Skill rating system - myths and facts

General Discussion
1 2 3 8 Next
Highly Rated
Good evening fellas, today I finished my little experiment with playing 100 games and recording changes to my skill rating. Whole list can be seen below:

2893 -27 lucio
2866 -29 lucio
2837 +18 reaper
2855 -22 sold/roadhog/zarya
2833 -25 reaper/dva
2808 -30 lucio
2778 -40 lucio
2738 -46 fara
2692 -32 Zarya/Zen
2660 +25 zen
2685 -24 zen
2661 -26 junk/ana
2635 +23 sold/lucio
2658 -25 mercy
2633 -23 reaper/zarya
2610 -29 lucio
2581 -31 reaper/roadhog
2550 +27 soldier
2577 +32 soldier
2609 +30 zen/lucio
2639 +31 Zarya
2670 +35 reaper/zarya/harambe
2705 -23 lucio/roadhog/harambe
2682 -20 zarya/harambe
2662 -25 zen
2637 +22 mercy/reaper/soldier
2659 -24 reaper/harambe/roadhog
2633 -25 reaper/sold
2608 -26 sold
2582 +23 mercy/lucio
2605 +20 soldier
2625 +22 reaper
2647 -22 zen
2625 +23 lucio
2648 -22 soldier/zen/lucio
2626 -29 reaper/roadhog
2597 -21 ana
2576 +29 zarya
2605 -23 reaper
2582 -22 harambe/dva
2560 -28 soldier/zarya/rein
2532 +23 reaper
2555 +19 reaper
2574 -22 reaper
2552 +20 reaper
2572 +20 reaper
2592 +21 soldier/phara
2613 +28 soldier
2641 -19 soldier/zarya
2622 +27 zarya
2649 +28 soldier
2677 +30 soldier
2707 +30 soldier
2737 -26 soldier
2711 -27 soldier
2684 -30 mei/sold
2654 -29 zen
2625 +25 zen
2650 +27 zen
2677 -21 zen
2656 +29 zen
2685 -25 junk/zarya/roadhog
2660 -24 lucio
2636 -24 lucio
2612 +22 lucio
2634 -22 reaper/soldier
2612 -30 reaper
2582 -26 pharah
2556 +23 pharah
2579 -29 reaper
2550 -27 reaper/tracer
2523 -29 reaper/lucio
2494 -29 reaper/zarya
2465 +24 lucio
2489 +19 lucio
2508 +27 pharah/reaper
2535 -20 pharah
2515 -19 mccree
2494 -20 mccree/dva
2474 -21 reaper
2453 +23 soldier
2476 +20 rein
2496 +25 roadhog/reaper
2521 +33 zarya/roadhog
2554 +33 lucio
2587 +39 zen
2626 +54 lucio
2680 +52 reaper
2732 -26 reaper
2706 -26 soldier/lucio
2680 -26 soldier/reaper
2654 +20 roadhog/rein/harambe
2674 -25 zarya/rein
2649 +26 lucio/zen
2675 +19 lucio
2694 -23 rein/roadhog
2671 +20 soldier/zen
2691 +19 reaper/lucio
2710 -25 lucio
2685


Now some conclusions:

1) Medals earned have no effect on your SR increase or decrease.
Example:
Win with Reaper, 4 gold medals, 18 pts gained.
Win with Soldier, whom I didn't play well, changed into Lucio later and only a few bronze medals - 23 pts gained.
Win with Zenyatta, only silver medal for healing gained, 25 pts gained.

I suppose that with Season 2 patch and in regards of problems with support's SR gain, Blizzard chose an easy way out and ditched the idea of personal performance being counted in when it comes to SR gain or loss to standarise their system.

2) Chosen hero class has no effect on SR gain or loss. Like at all, in current state of the game it doesn't matter whether you play Lucio or Soldier.
Example:
2625 +22 reaper
2647 -22 zen
2625 +23 lucio

Another example:
2453 +23 soldier
2476 +20 rein
2496 +25 roadhog/reaper
2521 +33 zarya/roadhog
2554 +33 lucio
2587 +39 zen
2626 +54 lucio
2680 +52 reaper


3) The biggest impact on your SR gain and loss comes from win and lose streaks. Hat's off, ladies and gentlemen - the easiest way up is to have a massive winstreak. And the easiest way down is a lose streak. There is no "omg I got 20 pts for a win and -50 for a loss, blizz fix ur !@#$". Ofc you can lose that much but it happens only if you get yourself in a lose streak.

