CS2 team rankings

Every completed match in our history is replayed in order, and each map is scored as one game of Glicko-2 — a rating system in the Elo family that tracks not just how strong a team is, but how sure of that we are. Beating a top side moves your rating a long way. Beating a team nobody rates barely moves it at all.

Why not win rate?

Because a win rate has no idea who you played. Go 8–1 against tier-3 opposition and a win-rate table will put you above a team going 15–8 against the best in the world — it counts the wins and stops thinking. That isn’t a ranking, it’s a scoreboard. Strength of schedule is the whole problem, and Glicko-2 solves it by construction.

What the rating deviation does

Glicko-2 carries an explicit uncertainty term — the rating deviation. A team we have watched for fifty maps has a small one; a team fresh off a three-map run has a large one. We rank on the conservative figure — the rating minus twice its deviation — so a side has to actually earn our confidence, not just get hot for a weekend. A short unbeaten run will not vault a new team to the top of this board, and that is deliberate.

That uncertainty also grows back when a team stops playing. The longer a side has been off the server, the less sure we are it is still that strong, so its conservative figure eases down and its ranking fades — a team quietly climbs by winning and slips by going quiet. Two teams on the same raw rating are not ranked equally if one played last week and the other two months ago.

The board is rebuilt from scratch every day, so it is deterministic: the same match history always produces the same ratings. A team with too few rated matches in the window is left off entirely rather than given a flattering position on a handful of games.