Healer’ and at least one (distinct) player who has checked ‘Tank’ on each team.
Thats it. No hero restrictions, no indication to teammates of who has checked which box(es), no metagame enforcement beyond the one tank one healer minimums.
Overwatch is radically deep and continually surprising. To this day I continue to see creative players play and win the game in ways previously unimagined. Sometimes even your healers should respawn as Mei or Tracer for a game-changing point contest. The game should never limit the options of players to make these kinds of creative decisions on the fly. Furthermore, any proposed systemic change that affects the matchmaker must weigh the impact on queue times against the theoretical match quality improvement.
I would contend that such a system would have a very small negative impact on queue times. In the majority of my games, propositions one and two are never questioned because both teams virtually always have team compositions that satisfy them. Thanks to Overwatch’s quality game design, Tanks and Healers are fun and rewarding to play compared to other games’ implementations of these roles. This diversity in player preferences means that most matches the matchmaker considers will likely already satisfy the one healer + one tank minimums, thereby having no impact on the queue times of these matches.
In some games, though, these compositions are a result of mutual understanding rather than natural player preference. In others, natural player preference trumps desire to win and extremely poor team compositions are fielded. It is in these games that I would argue the matchmaker has erred.
Consider a hypothetical matchup: Red vs Blue. The variables looked good enough: the players had been waiting for some time and of course the matchmaker hates to delay. Excitedly, it assembled two teams of 6 that, oh joy, had equivalent average MMR! In the eyes of the matchmaker, this is a perfect 50/50 game. The best possible way to measure the relative ability of the players of Red Team versus those of Blue Team. This time, though, something is wrong. Blue Team has tragically found itself with six Mercy-Only roleplayers! While Red Team readies it’s aggressive dive composition (featuring Winston and Lucio as its core enablers) and prepares a strategy to assault the first point, Blue Team is mired in an extended discussion of who has not yet polluted his/her career profile with non-mercy play time.
I believe that the players on Blue Team, roughly 60 seconds after the gates open, would prefer to have waited a bit longer in the queue so that they could each find a team that would permit them to victoriously fulfill their healer fantasies. Perhaps Red Team enjoys such a matchup, but even they fall victim to an artificial inflation of their skill rating. Were Blue Team distributed across a few different matches, each could have used their talent for supporting to defend their teammates against the assault of the Red Team and perhaps emerge victorious. It is not the ability of Blue Team that has led to their defeat, it is simply their misfortune to have specifically been placed together. In this sense then, all 12 players’ Skill Rating has been distorted from its ‘true value’ as a consequences of the randomness inherent to the current system. In this sense then, the match was bad from a system wide perspective. I would contend that the playerbase is intelligent enough to desire the defense of such small guardrails
Some might worry that this argument could theoretically be extended to defend even a full metagame enforcement system. However, the concerns of creativity in exigent circumstances and the incredible diversity of situations in Overwatch would make such a system disastrous. The extremely limited implementation of the one healer one tank minimums that I defend is intended to slightly increase the lower bound on match quality without significantly negatively impacting queue times or player freedom. I would argue that those extra seconds spent in queue when a ‘quality match’ under the current system is rejected for the failure to meet the one tank one healer minimums are worth it from a player perspective. No one wants to play without a tank to peel for them, no one wants to play without a healer to restore their HP and enable their plays.
My suggestion rewards players who are most willing to adapt to the needs of their team with shorter queue times, while ensuring that those who are less willing nonetheless find themselves in ‘winnable’ games. Skill Rating distortion would be reduced, yielding further gains to the quality of the matchmakers judgement and thereby the quality of the player experience. The cost is small, the gains are high. In my judgement, this system or something similar would positively impact the player experience in Ranked Matchmaking in Overwatch.
Leave a comment with your thoughts (or feel free to mock my spelling and/or grammar)!
2 thoughts on “Quality over Quantity”
i really hope this post can reach the dev team. great job jake!
I think the example of six roleplay Mercy does almost nothing to recommend this preferential queue. The reason is while you may want to ensure each team has at least one tank and one heal, there would be no measure to prevent five players who selected healer to be on the same team, if their queue is at the same duration. Therefore, you end up with five Mercy who selected Heal, and one individual who said they are willing to tank. I would in fact argue that the tank player will now have an even worse experience, as they now have a team from whom they expected balance. I understand this is an outlier example, but so is the “six Mercy” example. Each of these examples, as a statistical possibility, will represent the same probablility in the anecdotal environment.