Example:
2855 -22 sold/roadhog/zarya
2833 -25 reaper/dva
2808 -30 lucio
2778 -40 lucio
2738 -46 fara
2692 -32 Zarya/Zen


My avarage rating for these 100 games was 2632,29. As you can see above, as I get closer to my avarage, system punishes me less for my lose streak.
This is a way of system giving us a chance to not fall that much before we stop our lose streak. It's also worth saying that system tends to punish us much more as we get close to our highest rank in the season. On the other hand if we fall way below our avarage rating, our punishment for lose streaks is minimised if not non existent.

Example:
2535 -20 pharah
2515 -19 mccree
2494 -20 mccree/dva
2474 -21 reaper


System ofc works in both ways - if you fall below your avarage rating and will get a good winstreak, you can count on nice income of points.

Example:
2453 +23 soldier
2476 +20 rein
2496 +25 roadhog/reaper
2521 +33 zarya/roadhog
2554 +33 lucio
2587 +39 zen
2626 +54 lucio
2680 +52 reaper


This boost of income gets smaller or stabilises as you surpass your avarage rating. I suppose that when having a massive winstreak, the system may give you a chance and give you even more points even though you are way above your avarage. But on the other hand if you happen to stop your winstreak when being near your peak rating, the system will question your ability to climb.

Example:
2893 -27 lucio
2866 -29 lucio
2837 +18 reaper


This behaviour leads to much lower increase than decreace of your ranking points and the only way to break through it happens to be utilising winstreak boost. If you win/lose one after another when being close to your peak rating you will start dropping in your's avarage rating direction.
Proof: after surpassing 2690 rating, my avarage gain per match was ~22 pts, while avarage loss was ~-28. I could surpass this barrier eventually when I climbed on massive win streaks from lower rank.

It doesn't look that fair and good for health of the game, but the system happens to work in both ways - if you fall way below your avarage lvl, your last loses in the series will be less severe, and the winstreak feature will kick in much faster and with bigger numbers.

Example:
2453 +23 soldier
2476 +20 rein
2496 +25 roadhog/reaper
2521 +33 zarya/roadhog


vs

2552 +20 reaper
2572 +20 reaper
2592 +21 soldier/phara
2613 +28 soldier


and

2855 -22 sold/roadhog/zarya
2833 -25 reaper/dva
2808 -30 lucio
2778 -40 lucio


vs

2535 -20 pharah
2515 -19 mccree
2494 -20 mccree/dva
2474 -21 reaper


TL;DR Problem with differences of SR gained and lost based on chosen class is imaginary, also personal performance is irrelevant on that matter. The most important things are:
1) if you have won
2) if you have won in the series
3) how close you are to your peak rating.


Chapter 2 - influence of being in lower avarage rated premade when compared to your personal rating:

I've also tested how system behaves if you match yourself in full premade while it's avarage rating is lower than yours current rating.
Let's compare both solo and premade situations:

Solo:
2660 +25 zen
2685 -24 zen
2661 -26 junk/ana
2635 +23 sold/lucio
2658 -25 mercy
2633 -23 reaper/zarya
2610 -29 lucio
2581 -31 reaper/roadhog
2550 +27 soldier
2577 +32 soldier
2609 +30 zen/lucio
2639 +31 Zarya
2670 +35 reaper/zarya/harambe
2705 -23 lucio/roadhog/harambe
2682


Full lower ranked (200-300 pts less) premade:
2532 +23 reaper
2555 +19 reaper
2574 -22 reaper
2552 +20 reaper
2572 +20 reaper
2592 +21 soldier/phara
2613 +28 soldier
2641 -19 soldier/zarya
2622 +27 zarya
2649 +28 soldier
2677 +30 soldier
2707 +30 soldier
2737 -26 soldier
2711 -27 soldier
2684


Avarage income in full premade: ~24,6pts
Avarage loss in full premade: -23.5pts

Avarage income solo: 29pts
Avarage loss solo: ~25.8pts

It's not that super representative comparision since it's super hard to get 100% equal situation with same win/lose streaks, but we can see following patterns:
1) Playing in lower ranked premade slows the moment that your's winstreak boost kicks in
2) Playing in lower ranked premade you get and lose less points than if you were playing solo. It's not that big difference though to be not beneficial in the end if you can hold a strong winstreak with your friends.

