Supports and Skill Rating

Announcements
1 2 3 26 Next
Over the past few weeks we’ve been reviewing reports from players about how their SR gains after a win appear to be lower when they play as a support. I wanted to share some of the results of the investigations so far and clear up some community misconceptions along the way.

As part of the 1.9 Orisa patch, we made a change to how assist scoring was handled to address what we honestly considered to be a bug. Players were getting full assist credit even if the player being assisted did very little to the target. This change, along with other more significant balance changes in the patch, meant that we needed to recalibrate the tuning for the systems that calculate a player’s contribution to the match. This was performed for all heroes several weeks ago, and we’ve already recalibrated once more after the recent 1.10 patch.

The change to the handling of assists also affected how often a player might earn “On Fire” status during a match, but the amount a player is “On Fire” does not directly affect SR adjustments due to player performance. There is some correlation between the two because both systems are trying to measure “How well are you playing?”, but there are significant differences. The determination of being “On Fire” examines not just your own performance, but your performance relative to your teammates. The calculation of your SR adjustment after a match doesn’t look at your teammates, but instead compares you to the performance of other similarly skilled players with that hero across an enormous pool of competitive matches. So, we compare your Genji play to the play of other Genjis, Ana vs. Anas, etc. Since we’re comparing “apples to apples”, we shouldn’t see any kind of support specific bias in SR adjustments due to player performance.

We’re still seeing anecdotal reports of some players experiencing lower SR gains on wins, but we’ve also been seeing other reports from other support players that their SR gains look correct. Based upon our investigations so far doesn’t look like there’s a broad systemic issue affecting all supports across all competitive matches. There might be a more localized issue affecting a specific hero, or a certain type of play style or game situation. It also might be something completely unexpected, so we’re doing a thorough examination of all the code that affects SR adjustment.

In the meantime, we do appreciate everyone’s continued feedback and reports about this. They do help, we are looking at them, and if we find issues we’ll let you know. Thanks!
Thank you for the update! Only place to go is up from here :-D
Thanks, this is great.
Highly Rated
I appreciate this. I can personally tell you that I main Zen. I was always averaging 18 - 20 SR per win regardless of being on fire or performing well. Same was true if I performed very badly. Still got 18 - 20 SR regardless of skill. I have video proof if you want to take a look at my YouTube videos. Some of the matches I did poor, average and one I got four Gold at the end of the match we won but still only got 20 - 25 SR for the win.

I stopped playing Zen for a while and started playing DVA. Suddenly I am getting at least 30 SR for wins even if I did average. If I did really good then I would get up to 40 SR for a win and not from streaks.

Please investigate Zen.

I would have 1300+ Transcendence healing, 15,000 healing, 50 Defensive Assists and 40 Offensive Assists as well as 15,000+ damage and still only get 20 SR for the win. Then another game I did very little of anything with Zen and got 18 SR. Only a 2 - 5 SR difference from performing really well to poorly.

YouTube Video Proof

https://www.youtube.com/channel/UCuHFc4ET1ZpAQKxETupx7MQ

Prime Example Below (Only got 24 SR):

https://www.youtube.com/watch?v=HbAtRLEqKh8 (42 Gold Elims, 16 Gold Obj Kills, 14,000 Damage, 6,300 Silver Healing, 17 Best Kill Streak, 40 Offensive Assists, 24 Defensive Assists, Bronze Obj Time)

VS (Got 22 SR below)

https://www.youtube.com/watch?v=j0YJucw2D7U (Terrible Stats )

Only a 2 SR difference between performing well and not so well! No streaks!
thank you for finally addressing the issue. :) We appreciate it.
Thank you for the thorough explaination.
04/18/2017 04:25 PMPosted by KingSlayer
I appreciate this. I can personally tell you that I main Zen. I was always averaging 18 - 20 SR per win regardless of being on fire or performing well. Same was true if I performed very badly. Still got 18 - 20 SR regardless of skill. I have video proof if you want to take a look at my YouTube videos. Some of the matches I did poor, average and one I got four Gold at the end of the match we won but still only got 20 - 25 SR for the win.

I stopped playing Zen for a while and started playing DVA. Suddenly I am getting at least 30 SR for wins even if I did average. If I did really good then I would get up to 40 SR for a win and not from streaks.

Please investigate Zen.

