How Competitive Skill Rating Works (Season 10)

Competitive Discussion
11

Is that exclusive to each comp mode or do they use the same mmr for elims?

12

Blizzard has said exactly zero about MMR/SR in competitive 6v6 and capture the flag. And I can’t be bothered to take data. But what I have gathered from forum anecdotes is that your first game in the special competitive modes is seeded with your MMR from normal competitive, but with a high uncertainty until you have played a sufficient number of games in the special competitive mode.

There is zero evidence that doing well or poorly in a special competitive mode will influence normal competitive.

13

I love how the whole matchmaker discussion makes it out to be like some intelligent being struggling to make fair games. In reality, matchmaker groups low MMR players with high MMR players to “make the overall gameplay experience smoother for the majority”. Meaning serious players get to carry casual players. There are more casuals, they buy more stuff, so keep them happy.

4 Likes
14

You don’t have a well formed theory here of what you think is happening. But it is probably addressed in the “Popular Myths” section of the original post.

Regardless, unless a player is at the extreme ends of MMR (and assuming he is not grouped), he is equally likely to be the lowest MMR or the highest MMR player in any given match. Some matches he will be the carry, some matches he will be the carried. Most matches he will be in the middle somewhere.

2 Likes
15

I ranked high platinum in CTF comp and went to diamond fast then stayed there and still had no trouble. This was just after i had grinded out of low silver in regular comp. I felt more at home with the diamond players than i do gold players, but Blizz wants me with the silvers. I can’t argue with any of the OP but would just add that the system isn’t working well for solo queue. So then, also penalizing the groups that don’t have much time together is going to kill the game. Improve solo queue or stop giving skill penalties to groups.

18

My two cents on a couple things:

First of all, you say the performance adjustment is minor. My experience (over an absurd number of games recording the SR gain/loss on almost every single one of them, so I have a very good feel for “expected” gain/loss per game) is that the adjustment is very significant, and can very easily make up the difference between 15 points and 30.

Here are some stats to back this up, after entering all of my season 9 games (nearly 600 games) into a spreadsheet. Unfortunately I was tracking them all on post-it-notes by my computer, and after entering season 9 have exactly zero desire to go back to previous seasons, but I’ve started entering each match directly into the spreadsheet rather than on post its (while wondering why I didn’t just do that it the first damn place).

Games 585
Wins 302 52% 22
Losses 279 48% -23
Draws 4

Hog 153
W 75 49% 20
L 77 50% -25
D 1
Soldier 298
W 148 50% 21
L 150 50% -24
D 0
Moira 191
W 96 50% 25
L 93 49% -22
D 2
Winston 33
W 14 42% 22
L 19 58% -23
D 0
Mercy 34
W 11 32% 22
L 21 62% -24
D 2

The last number on the w/l rows is the average gain/loss per game.

The interesting bits here are that Moira is by far my best hero, that I consistently get the best stats with, Soldier is my “main” (the one that I want to play the most often) but that I know I don’t do as well on, and I just picked up Hog during the last part of the season, so obviously I’m not as practiced with him. Even if the difference was only between 20 and 25 points, that is still a 25% bonus/penalty, which is quite significant.

Second of all: I started recording stats because I was convinced that there is something seriously screwy with SR gains and losses, where I was consistently losing far more points for losing than I was gaining for winning. That is, in general, absolutely 100% true, and it is extremely frustrating. The overall difference between gaining 22 per win and losing 23 per loss might not seem that significant, but over 585 games that is over a full ranking level!

Here’s what I do know: that discrepancy is not because of the stats adjustments. I track stats pretty consistently on masteroverwatch, and here’s where some of the key stats lie percentage-wise:

Soldier
Elims: 1.96/min (top 65%)
Damage: 1149/min (top 71%)
K/D: 2.58 (top 47%)
Accuracy: 31% (top 75%)

Roadhog
Elims: 1.62/min (top 74%)
Damage: 1152/min (top 59%)
K/D: 2.59 (top 56%)
Enemies Hooked: 1.03/min (top 92%)

Moira
Elims: 2.27/min (top 41%)
Damage: 650/min (top 33%)
K/D: 3.76 (top 21%)
Healing: 739/min (top 86%)

Mercy
Healing: 1041/min (top 48%)
Ressurects: .7/min (top 14%)

Winston
Elims: 1.89/min (top 69%)
Damage: 711/min (top 62%)
K/D: 1.94 (top 75%)
Blocked: 929/min (top 75%)

With the exception of Moira healing and Hog hooks, my stats are all in the top 50-70% range. But my SR is low to mid silver, which on masteroverwatch is about top 90%, so my stats are generally going to be higher than others at my SR.

