How Competitive Skill Rating Works (Season 10)

High MMR players are matched with other high MMR players. It isn’t “not friendly”. It is just trying to make matches fair. See the summary section in the original post for how MMR and matchmaking work to make fair matches.

I can’t look up your stats for some reason. Maybe you changed your battletag recently. If you link them (surround the link in `` to evade forum filter) I can comment.

Also, from the original post (matchmaking section).

(34) Overwatch Forums

It loads now, but only based on your 6 minutes of reaper. It looks like the problem is that Brigitte is not supported yet, and almost all of your play this season is on Brigitte. Don’t get too exited that it says that you belong in gold. Six minutes of reaper is not enough data to make a determination with any accuracy.

I get the fact that creating a competitive system is in fact extremely complicated. There are multiple variables and human nature to contend with. My question is why have the placement matches if the end result will put you ± 10 SR from where you finished last season? IS there really any value in it? For example, last season I went 7-3 in placements wound up silver to start. Ok that works, (I hadn’t played since season 1) By the end of the season I finished at 1247 SR Bronze. Fast forward to this season. Went 7-3 placed 1257 SR Bronze. I am not understanding how placements matter? It is very demoralizing to do so well during placements to only be so LOW in the SR ratings. I wasn’t expecting to get Gold or anything, but I would expect to be Silver or higher Bronze. Any information and incite would be appreciated.

If you are an active player, each placement games matter about as much as a normal game, on average. See the section on “Season Transitions”.

If you are a placements only player, they matter substantially more.

Hiding your SR during placements is kinda silly, for all but new accounts.

If you wish to climb, it is best to not rely on placements, but play many competitive games in a season.

THIS IS THE UNDERLYING REASON PEOPLE IN LOW RANKS HATE THE COMPETITIVE EXPERIENCE. I should be able to flex and gaurantee myself a better than 50% winrate and climb. Not win 55% just to stay even.

1 Like

Hello, Kaawumba!
I’ve been recording gameplay data in season 10. Could you please apply your stat model to check for anomalies?

(there is one obvious anomaly this season - a sudden jump of 20+ win streak, taking me 100+ SR from my placement)

The data is here: https://docs.google.com/spreadsheets/d/1TBdpG3ahtD31QZ0Xn6HMygxruuIM1285n2cHCNHsNwg/edit?usp=sharing

I added your season 10 data to the season 9 data to get smaller error bars. It’s at Win/Loss Simulation and Data - Google Sheets

I don’t see any 20+ streaks. The longest streak I see is 8, which is long, but not anomalous (difference between the model and the data is still within the error bars).

The worst part about climbing in competitive from my experience is that instead of giving you higher skilled games as you climb, it seems like blizzard would rather determine your skill based on how many potatoes you can juggle.

1 Like

No. This is not compatible with Blizzard’s statement: “All the system does when it comes to matching on skill is attempt to match you with people of a similar number” (Overwatch Forums).

Unless your rank is very high (or low) you are equally likely to be the highest ranked as the lowest ranked person in any given match. Most matches you will be somewhere in the middle. This assumes you are not grouping.

If you find that you keep getting pushed down from a certain rank, you are the potato (that is, the person who does not belong at that rank), not your teammates.

1 Like

This would be great if the ratings actually meant anything. But since they don’t it’s just a random dice roll of who you’re gonna get. Skill variance in mid ranks are insane. Good players keep the potatoes up while potatoes keep the good players down.

1 Like

Yup. Last comp game I played (and likely ever until we get guilds) had a Rein that just yolo charged in all game, with no heed of grouping up or even using his shield. Also had a terrible Dva going in 1v6 all game. Enemy Hanzo had 5 Ults in 5 minutes thanks to these 2 sub bronze potatoes on my team.

And the enemy team was so bad we still nearly won the resultant 4v8. I’ve improved this season but I’m still dropping SR and still losing games vs. Worse opposition as I am getting throwers and potatoes on my team. Bad of shtte matchmaker. I’ve had enough.

Skill variance can be high, yes. But with more games such variation averages out. If a plat player is in gold, he will win more than 50% of his games and climb out, with enough games played.

Nicely done! compliments…

1 Like

The problem I see is, what constitutes enough games? Let’s say someone spends 21 hours a week playing (3 hours per day). Assume you can do 2 comp matches per hour. This would include queue times, bathroom breaks etc. That’s 42 games a week.

Is playing enough hours a week equal to a part time job enough for the system to average out the variations? If so, this seems excessive for anyone who has a full time job and life outside of the game, like me.

I play as much as I can, and I have only 82 games in this season. Overwatch is basically the only video game I play regularly. I know there are students and other people with more free time to play, and I expect them to be the majority of people on comp. Even then, I wouldn’t think they’d be playing more than 20-ish hours a week, every week.

So, is comp set up to average out these variances only for people who can play comp like it was a 2nd job? If so, that doesn’t seem to be a very good system for the majority of players. In stats a large sample size is usually at least 30. So, shouldn’t the averaging out of variances take around 30 comp matches?

That doesn’t seem to be what is happening here. If it was I don’t think we’d see as many complaints about the wide variation in perceived player skill.

1 Like

This is just speculation. Anyone can grab numbers/stats and then bend them to their argument given enough time. The fact is, we won’t know anything until Blizz releases official specific details on exactly how it all works.
Also, there are rank variables that don’t show up in charts correctly or at all.
I.e. Bronze/silver will most likely have a way higher chance of leavers/throwers because at those ranks “caring is pointless”. Also, in those ranks you have a ton of smurfs that throw off the “curve” a TON. Instead of 25-30 elims being “GOOD”, smurfs come in and get 60+ elims and make the mmr bent.
Also, after every ban wave where THOUSANDS of players get banned, the mmr is never reset. SO those stats also effect the curve.
There’s alot of variables that aren’t accounted for.

1 Like

It depends on how accurate you want to be and how close you are to your true rank. A GM in bronze will never lose a game and will climb to silver (or at least diamond) quickly. A silver in bronze may take hundreds of games to get to silver.

This is one of the reasons that we don’t have a reset every season. It may take multiple seasons, but you’ll get where you belong. And if you improve, your rank will eventually reflect that.

The rule of thumb doesn’t apply here, as the situation is different, and much more difficult. I wrote up one analysis here: Overwatch Forums, which implies that you need hundreds of games to make an accurate measurement of someone’s skill. I have another, more complete, simulation with a similar result that I never got around to posting.

I have sympathy for you. I have a similar “problem” of a full-time job and a family. But it is what it is. It’s a fundamentally very difficult problem to rank people quickly and accurately when you have teams of six, solo-queue, 27+ heroes, team-focused game. All proposed ways of improving this: role-queue, more performance SR, easy grouping, heavier punishments for toxic behavior, etc. have negative side effects that you may or may not find acceptable.

2 Likes

You’re welcome to read through all the references (many of which are blue posts) and make up your own mind.

All of these effects affect everyone equally. That is, it won’t hold people away from their true rank. What these effects do is increase RNG per game, which makes each person’s rank bounce around more. Not desireable, but not a fatal flaw with the ranking system.

1 Like

Just gonna leave this here, 'cause it might have some stuff you’d want to add:

1 Like

Thanks. I just saw this. I will definitely be updating my post.

1 Like

I updated the post with the information from Groups and Matchmaking in Overwatch - #215 by Kaawumba-1133. Most of the changes are at the end of the matchmaking section, but I’ve made small changes here and there.

1 Like