Appogee, on 06 October 2016 - 09:54 AM, said:
It seems to me that, even though Elo is gone, the matchmaker still creates two teams of what it believes to be equal skill. by determining the average skill level of both teams.
Eg.
3 good players + 9 bad players = 12 average players.
6 really good players + 6 really bad players = 12 average players
3 really good players + 9 average players = 6 good players + 6 above average players
etc.
Net: the better you get, the more potatoes you are expected to be able to carry.
But if you're levelling any of the Mechs PGI encouraged you to buy, or some of your 'good' team mates are levelling, then you're probably not contributing to your expected level. And then during leaderboard events you get some (not all!) top level players creating low level alts to achieve high scores.
The continually uneven matches show that this approach is flawed.
This is not correct.
It doesn't "add people to get the average where it wants it". That's flatly wrong.
What's amusing here is people use that to try to explain how the MM works, but it's unnecessary: The results we get are the exactly what you should expect from what the algorithm actually is.
It takes the oldest player in the queue, that's the match seed. It then adds players starting from that seed rating, and adding at ever increasing delta from that specific rating.
It goes up and down from the rating together.
However, if the match is seeded at high rating (not uncommon, if you think about it: Mid ranked players will tend to get pulled into matches quickly, so they're not going to sit around long to actually seed matches often - I can demonstrate this more clearly if anyone has questions, but don't really want to go into detail if not necessary) then it'll basically start building teams with the
strongest highest rated players, then add people at constantly lowering ratings until it's got the teams made
Note: it'll only go so far (one tier's worth of rating) then stop extending until it hits the setpoints, under normal circumstances 1 minute then 2 minutes for +/- tier delta, then continue extending. This allows players who queue after the MM starts building the match but within current bounds to get pulled into matches by delaying the process overall, instead of just quicking building a match out of the closest 24 players regardless of ranking delta
But this DOES mean that at "mid tiers" you'll have matches that are made of players almost entirely higher or lower than you, or you flush in the middle, fairly frequently. As your ranking progresses, though, there are ever fewer players above you and ever more players below you. As such, you're going to find more and more lower ranked players with you.
Now, the important thing to consider here is that there are two reasons a player is lower ranked than you, but still close enough to get into a match with you.
1) He's abysmally bad, but plays a lot. Thus, he's not T1/highly ranked yet, because he's so bad his ranking increases very slowly.
2) He's decent, but plays rarely. Mostly rusty, then, and likely out of touch with current strategies.
I fall into #2 above, for example - I play quite infrequently these days (ususally one session a week, for ~6-8 matches), and while I tend to lose rating on 1:10 games overall, I'm still T2 due to playing so few matches.
So, as time progresses, more and more really bad players end up in the #1 situation above. They keep creeping upwards, and populating T2/low T1 with players who just barely manage to increase rating overall.
Enjoy!
TLDR: The MM doesn't "try" to do anything; it doesn't add players to counterbalance other players, it's really simple. But low pops and a ill thought XP bar rating system gets us what it gets us.