“MMR” is always hidden for QM. People act like there’s other factors that they use to ‘tell’ what it is, but QM mmr regardless of their impulse is hidden. As such, people can only draw inference on what the game did for that specific match.
That aside, what you probably “mean” is that the game also tries to match on “confidence”. When players win or lose streaks of games, that loosens the game’s confidence in the mmr it is using, and it does try to match similar confidence.
So, when you’re on a winning streak and then think you’re ‘forced’ to lose, the game matched a set of players with similar ‘confidence’ because two sets of similar confidence makes a “fair” game since either side is capable of losing.
If either side can ‘lose’, then it is “fair” chance for both. Or ‘both’ are being “forced” to lose because one of them will lose at the outcome of the match.
However, people usually don’t extend the “forced” aspect of the matching to the other team since they see whatever sets them off, and assume that they are then going to lose, so they do. Or retroactively, they assume the game then ‘forced’ the loss once they lose that game. In any given game, one team is going to lose, so word-smithery aside, someone is always “forced” to lose, so the system tries to be ‘fair’ based on the metrics it prioritizes for any particular region, or time of day.
The game only has numbers to make predictions, it doesn’t actually know how someone is going to play. Matching is based on probability not on an accurate appraisal of “skill” since any given MMR doesn’t know how someone is actually playing for picks, counter-picks, or team cooperation.
Once people realize the game is more than just trying to ‘force’ one person to “lose”, the “forced loss” argument falls apart; it’s not out to get ‘you’, it’s just a sorting system using a metric that’s fairly easy to exploit. MMR is flawed because it cannot appraise “skill”, can’t actually see what players do, and simply tries to work with similar-enough numbers in a limited capacity that is further strained at 3:00 a.m.
The difficult aspect of the game trying to make “fair” matches is that grouped players can also manipulate the matching to put things better into their favor, in a way a solo-queued player generally does not. So the match isn’t as “fair” as the matching system will ‘think’ because it doesn’t actually “think”, it just puts similar numbers together, and the “Law of Large numbers” creates confirmation bias that the system looks like its working when it appraises thousands of games a day, and not just a specific match that gave a specific set of people a bad time.
https://www.heroesprofile.com/Match/Single/63837720
Here’s a game that you faced a 5 stack. The matched tried to make it closer to ‘fair’ by having 4 group players on the other side, two pairs. While that isn’t the same thing, it’s an example of the system trying to match party sizes as the matching tries to prioritize some things over others, especially for QM.
If the game doesn’t allow dynamic party sizes, then waiting times for a “quick match” would expand to the point players would think the game is dead, and they’d stop playing.
On the side of the bigger stack, one player has a high account level (which lends itself to assuming a higher mmr,) while the rest have low account levels. By playing with lower accounts, they’re either completely smurfing, or using smurf accounts to help the ‘newer’ players by lowering the mmr average to reduce the matching range they’ll face at 3:00 a.m.
If I look at the sylvanas’ game history for that same night, the party changes heroes from game to game, so it’s likely that those are players trying to teach/learn a variety of heroes instead of using a specific composition to stomp games all night.
Because ‘smurfing’ masks a players “true skill” the system can only try to compensate by trying to give a higher mmr handicap to the disadvantaged size, which means it’ll try to match against players have have higher mmr than it might otherwise match against.
It’s likely that the game underestimates that handicap, but the handicap is also limited by the available players since players that want to avoid late-night stacks either switch modes (aram or custom,) or play a different game in the wee hours of the night since it is ‘known’ that players are more likely to face unequal matches when the playerbase is low.