CS2 predictions

A model trained on our full match history combines each team's strength rating (built map by map from every result), average roster rating, head-to-head record, track record and recent form. The weights are learned from what actually predicted winners, retrained as new matches finish, and the blend is shrunk toward the bookmaker consensus when the sample is thin, so a two-game roster never reads as a lock. Every prediction shows the full reasoning behind the call.

Matches where the model gives one side a 70% or higher raw win probability. Historically those calls land about three times out of four; the hit rate shown on this page is graded automatically from every settled call, wins and losses alike.

The model's win probability plus the full rundown behind it: current form, map pool and veto projection, pistol-round conversion, head-to-head history and how the betting market prices the same match.

Every match on the schedule gets one automatically, from S-tier playoffs to qualifiers: today's matches first, then tomorrow and the rest of the week as the schedule fills in.