Which means that there is some other factor that unbalances gains and losses even more than what is shown here. I don’t know what that factor is, but it really pisses me off sometimes and frustrates the hell out of me. There is absolutely no good reason for Blizzard not to show the factors involved in SR changes.

1 Like
19

PS: I am quite positive that if I were to one-trick Moira, I could quite easily make it to Plat or Diamond. I don’t want to do that, and in fact am trying to add a wider variety of heroes to my pool, which is likely to keep me in silver for the foreseeable future. But at least I know that’s something I’m doing to myself, as opposed to some bull**** hidden factor that causes me to lose more than I gain.

20

Before I address your point, your numbers don’t seem to be adding up.

Win Loss Tie Games
Given Total 302 279 4 585
Hog 75 77 1 153
Soldier 148 150 0 298
Moria 96 93 2 191
Winston 14 19 0 33
Mercy 11 21 2 34
Summed Total 344 360 9 713

Adding the numbers together is significantly greater than the total you gave. Are you counting games multiple times when you switch heroes?

Could you share the raw data spreadsheet? Initial SR, Final SR, and heroes played for each game? Google sheets works well for this. If you need to post a link, surround it in `` to evade the forum filter.

21

They don’t add up because I played multiple heroes in many games.

(and incidentally, the numbers appear to suggest that I did much worse in games that I did play multiple heroes) :slight_smile:

Also, my totals don’t quite add up with my stats as reported by overwatch, it’s off by 9 games. Not sure exactly what happened there (other than the fact that I wrote down almost 600 games on post-it-notes and then manually entered them into a spreadsheet, the potential for some error is pretty high, but it should be fairly close overall)

22

Here’s a link: https://docs.google.com/spreadsheets/d/1G1TxRlIUdzxXiNm1gSSN5suKJmV2PDnVpji5iTALxTQ/edit?usp=sharing

There are a couple anomalies in there (some over-40 point losses that I feel like I might have actually forgotten to record a game), but with the total number of games, a few small discrepancies shouldn’t affect the overall averages much.

23

so the numbers of masteroverwatch and overbuff are for your rank not all ranks. from what i see is you die to much for being in that rank bring that down and it could also mean your out of position for those death…moria is your best hero right now.once you improve youll get sr from 80 to 60 to 35 and youll rank up fast even if you lose some games which they do happen from time to time.

24

Masteroverwatch stats are across everybody that they track (which admittedly is only a small percentage of the total playerbase), not just players at my rank.

I “die too much for my rank”!? 2.5 K/D is pretty good, and my Moria K/D is 3.76. (which is actually too high, and it’s one of the reasons my healing number is low, which is something I’m specifically working on).

Top 50 players are in the 2.5-4 K/D ratio range, I don’t think dying too much is my problem. (well, ok, it is, but not nearly to the extent that you’re claiming it is).

I know I’m a silver-ish player, and I know most of the reasons why. I also know that I lose more points for losing than I gain for winning, and that my stats are higher than the averages for my rank, so the performance-based adjustment doesn’t account for it.

REPLY TO POST BELOW THIS ONE because apparently I’m limited to 3 replies as a “new user”. (which is amusing, because I have been using the battlenet forums on and off since Starcraft, but whatever)

Ah, ok, yeah, I see what you’re saying. The ratings are measured across everybody they record, but my stats were gained against bronze and silver players mostly.

Still, I think a lot the stats are more dependent on my ability than the people I’m playing against (or at least weighted that way, although obviously there will be an influence), so I think the assessment that on average my stats are probably higher than people my rank still seems valid.