I would have 1300+ Transcendence healing, 15,000 healing, 50 Defensive Assists and 40 Offensive Assists as well as 15,000+ damage and still only get 20 SR for the win. Then another game I did very little of anything with Zen and got 18 SR. Only a 2 - 5 SR difference from performing really well to poorly.

YouTube Video Proof

https://www.youtube.com/channel/UCuHFc4ET1ZpAQKxETupx7MQ


https://masteroverwatch.com/profile/pc/us/KingSlayer-12345

As he said, you are getting compared to how well other players who play Zen do at your level. Your stats in master overwatch is compared to ALL brackets and generally at lower rating you would have relatively better stats because opponents are not as good at that level. Yet, your stats are about average so I think your SR gain/loss is justified.
Thanks for the update.

I was thinking about this issue today and had a suggestion. I know Blizzard wants to get Smurfs boosted into the correct ranking quickly and rate players based on their skill, which can lead to unusual amounts in SR gains and losses. That makes sense, but what it neglects is that Overwatch is a team game, not an individual sport. Grading everything on perceived individual contribution can be unfair to some players whose help is felt particularly by their gamesense and ability to get others where they need to be.

I propose that instead of possibly having wildly varying SR amounts, everyone gains the same SR except for the one or two players with the best and worst play. Giving extra SR to the best players gives them extra incentive and helps boost smurfs into the right rank. People want to play well because they'll get more out of it, beyond just average play. The worst player(s) would suffer a slightly larger SR loss. This would help deter people who throw games and troll because it would be much more painful than just playing an average game or even leaving. That is key: The penalty for throwing would have to be higher than the penalty for leaving.

Just some thoughts. I obviously don't know as much about the game as y'all. Hope this gets sorted out!
Highly Rated
but then theres still the issue of 18 minute fiesta 20k healing + 10 rezzes, vs 4min defense of "i sat in the corner and boosted our bastion the whole game"

you cant compare the two as one having done more because its a different map/team comp/situation on the whole.

you are wrong for comparing "other genjis to other genjis" when theres more relevance from competent play like "i got a double kill on both healers and then we advanced the payload after they backed off" which has more of an impact on the win when compared to racking up 5 high action kills only to fail at killing the 1 lucio on the point.

even blizzard has no concern over obj control.
Highly Rated
Thank you for addressing this issue. Really, thank you.

Just a small question though.
If the system compare the stats of the support player with the stats of another support players on the same skill level from a pool of competitive matches to determine the SR, doesn't this mean it compares the stats with the stats from BEFORE the assist change? Doesn't this give a false image of the player because those stats aren't achievable anymore?
Thank you. The big problem I've had is that Blizzard's been very mum about the situation. Everything is working as intended doesn't really assuage beliefs when things slip through the cracks. Hell, season one and part of two Lucio and Mercy weren't getting the gains they should have and those had to be fixed. This constant speculation that it's on fire but no one saying anything just gave the theory credence when you have people playing characters and getting 5 rank from a win.

Communication is what was bugging me, and now we have something more than speculation and frustration. Maybe now I'll actually jump back into comp.
SR gains do not make much sense for a good player plagued by bad ones. Ive had some bad games however my stats(top 10-20%) do not match my SR(top 71%). Ive tracked all my SR gains and losses along with leavers, trolls(people who do not play for the team, negative asses that just sabotage the game). Ive had many matches holding by myself but as soon as im killed we lose a ton of ground with the rest of the team being terrible.

The SR calculation is not accurate to what it is. Its supposed to be personal ranking except its crushed by people around you who underperform or just dont care. SR is more of a team ranking since you need a good team to even consider getting anywhere. Best yet is when you solo queue and get paired with those korean gang bangs. They dont speak english, they are terrible, mostly trolls playing widow + hanzo and refuse to switch.

People who perform well even in a losing team should lose less than the ones who dont perform well. Ie. Ana with 4 golds should only lose 10SR not 30-40 because their team was horrid.

The matchmaking is also broken as well. Most games are very one-sided its rare to have a match that isnt(you can very easily see this on the spreadsheet). Mostly good on one side and mostly bad on the other. Its almost like youre matching is like 4bads + 2good vs. 4good+2bad. The 1000 SR spread on matching doesnt work at all it should be 500 spread for all tiers.

https://docs.google.com/spreadsheets/d/17in6Ad0kYjgEy5Nr9WjtSoizxAE4pvSOk8R9DWmsKgs/edit?usp=sharing
04/18/2017 04:25 PMPosted by KingSlayer
I appreciate this. I can personally tell you that I main Zen. I was always averaging 18 - 20 SR per win regardless of being on fire or performing well. Same was true if I performed very badly. Still got 18 - 20 SR regardless of skill.


