Overwatch Competitive Play Guide

Overwatch Competitive Play Guide by ExcaliburZ

Greetings and welcome to this comprehensive guide to Competitive Play in Overwatch. This post will attempt to explain the design and functionality behind the Competitive Play system. Data in this post comes from empirical research as well as posts from Blizzard employees.

INTRODUCTION

In order to queue for Competitive Play, you must be at least level 25. If you are part of a group, your group must be within 50 SR of each other. You receive 1 Competitive Point per win, and this currency carries over across seasons and never expires. It can be used to purchase exclusive rewards such as golden versions of each hero’s main weapon. In addition, a lump sum of Competitive Points are awarded at the end of a season based on your peak SR, NOT your current SR.

This is the table for Peak SR/Competitive Point rewards for the Summer 2016 Season, also known as Season 1:[1]
SR 0-39: 10 CPs
SR 40-45: 20 CPs
SR 46-49: 40 CPs
SR 50-52: 60 CPs
SR 53-55: 80 CPs
SR 56-59: 120 CPs
SR 60-64: 200 CPs
SR 65+: 300 CPs

Completing all of your placement matches for a season rewards you with a seasonal icon, spray, and a lump sum of Competitive Points, all delivered at the end of the season. If you ever reach the Top 500 AND are at least SR60, you will receive a special animated version of the icon and spray. It is not required to hold your position within the Top 500 for any particular length of time: just getting in is enough to earn the reward.[2]

Leaver Penalty

Players who leave Competitive Matches are temporarily suspended from queuing for Competitive Play for 10 minutes.[3] This penalty increases each time a player leaves during a season, up to a full season ban. Banned players forfeit their end-of-season Competitive Play rewards. When a player leaves a Competitive Match, teammates can leave the game without sustaining a leaver penalty, however they will still receive a loss (including losing SR).[4]

MATCHMAKING RATING

Competitive Play in Overwatch is a dual-layered system. You have your Skill Rating (SR) which is visible, and that’s straightforward. Below that, though, is a hidden matchmaking rating (MMR) which is invisible. MMR determines your opponents, and is a more complex representation of skill level due to incorporating additional factors. In the case of groups, the group’s average MMR is used, since the entire group queues together. Your Competitive Play MMR is separate from your Quick Play MMR.

The core concept behind MMR is that wins and losses carry the same weight.

First, let’s learn about what exactly SR represents. There are 100 SRs: SR0 is the lowest and SR99 is the highest. 1 SR represents 1% of the predicted MMR spread of the playerbase at the time of the season roll. If the best recorded player has 7000 MMR and the lowest-rated player has 2000 MMR, then 1 SR would span (7000 – 2000) / 100 = 50 MMR for that season. There may be buffers added in both directions to allow for expansion beyond the predicted model. This allows the numerical definition of 1 SR to remain flexible per region and between seasons.

Visually, the SR distribution somewhat resembles a normal distribution curve.

When queuing up for Competitive Play, the matchmaker will find other players who are:

• a similar MMR (or in the case of groups, a group whose average MMR is similar to yours)
• a similar group size, if applicable (this also applies to other groups in the game, not just your own)
• close in ping to you
• queued up at the same time as you (obviously!)

Player level is not part of the matchmaking criteria. You could be level 50 and matched with level 500s, as long as your MMRs are similar.

Placement Matches

When you start your first placement match with no prior Competitive Play history, you are assumed to be average. It is believed that you start at the mean seed value of SR50. Your placement matches carry moderately more weight than a standard match.[5] This allows you to reach your potential, positive or negative, fairly quickly.

Rating Calculation

When updating ratings, each player’s SR is compared against the other team’s average MMR. Rating gain is applied individually. There are three contributors:

1. The gap between your SR and the opposing team’s average MMR
2. Individual performance
3. Streaks

First, the SR-MMR gap. If your personal SR is lower than the other team’s MMR, you lose less SR for a loss and gain more rating for a win. If your SR is higher than the other team’s MMR, you lose more SR for a loss and gain less rating for a win. MMR is invisible, but SR is derived from MMR, so looking at the average SR of the enemy team can usually give you a rough estimate. If someone in the game is SR42 and everyone else is SR30 (assume no groups), you can infer that the SR42 likely has an MMR close to SR30.

MMR travels faster than SR, and SR chases MMR. You can think of SR as the “reality” and MMR as the “potential”. If we use the hypothetical example of 1 SR = 50 MMR, then on average we can expect your MMR to change +/-50 per game even if your SR changes only by 19 (0.38 SR in this model).

Individual Performance Modifier

Individual performance affects your MMR. Your performance — most likely measured by your average score in the match — is stacked against other players who have used the same hero on the same map on the same side (attack/defense). Presumably your performance is expressed as a percentile and applied to your MMR as a modifier with unknown weight. If your score per minute as Soldier:76 on Hanamura Attack was 230, and this was in the top 11% of Soldier/Hanamura/Attack players, and the weight of the modifier was [25%], your MMR gain for a win earning 50 MMR would be ([0.25] * 0.89 * 50) + 50 = 61.125.

