Five-year world ranking

World ranking

Score first: it blends a team's own rating with its league's. The two components are shown next to it.

League ratings

A league rating only moves when two regions actually meet, at Worlds, MSI or First Stand.

How it is computed

Every setting below is measured, not chosen by feel. The protocol is stated with the figures.

• Every team starts at 1500. Each game moves both ratings by the standard Elo formula.
• The K factor is 26 domestically and 30 at Worlds, MSI and First Stand: those are the only games that actually compare two regions.
• At each year boundary, ratings regress 10% toward 1500, to account for roster turnover between seasons.
• Blue side wins slightly more often than red, so blue is credited 20 rating points before each game — otherwise that edge would be charged to the team. A 300-point gap means a one-in-ten chance of losing; the chess value of 400 predicts worse here.
• A team that never leaves its region is only calibrated against its neighbours. Its final score therefore blends its own rating with its league's, at 25%. That league rating only moves on inter-region games. Without the blend, international matches are predicted correctly 61.5% of the time; with it, 65.1%.
• Only teams with at least 40 recorded games and recent activity are listed.
• None of these settings were picked by eye. They come from a search run on the 2022–2024 seasons, then checked against 4,334 games from 2025 onwards that the search never saw: the favourite wins 65.0% of them.

Known limits

What this ranking cannot tell you. Worth reading before quoting it.

• A region that rarely plays internationally is poorly calibrated: its teams are rated mostly against each other. Check the "Intl. games" column before comparing across regions.
• A roster change is invisible to the model. Only the yearly regression accounts for it, crudely.
• This is not an official ranking. Riot publishes its own, the Global Power Rankings, which blend a team rating with a league rating the same way — at 20% — for a stated 65% accuracy.