Choose a skill-rating algorithm (Elo, Glicko-2, or TrueSkill) for a game's matchmaking system based on team size and match cadence

Steps

1. Use Elo when matches are strictly 1v1 and you want a simple, well-understood rating update with minimal computation and no notion of rating uncertainty.
2. Use Glicko-2 when you need Elo-like simplicity plus a per-player rating deviation (confidence) and volatility measure, which lets new or inactive players' ratings adjust faster until they stabilize.
3. Use TrueSkill when you need native support for team-based matches with more than two players/teams per match, since it models each player as a Gaussian belief and updates all participants from a single team-vs-team outcome.
4. Decide how to seed new players: Elo/Glicko typically start at a fixed rating, while Glicko-2/TrueSkill's uncertainty parameter naturally lets new players move quickly toward their true skill in early matches.
5. Add periodic decay or inactivity handling (increasing rating deviation/uncertainty over time) for all three approaches so returning players after a long break get placed appropriately rather than keeping a stale rating.