Skill-based matchmaking (SBMM), also referred to as matchmaking ranking, is a form of matchmaking dependent on the relative skill level of the players involved.

A common rating system in chess is the Elo rating system, developed by Arpad Elo. Former International Chess Federation president Florencio Campomanes described it as an "inseparable partner to high-level chess". In 2006, Microsoft researchers proposed a skill-based rating system using Bayesian inference and deployed it on the Xbox Live network, then one of the largest deployments of a Bayesian inference algorithm. The researchers were displeased with the ranking system in the beta of Halo 2 (2004). By the time Halo 2 launched, it was using TrueSkill.

The term skill-based matchmaking first appeared in a 2008 interview with game designer John Carmack in which he emphasized its importance in Quake Live (2010). Upon setting up an account with id Software, the game will ask the player for their skill level and judge accordingly depending on their performance from that point forward. The presence—or lack thereof—of skill-based matchmaking became a point of contention. During the development of Dota 2 (2013), Valve Software believed that the barrier to entry could be solved with, among other things, skill-based matchmaking through its Steamworks service; when Call of Duty: Black Ops (2010) developer Treyarch was asked why the game wouldn't include skill-based matchmaking unlike Halo 3 (2007), multiplayer design director David Vonderhaar said that speed was "more important than anything else".

The skill rating of a player is their ability to win a match based on aggregate data. Various models have emerged to achieve this. Mark Glickman implemented skill volatility into the Glicko rating system. In 2008, researchers at Microsoft extended TrueSkill for two-player games by describing a number for a player's ability to force draws. Variability in map, character, and server effects have been considered in at least two research papers. In 2016, two Cornell University graduates modeled skill rating as a vector of numbers, showing "substantial intransitivity".