Additionaly it also slows down the moment the punishment for lose streak kicks in.

Example (no lose streak kicking in):
2732 -26 reaper
2706 -26 soldier/lucio
2680 -26 soldier/reaper


Conclusion: if you have good friends who are sitting in lower ranks but you believe that they play way above their current lvl, it's worth teaming up if you can hold a stable winstreak. It will take more games to gain the same amount of rating than if you were playing solo and managing to get the same score, but it's still low-risk. Bear in mind though that if you won't manage to get this winstreaks, you will probably stall your rating with low loss/income and push the avarage rating towards your current SR what can lead to getting your rank stuck.
Highly Rated
Nice Work man, and well said too
Bumping this up cause, it really does seem very interesting to take in and notice these things.
This explains why solo queue being garbage. They removed personal skill from MMR and placements, etc.

Good job Blizz, way to be lazy when trying to solve the support ranking issue.

How in the world do you gauge solo players if you don't measure their performance at all?
09/13/2016 04:44 PMPosted by Poopfart19
This explains why solo queue being garbage. They removed personal skill from MMR and placements, etc.

Good job Blizz, way to be lazy when trying to solve the support ranking issue.

How in the world do you gauge solo players if you don't measure their performance at all?

By measuring how well they work with a team.
JEFF WE DEMAND A STICKY
09/13/2016 04:47 PMPosted by TornadoTitan
09/13/2016 04:44 PMPosted by Poopfart19
This explains why solo queue being garbage. They removed personal skill from MMR and placements, etc.

Good job Blizz, way to be lazy when trying to solve the support ranking issue.

How in the world do you gauge solo players if you don't measure their performance at all?

By measuring how well they work with a team.


But they are solo, so if the team chooses not to work with them, then what? And this game often puts solos against premades. So what then? If you want to measure how well solos work as a team the system should only queue solos vs solos. No more dynamic queue with this new system. Solo queue only and groups only.
09/13/2016 04:44 PMPosted by Poopfart19
This explains why solo queue being garbage. They removed personal skill from MMR and placements, etc.

Good job Blizz, way to be lazy when trying to solve the support ranking issue.

How in the world do you gauge solo players if you don't measure their performance at all?


This is why I suggested a solo/duo queue and a team queue, with separate rating systems.
09/13/2016 04:58 PMPosted by Widgit
09/13/2016 04:44 PMPosted by Poopfart19
This explains why solo queue being garbage. They removed personal skill from MMR and placements, etc.

Good job Blizz, way to be lazy when trying to solve the support ranking issue.

How in the world do you gauge solo players if you don't measure their performance at all?


This is why I suggested a solo/duo queue and a team queue, with separate rating systems.


Yup, with the new system, it has become a must. There's simply no other way to measure solo queue players with these changes (if true).
09/13/2016 04:58 PMPosted by Widgit
...

This is why I suggested a solo/duo queue and a team queue, with separate rating systems.


if we have solo and duo queue explain to me how groups of 4 or 5 will find games, and I also want to work on my rating with a group or on my own

Yup, with the new system, it has become a must. There's simply no other way to measure solo queue players with these changes (if true).
[quote]
<span class="truncated">...</span>

This is why I suggested a solo/duo queue and a team queue, with separate rating systems.


if we have solo and duo queue explain to me how groups of 4 or 5 will find games, and I also want to work on my rating with a group or on my own

Yup, with the new system, it has become a must. There's simply no other way to measure solo queue players with these changes (if true).