I'm also a Zen main. Sometimes I get 18 or even less and it sucks, but if I'm honest with myself it's cause I played badly, or wasn't able to have much of an impact. I have also had amazing games where I can get in the 28-30 range but those are rare. A player who consistently gets 30 is at the wrong rank... I would say I average around the 23-25 range in the last few weeks. Over all of Season 4 I have 506 games as Zen and my average gain is 23.6, average loss 25.98. Of course that includes a lot of streaks from earlier in the season as well as the early bonus SR. Next season I expect this to be better.

Since the streak change patch, 30 feels like the most SR you can gain per match (barring any very long streaks which I haven't had). Losses seem capped at 30 as well. Unfortunately while I have a +18 game more often than I'd like, I have basically never seen a -18. Losses are always well above 20, often 25 or more, and I guess it makes a bit of sense. If your team is losing, you're just not going to be performing well above average enough to merit losing the minimum.

I'm going to start taking better notes about my own performance and see if I can spot what I'm doing wrong in those +18 or -30 games.
Highly Rated
04/18/2017 04:03 PMPosted by Scott Mercer
The calculation of your SR adjustment after a match doesn’t look at your teammates, but instead compares you to the performance of other similarly skilled players with that hero across an enormous pool of competitive matches.


Does the data used for that comparison get purged after a patch that radically changes certain hero mechanics, or general mechanics like the ability to gain fire? If the historical data says that a certain hero averages x% of fire, and then a patch like the most recent one makes it twice as hard for that hero to achieve the same numbers, being compared to that obsolete data could lead to an extended period certain players being deemed subpar for what would have previously been considered average play.
Same here with Zen, I never get good SR even when carrying. Something with performance calculation must be off.
"There might be a more localized issue affecting a specific hero" - are you implying this hero to be Mercy, as a large emphasis has been on her being 'broken' the past month? I'm still concerned that the new fire formula has negatively impacted her (regardless of how SR gains and losses correlate), we now get little to no points for healing/defensive assists and it is challenging to get on fire as Mercy compared to other heroes - requiring to do a 3+ res in a lot of instances, whereas before it was a slow build of fire which felt more appropriate and rewarding.

Can you please confirm if this will remain in place from here on? Was the design intention to shift the focus on utilizing a more aggressive play style with Mercy in order to achieve fire? e.g. more blaster kills/offensive assists. Just so we know how to adapt, really.

Many thanks.
maybe you should take up a character that takes skill, might help you in the long run
Thanks for the effort, and for the update.

However, I expect that performance is too hard to measure accurately to be a reliable measure with which to influence SR (unless someone is very far from their correct rank). For example, consider a player with poor mechanics but excellent leadership skills. He can get almost any group of players to work well together and win, but you will never be able to measure it from his statistics.

The gold standard for error detection here is if someone has a 50%+ win rate, solo queue, over many games, but is losing SR (or vice versa). A similar case would be a person whose SR is stable, but his win rate is not 50%. I don't see how you will be able to get rid of this completely unless you don't take performance into account except in extreme cases.
Finally people will stop complaining about the fictional "support bug".
04/18/2017 05:40 PMPosted by Katsu
"There might be a more localized issue affecting a specific hero" - are you implying this hero to be Mercy, as a large emphasis has been on her being 'broken' the past month? I'm still concerned that the new fire formula has negatively impacted her (regardless of how SR gains and losses correlate), we now get little to no points for healing/defensive assists and it is challenging to get on fire as Mercy compared to other heroes - requiring to do a 3+ res in a lot of instances, whereas before it was a slow build of fire which felt more appropriate and rewarding.

Can you please confirm if this will remain in place from here on? Was the design intention to shift the focus on utilizing a more aggressive play style with Mercy in order to achieve fire? e.g. more blaster kills/offensive assists. Just so we know how to adapt, really.

Many thanks.


I played four games of Comp about an hour ago and I can confirm you still get about 18-23 SR for wins if you play Mercy. My highest SR gain was the game where I managed to pull off a couple of rezzes, one was a 4 man rez which got POTM, and I did kill some people in it when it was safe to do so.

Join the Conversation

Return to Forum