Please bare with me… it’s going to be a simple question and I’m not suggesting to change it I just want to know more. Any personal attacks will be flagged as trolling.
We know we all have have mmr values from -3 to 3, it’s stored at .5 intervals.
So there is a team of 1 vs a team of 1, does having the same mmr mean it’s a 50/50?
If you don’t know don’t replay, and again please keep it civil - it’s a basic question.
Below 3K, basically the matchmaker does a quick calculation on expected performance based on past matches and works out a chance for a team to win. If it’s less than 40%, the match will be rejected and the queue restarted. If you were put in a team that was expected to win, you’ll get slightly less SR than if you get put in a team that was expected to lose, the rest of your SR, below 3K of course, is based on your personal performance in a match.
The public part of the calculation (SR) should be equal on average on both teams. That is all.
MMR is not a value, it cannot be averaged or compared.
MMR calculations are carried out across all ratings and have an impact on all those playing at any given time. In lower ranks, these calculations also affect the SR distribution.
The game leap guy? Yea… hey I said it, I have an account I think he is one of the greats… Just glad I wrote something worthwile at the time for it to be read back to me. Looking at the vid now.
I’ll try to re-word it, but there were some good explanations.
First I am simplifying the equation of saying a team of 1, instead of 6. So if a team 1 person, who has an MMR value, compared to another team of one person who has the same MMR value.
Is that match considered to be 50/50. Is the math that straight forward? Also as someone pointed out, if team A is determined to have a 60+ percent chance at winning. That game is thrown away.
I am just curious if MMR IS the percent winning or … if there are other factors outside of MMR? or is MMR already has those factors rolled into.
With how much time you spend doing this annoying analysis on the forums you could have instead spent that time trying to get better and then you wouldn’t need to complain anymore.
Knowing Blizzard, it’s probably this straightforward. If person A was on a 5 match win streak and person B was a on a 5 match losing streak, and they crossed paths at “1” mmr, I don’t think Blizzard would factor in anything else besides the 1. No one knows for sure, but they don’t even try to make sure that DPS across matchmaking is equivalent. They might get a diamond tank while you get a diamond healer, so I really don’t think Blizzard is overthinking anything in general.
They seem pretty low effort across the board. I mean hell, 30 seasons in and you’ve still got UI covering up player names.
A) this was said quite a long while ago. No reason to assume it’s still in effect. B) I think there are a ton of caveats here considering how often matches are blowouts. Blizzard cares an awful lot about queue times (long dps waits are mathematically unavoidable) and at this point I think they’re more interested in shipping matches quickly than they are with match quality.
why bother trying to improve/climb the “ranks” lmao.
when a quick google search lists several patents that rig the matches towards 50-50 and the devs have stated SR is just a fake per account no-reset number. Ur real skill is hard crunched and unexplainable as mmr and no1 will ever kno/see it.
if it’s about doing w/e it takes to win games, then we don’t ever need mmr.
historical perf data shouldn’t matter, at all, only the sr u are currently at.
that’s what ladders are supposed to be about.
ship 12 randos around the same sr, let them play, those who do w/e it takes win and get sr+pbsr rewards in accordance with the mismatch. nothing need be tracked.
no need for hidden mmr unless to rig the match probabilities.
which is basically illegal in every competition.
in this case the MM would ship for within the interval [40,60] to [60,40].
and the victor would get adjustments in accordance with that mismatch (so more mmr, pbsr, sr). vice-versa for the defeated.
we don’t know if mmr is s.d. around some active population mean, or some s.d. around all historical rank averages, all data on file, or what. it could be the output of some logistic regression or other classifier. or some kind of simple hextile range, or just another kind of weird nonlinear mapping.
if it’s not a “linear” splay out than the “team average mmr” won’t work right.
for 6v6 then it’s the same principle, however we suspect they use arithmetic mean to average mmr. This is what a CS grade might do, but every1 with math knows when comparing rates, ranks, and ratings, you typically want the harmonic mean. So they might average but use the wrong kind of average.