I don't know, and I don't really care. Right now every change is made for groups. They could allow for an option so people can pick both solo and group queue. I'm sure there would be enough to fill spots still. However, right now solos are just used to make teams go against groups. The group balance is almost never equal, so I apologize if I seem unsympathetic towards groups. The only problem would be with 5 mans. 4s would still get a duo que to fill them.

Or they could keep it as is, and simply have a separate rank for solo players. Not a separate queue, but rank them differently.
thanks for the work and dedication :D
So your not posting with your OW profile and you got no link to it? I guess we should just take random stats from some randy because why the hell not.

Edit: Thanks for providing a source.
09/13/2016 05:45 PMPosted by spidey271
So your not posting with your OW profile and you got no link to it? I guess we should just take random stats from some randy because why the hell not.


This guy does have a point. You could have just made up all of that data. I'm not saying you did, but it wouldn't be the first time someone has just pulled stats out of nowhere.

75% of stats are 60% accurate. - Like this.
So your not posting with your OW profile and you got no link to it? No logs or anything. I guess we should just take random stats from some randy because why the hell not.


I'm on EU and it seems that while I can post on US forums as well, it doesn't link me to my EU profile.
If you really want to see it, here you have it:
https://playoverwatch.com/pl-pl/career/pc/eu/Morgoth-2521

Also how could I record my match history along with income for so long if not by doing it manually?

If you want another proof, here you have:
1) Another topic, sadly in my native language, when I was challanged when saying that it doesn't matter what class you pick and I promised to make a record of 100 games to prove myself right in that matter.
http://eu.battle.net/forums/pl/overwatch/topic/17613411893?page=3

2) a text file I created day before to record my games:
https://www.dropbox.com/s/ydhotq5ji7dp1df/match%20history.txt?dl=0
09/13/2016 05:50 PMPosted by Morgoth
So your not posting with your OW profile and you got no link to it? No logs or anything. I guess we should just take random stats from some randy because why the hell not.


I'm on EU and it seems that while I can post on US forums as well, it doesn't link me to my EU profile.
If you really want to see it, here you have it:
https://playoverwatch.com/pl-pl/career/pc/eu/Morgoth-2521

Also how could I record my match history along with income for so long if not by doing it manually?


Thanks. You can see someones match history on master overwatch so that's alright. I just needed some source to back your statistics before I could take it seriously.
This test is also inaccurate because of you doing this on one profile. I'm sure you did your placements on reaper which it looks like. Also how can this be a myth if Blizz said supports were getting less points and ranked lower?

Are you saying blizz is wrong?
09/13/2016 05:57 PMPosted by Kevo149
This test is also inaccurate because of you doing this on one profile. I'm sure you did your placements on reaper which it looks like. Also how can this be a myth if Blizz said supports were getting less points and ranked lower?

Are you saying blizz is wrong?


They said they made changes to that in PTR.
I can report that 2800+ results in plus or minus 20-30 most of the time, however there is a few games for some reason I only lost about 10 or so. But only on losses usually I always gain 20+.

Also I haven't noticed anything regarding playing support resulting in gaining less, however most of the time if I played support I also played dps or tank classes in the same game.
09/13/2016 05:57 PMPosted by Kevo149
This test is also inaccurate because of you doing this on one profile. I'm sure you did your placements on reaper which it looks like. Also how can this be a myth if Blizz said supports were getting less points and ranked lower?

Are you saying blizz is wrong?


They made adjustments after saying that.

Also to the OP, I appreciate more data in measuring the rating system. Mine's here: https://docs.google.com/spreadsheets/d/1TYxqOPL7zBauizxk7js5OH-CCbncW14HNOZkt594jLg/edit?usp=sharing

Most of this has been documented in the past (I have a pretty lengthy post here with references: http://us.battle.net/forums/en/overwatch/topic/20748175363 ), but remember that these are the only things that affect your SR:

1. Whether you won or lost.
2. The gap between your personal SR and the other team's average.
3. Your uncertainty (basically, how confident the matchmaker is that you belong at your current rating).
4. Your individual performance as compared to historical data of other players using the same heroes on the same map (NOT by medals).
5. The length of your current winning or losing streak.

Join the Conversation

Return to Forum