And that was the only point. I wasn’t calling out stats to say “hey, I should be ranked higher”, just to say that they are likely to be higher than other people at my rank, so it shouldn’t be the performance-based adjustment that’s dragging me down. Honestly, I just wish they’d get rid of that entirely, across the board. It’s stupid. And not being able to even see if that’s what’s impacting you is even worse.

Maybe I’ll have to rethink a bit though. My thought process that my aim would still be my aim against a gold player, for example, could be skewed. Because that would be true, but their team would probably be better at killing me, which would reduce the amount of damage I do overall.

Another thing I do frequently though check out other people’s stats in the games I’m in that play heroes that I play and compare. And what I’ve seen generally reinforces the idea that my stats are higher than the average player at my rank.

So maybe it is the performance-based adjustment leading to the discrepancy after all, but it doesn’t seem that way, although I obviously don’t have a clear-cut way of determining or demonstrating that.

If only I could be told exactly what was affecting my SR changes…

:slight_smile:

You have given me much food for thought though.

1 Like
25

yes top 500 are 2.5-4 k/d in that rank if they were at low ranks ill be like 10 to 15 k/d…yes yours are not that bad but if you want to get more sr and rank up you need to be up from 3.7 and more so even when you lose its only 15 to 20.see the think is your k/d is vs other at your rank if you go up in rank then your number would be less not higher so you have to show it that you can hang with the higher sr… u get me…sorry im at work and no time to reread what i write lol

2 Likes
26

Thanks. I checked your math, and got
With outliers:
21.88 +/- .23 on a win
-23.18 +/- .24 on a loss
Outliers (change of less than 16 or more than 39) removed:
22.38 +/- .19
-23.10 +/- .17
Which basically agrees with your result, though more precise.

I also charted your SR over time:

This indicates that the main thing that is holding your back is not the performance modifier (which is indeed minor as it only costs you about 1/3rd of an SR per game), or other mysterious factors in the algorithm, but the fact that you can’t seem to win a game when your SR is approaching 1900.

Though there is some evidence that you improved, as you were capped at 1700 earlier in the season. This is how you rank up: improved play combined with enough games to overcome random factors and any SR asymmetries / bugs / etc.

You have no evidence to support that claim. But if you believe it, why not try it? Maining Zenyatta is how I got out of bronze/silver. Just to warn you, though, if you drop a plat player in silver (even a support main) they will generally win 75%+ of their games until they are well into gold.

2 Likes
27

i know what you mean …but if you can record your games people on this forms can help you out or even you will see what you can do better next time and improve also at the same time.i have friend in all ranks and some are better then there rank but stuff happens or they go back to there bad habits again or they rank up and they get to were they have a hard time like they got to were there skill level is.

28

Thanks for all the input. I clearly had not given enough consideration to the fact that the stats I’ve been watching were being measured against the entire database, but acquired against players just as bad as I am.

As far as recording games goes…I actually started recording all of my play sessions, and am seriously considering starting a “UnScrubbed” youtube channel where I post all of my gameplay and review some of my own games on the channel. Sort of a history of somebody escaping the lower ranks and how it happened. :smiley:

Also, the biggest thing holding me back right now is playing drunk. That steep drop from almost 1900 all the way below 1600 that I’ve been slowly recovering from?

Yeah, that was all in one night where I didn’t think I was nearly as drunk as I was. The next day I decided to really stop doing that. Hahahaha

2 Likes
29

dude, you’re my hero

1 Like
31

I was so excited to see that someone was right about something on these forums that I forgot to read the subsequent posts…

Not realizing that your stats are meaningful only against players of your own skill level is a common error.

Good discussion! It’s always good to have more data to look at.

32

Been there brother…check out this post…
https://us.forums.blizzard.com/en/overwatch/t/pc-na-2350-and-below-bronzers-ow-team/84806

I’m going to buy an alt so I can play tired. I still won’t play drunk, but I don’t want to tank my SR just because I had a long day.

33

In hindsight, that should have been way more obvious to me.

Because duh. Lol

Now I kinda wish the stat sites did actually give you a rating compared to people around your own skill level, rather than a global comparison. :smiley:

Still though, way the best would be to get actual feedback about why you gained or lost the amount that you did. It’s always bothered me that Blizzard believes in hidden MMRs. :confused: