Thanks! I was actually going through your post looking for the right citation.
However, I can’t find in that ref where it says anything about this specifically.
I’m pretty sure I read or saw in a developer update video that part of the reason SR is used is because a player’s MMR is a complex description rather than just a single number. The wording was something along the lines of, “SR is a more digestible number”.
EDIT: What you’ve quoted refers to the matchmaker, not matchmaking rating. I think “match you with people of a similar number” is likely a simplification.
Even if I’ve slightly misinterpreted this, I really doubt that MMR is an integer between 0 and 5000 like SR is. For a start, there’s uncertainty tied to it. At the very simplest it would be X +/- sigma. And it’s likely to be more complicated than that.
EDIT 2:
There’s a couple of other things at work. One is the matchmaker’s confidence in what your MMR should be
yeah, i’m not saying i’m right. i seriously doubt it’s that simple, but i just think it’s pretty clear that players are not always playing at around the SR they are at. I think this is also why so many players argue that there is a huge variance in skill in gold, yet the players who hit plat say the same thing, and i bet diamond players feel the same way. go on a losing streak and suddenly you are not playing with the skill level you expect to see at your MMR.
i don’t think what i said is so controversial that it needs to be proven, but it is quite easy to test that mmr moves fast when you win even without smurf level performance…just win games without overperforming and doing minimal work and you still see the same massive increase in your teammate’s skills over just 5 games of winning straight
anyway i think most people would be less frustrated by the system if they realized this about it.
There are some who put it another way, that the matchmaker gives you better teams if you are doing well, but i think it’s not based off performance per se. i measure my performance every game and it doesn’t have the huge difference in SR gains I get for streaking.
you want evidence go look up falerii (my widow throw smurf and see my gains for today with lucio) 30+ average…but the thing is I gained 30+ in the hard games that i won, regardless of whether i performed well. But you will just argue that the standard for what constitutes good performance in gold is really low so that actually streak bonuses are gone and it’s based on performance.
however, my experience doesn’t match that.
as for you disliking that i refuse to present you with data. when something can be tested in under 3 hrs very few people are going to bother making up a spreadsheet for you just because you don’t want to acknowledge their experiences
A friend I duo with a lot has recently done his first ever placements on a new account. We have not experienced any “forced losses” after ~20 matches post placement. There were clear reasons why our team won the matches we won and lost the matches we lost. None of those reasons had anything to do with smurfs or any other things you mention. When we lost it was mostly because our team was being countered by theirs and we didn’t sufficiently counter-pick. When we won it was most often the same reason but vice-versa.
So, I have conducted your experiment and the findings are contradictory to your claims.
also where the hell did i say anything about forced losses. i said that the MMR goes up and you have to play against better opponents.
anyway i have climbed 280ish today with massive gains and those include massive gains in games that I barely did anything in.
anyway i will do my placements with bastion next time on a fresh account and then do this experiment and get you the data you want, but it’s not hard to test yourself
2 people already misunderstood the purpose of my post and judged it off the title. What i conclude is that posting anything on the boards is meaningless unless I can post what i mean in under 50 characters.
GG
anyway, fun day i will soon have no accounts in gold…i decided it wasn’t very good practice. i will just play quickplay with my low rank friends
This is true, but not necessarily because people are boosted, throwing, smurfing, or what have you. It’s because SR is a singular number that averages out the impact on games you have over time. On a day to day basis, people perform differently. Some days my aim is like a gold scrub. Other days it’s on point and GM level. People also perform differently on different heroes. Just because you play Zarya at a plat level doesn’t mean you play every other hero at a plat level. Some might be diamond, or gold, or silver. People often think that they play the exact same but reality is they are not. Even on the same hero.
What I conclude is that you need to make better titles.
My secondary conclusion is that your writing is so poor that people can’t properly understand you, thus having to infer your meaning from the title of the thread.
You are matched by a single number (MMR) but the changes in that number are complex. One factor in the changes is the uncertainty in MMR.
I expect that MMR is floating point number (computer programming term, look it up if you are unfamiliar), and has no unnatural end points like 0 and 5000. In addition, given any specific MMR, there should be a strict mapping to SR. However, I usually simplify and express MMR in the units of SR. This can be done because of the strict mapping (just like you can express Fahrenheit in terms of Celsius and vice versa).
Whether you consider sigma to be part of MMR or not that important to me. However, sigma is not taken into account when making a match, only when determining how many points to win or lose. This is called K-factor in chess/Elo and works similarly.
From what I’ve gathered from my experience and various forum posts is that the higher you are on the rating scale, the less variable the matches feel.
High variance (both in individual players and match quality) has many causes, many of which I list at How Competitive Skill Rating Works (Season 12) → “Matchmaking” → “If the match-maker says most games are fair, then why are there so many stomps?”.
A large fraction of the controversial posts on this forum are what exactly MMR does and how it behaves. Maybe you are new here.
The thing is, is that you have somewhere between anecdote and nothing, and I have pages of data and analysis which contradicts your conclusions.
I also have two smurfs, but they just placed gold and didn’t have win streaks of significant length, so they won’t show the effect that you claim, even if it exists. That’s why I showed the streamer data, which contradicts your conclusions.
Yes, I understand what a floating point number is, I do quite a lot of programming in my line of work.
You make a lot of assumptions which are not supported by the refereces you cite in your other thread.
It seems reasonable that there could be a direct relationship between SR and MMR, but there has never been any suggestion of that by the developers. The closest would be the assertion from Jeff that “SR closely follows MMR” (this is paraphrasing because I can’t be bothered looking up the exact quote, but that’s basically what he said).
However, sigma is not taken into account when making a match, only when determining how many points to win or lose
Yeah, good point. The matchmaker would be far more complex than required if MMR were as I described. I imagined MMR to be a multi-dimensional description of a player’s performance. I now see that wouldn’t be very useful and agree with you that it is likely to be calculated from an array of complex quantities, but must be expressible in a simple fashion for it to actually be usable by the matchmaker.
All of this discussion with you is very interesting, but I still stand by my comment about the OP being rather pointless.
anyway stevo’s data seems to disprove the idea of streak bonuses, so that would disprove my theory, but without that I have a hard time making sense of why my gains are so high. I can only conclude that gold level lucio players are so bad that a player can be texting with a girl 30% of the game, or afk riding the walls away from the fight because they know their team is going to win anyway and still get high performance gains.
that or MMR massively adjusts for exceptional performances and afterwards you continue to get high SR gains despite not performing well.
New accounts have exaggerated gains for 18-25 games (including placements). Performance SR matters below diamond and gets exaggerated as well in the early games. Compare to Streamer Data - Google Sheets → Dmum 11 and Healgod 12. These players get into diamond+ extremely quickly, though, so performance SR is less of a factor.
forced loss is only in QP. My experience in Comp has been that teams are better coordinated after you win some matches, you go against 4 and 5 stack players in comms if you win too much. The reason I consider it forced in QP is because the matchmaker puts AWFUL people on your team instead of letting you fight more and more skilled enemy players.
That is not true at all. I’ve had games after a winning streak that gave me teammates that looked like bots when I watch the replay.
The reverse is true. After a losing streak I had an enemy team that was so bad I stood on the cart and didn’t move and never died. I just shot them all without WASD once.
![:slight_smile:](https://d74zh8edaijub.cloudfront.net/en/overwatch/images/emoji/twitter/slight_smile.png?v=5" ":slight_smile:")