Score is generated a number of different ways and is denoted by a flame icon (). Getting full and partial kills (expressed as a percentage of the target’s max HP in damage that contributed to the kill, so two characters dealing 200 damage each to a 400-hp Sound Barriered Lucio would each get 100, not 50), blocking damage (including Mei’s Ice Wall), capturing the point, moving the payload, healing damage, resurrecting allies, shielding allies as Symmetra, and giving allies armor packs all generate score. Just like the restriction on charging ultimates, score cannot be earned in the pregame setup phase. Score decays at a rate of about 4 per second. Although the score meter below your health bar only goes up to 300, score appears to have no cap (if you get to 400, for example, the meter won’t decay for 25 seconds). The “on fire” (no functional purpose, just a cosmetic animating flame on your portrait) marker appears to be around 210 score.

Medals (both number and quality) do not factor into the individual performance modifier because they compare you to the rest of your team rather than the historical data of players using the same hero in the same situation.

If you used multiple heroes during a match, they are weighted based on percentage of time played. If you played Mercy for 30% of a match and Zenyatta for 70%, Mercy would carry 0.3 weight and Zenyatta would carry 0.7 weight.

Streaks

If you are on an extended streak, winning or losing, the matchmaker handles this by further increasing or decreasing your MMR and SR in order to reflect the increasing uncertainty surrounding your skill level.[6] Win streaks carry the same weight as loss streaks. Humans tend to be streaky by nature, so it’s not uncommon to have a 5-win streak followed by a 5-loss streak, but this would result in a zero net change.

On average, streak bonuses kick in beyond the third win or loss in a streak.[7]

Because MMR travels faster than SR, a streak necessarily means that the gap between your SR and MMR (and therefore the MMR of your opponents) widens. As a result, breaking the streak often results in only a minor change to SR, allowing the gap to close a bit.

1 SR appears to be the maximum that can be gained or lost in a single match, attained through a combination of the SR-MMR gap, individual performance modifier, and streak modifier.

Underdog Label

In the pregame loading screen, the SRs of all individuals are shown, along with the averages of both teams. These averages are rounded to the nearest integer. If one integer is lower than the other, then that lower-average team will be labeled the “Underdog ”.

The Underdog label does NOT impact gains and losses of any kind on its own (remember that all gains and losses are individualized). The only time the Underdog label actually represents a reduced probability to win the game is when all 12 players have SRs that accurately reflect their MMRs, which is an unlikely scenario.

MATCHING PLAYERS

Ping Priority

When matching players together, the top consideration goes to finding players of similar ping.[8] This creates a level playing field in terms of potential. Within comparable ping buckets, players of similar MMR are located.

Group Handicap vs. Solo Players

When a group of players queues together, the matchmaker will always seek to find another group of equal size to match them against.[8] If your group is an unusual size (3, 4, or 5), there may not be an ideal group in queue at the same time to play against. In those instances, the matchmaker will broaden the search after a certain time threshold is reached. The broader search can result in differently-sized group matches or matching groups against solo players of higher skill.[9]

Inactivity

Uncertainty increases with inactivity.[8] For players who have not played in an extended amount of time, the default search range could be wider, or MMR changes could be “hedged” (since the confidence that your skill level is being properly defined is reduced).

FAQ

I grouped with a friend who has a higher SR, but he gained more rating than I did when we won our match. Why?
Generally, this happens when the higher-rated player has an existing win streak.

If the SR-MMR gap, individual performance, and streaks drive rating change, can’t that be exploited by boosting low-rated players with high-rated teammates?
Yes, this is a potential concern on the extreme end of the skill spectrum, where pro players are good enough to control the tempo of the match by themselves. And it does happen. Some changes are coming in Season 2 to ameliorate this problem.[10]

Why does it take 1 loss to offset 3 wins?
This is a result of your SR being “inflated”, meaning your underlying MMR is below your SR. This question also tends to pop up when players experience an extended losing streak followed by a win (losing 0.9 SR followed by gaining 0.3 SR), so there’s often a connotation of disillusionment as well.

Why am I matched against SR80 premades as a SR60?
There are many possible reasons. It’s possible that the SR80 premade has someone much lower than them to drop their average MMR. It’s possible that the SR80 premade has been in queue for a long time and their search range has expanded to include you. It’s possible that you queued at a time where the player pool was smaller, so options were more limited. It’s possible that your MMR is high enough to be matched against them normally. One somewhat unfortunate side effect of the very top skill levels is that everybody knows each other, and they tend to queue together rather than meet as opponents (part of this is because at the very top, the skill gap is wide enough that 1 loss can erase 10 wins).

I played a game where my team was all solo queuers and the other team was two groups of three. Why didn’t it just put one group on each team?
This is a little complicated to explain because MMR is invisible, but the answer lies in the group handicap. Let’s say 2 groups of 3 players queue up and their average MMR is 4000 each. If they were to be matched against each other, the matchmaker would need to find 6 other 4000-MMR players to fill slots. That’s one possible outcome. If it can’t, a second possibility emerges: apply the handicap, and find 6 solo players to match them against whose MMR averages out to 4100. The match will start when either of these conditions are met, but statistically they’ll produce comparable games.

Is it true that healers gain less rating than other heroes?
In the early days of Season 1, this was believed to be the case because teams that did very well didn’t require much healing, which is a healer’s primary source of score. The individual performance modifier was also believed to have a stronger weight coefficient for placement matches. Those two things combined could hypothetically translate into slightly weaker initial placement for healers who are part of extremely dominant premade teams. Beyond initial placement though, win streaks and beating higher-MMR teams is where you’ll get most of your rating (and since the SR-MMR gap is wider for you than the rest of your team, this is what will allow you to catch up to them).

Thanks for reading. If you have any questions, please ask!