Chess1 rates three different activities independently: rated online games, puzzles, and games against the built-in engine. All three use Glicko-2, the same system as Lichess — an improvement over classic Elo that tracks not just your strength but also how certain the system is about it.
The three numbers
Every performance has a rating (your strength estimate), a rating deviation or RD (the uncertainty around that estimate), and a volatility (how much your rating tends to swing). A new player starts at 1500 with RD 500 — which is why early results move the number by a hundred points at a time.
Certainty changes everything
The deviation is the key idea Elo never had. Beat a player whose rating is uncertain and you gain less: the result proves little. Beat someone with a rock-solid rating and the gain is real. The same logic applies to you — after months of inactivity your RD grows, and your first games back swing wildly until the system re-learns where you belong.
Why three separate ratings
Solving a fork puzzle says little about handling a Rauzer attack, so puzzle results feed only the puzzle performance. Online games update the game rating; engine matches update the engine rating, where your opponent is Stockfish pinned to a fixed level — from beginner (1000) to master (2700). The ratings are computed independently and none of them can bleed into another.
What moves your rating after a game
- The opponent's rating and certainty — beating a solid 1800 earns more than beating a provisional 2000.
- Your own RD — established players see small changes, fresh accounts see big ones.
- The result's margin — for engine games the server replays every move and verifies checkmates, stalemates and draws itself, so a claimed win that the board does not support stays unrated.
In rated online games both players' snapshots — before and after — are stored with the game, so a replay always shows exactly what was at stake when the final move was played.