What is MMR, and why is it different from my visible rank?
Matchmaking rating is the hidden number the system uses to decide who you play against; your visible rank is a presentation layer that may reflect it loosely, and the two are deliberately not the same thing.
What MMR does. It estimates your skill and matches you with players of comparable estimated skill, so that games are close. Systems derive from Elo and its successors — Glicko, which adds a confidence measure, and TrueSkill, which handles teams and uncertainty explicitly.
Why the systems track uncertainty. A new player's rating is a guess, so it should move a great deal on each result. An established player's rating is well estimated, so it should move little. Modelling confidence separately is what allows new players to reach their level quickly without destabilising everyone else.
Why the visible rank differs:
Progression feels better than accuracy. A visible rank that only rises, or that requires a series of wins to fall, is more motivating than a number reflecting every result. Systems frequently include protections against demotion and reward streaks.
Placement and seeding mean the visible rank starts conservatively and climbs toward the hidden rating.
Distribution management. Visible tiers are frequently arranged so that a target proportion of players sits in each, which is a presentational decision rather than a measurement.
It discourages obsessing over a number that fluctuates with every match.
What this explains:
Why you face opponents of apparently different rank — matchmaking used MMR, not the badge.
Why climbing feels slower than winning suggests — the visible rank is catching up to a hidden value it may already be near.
Why a losing streak can produce harder opponents, if your hidden rating is above your visible rank.
Why smurfing is so damaging. An experienced player on a new account has a provisional rating and faces genuine beginners until the system corrects — and the correction takes games, during which those matches are ruined. Systems attempt to detect it by identifying abnormally rapid rating gain, and detection is imperfect.
Team games complicate everything, since individual contribution must be inferred from team results.
Queue time is always a trade-off against match quality — the system can be more selective only by making you wait.