Chess metrics FAQ

They are the official ratings for that site for that time control. We read them from the API when you link your username; Nexa chess does not recompute them. They seed virtual FIDE and the rest of the metrics.

On Chess.com, the dashboard “Classical” column is daily (correspondence), not Lichess-style live classical.

It is an estimate of a FIDE-style Elo from your online rating (Lichess or Chess.com), using conversion tables from public rating-comparison data (e.g. ChessGoals), interpolated in bands by blitz, rapid or classical depending on platform.

Which online rating we use to estimate your virtual FIDE Elo: once you have imported games, we count games per time control (blitz, rapid, classical) and pick the control with the most games in your imported history among those with a valid virtual-FIDE conversion for your platform. If counts tie, we prefer rapid, then classical, then blitz (closer to long over-the-board play). If there are no counts yet (or all zero), we fall back to blitz → rapid → classical. That keeps one anchor aligned with what you actually play in the app, not only what the platform website shows at the top.

On Chess.com, the anchor’s virtual-FIDE conversion tables use blitz and rapid (we do not apply the same conversion to the daily mode behind the “Classical” column).

Important: it is not a registered or certified FIDE Elo. It is an approximate reference to compare you with opponents and with phase Elo on a similar scale.

It is the same calibration idea as your virtual FIDE, applied to the opponent when we know their platform identity and can tie them to an online rating. That lets you read their strength on a scale similar to yours.

If data is missing or there is no match, you will see “—”.

They are metrics built from engine analysis (Stockfish) stored on your imported games. The backend samples evaluations at four points and derives three segments:

• Opening: the position after fullmove 13 (26 half-moves: White and Black have each played 13 times).
• Middlegame: from the end of the opening to the next cut: usually through fullmove 40, or until “endgame entry” if the game is shorter.
• Endgame entry (intermediate point): from fullmove 20 onward, the first position with at most 6 queens, rooks, bishops and knights on the board (Lichess-style rule).
• Endgame: from that entry point to the final position of the game.

From how the evaluation moves across those segments and the opponent’s reference strength, we derive a FIDE-style scale number per phase.

Without Stockfish analysis these values do not exist: import and analyse from Home (“Update”) until games have evaluations in the database.

In the games table, each phase Elo can be compared to the opponent’s estimated virtual FIDE (when we have it). If the gap is at least +12 in your favour for that phase, the number tends to appear green; if it is −12 or worse, red; the band in between is neutral. It is a visual hint, not a verdict of “good/bad” on its own.

Centipawns (cp): quantify the engine evaluation from your side at sample points (fullmove 13, 40, endgame entry, final).

Deltas (Δ): measure how that evaluation shifts between phase cuts. From that shift and the opponent’s virtual FIDE we derive phase Elo (roughly 0.74 Elo per centipawn of advantage in the phase). Positive values mean an advantage from your side.

They group performance by calendar week inside the range you pick (e.g. last 3 or 6 months), split into opening, middlegame and endgame. They help spot trends; spikes or dips depend on how many analysed games you have each week.

On the openings page we group your games by ECO code and variation. Besides the dashboard’s global opening Elo, each line shows wins, draws, losses, percentages and an estimated Elo for that line. Expanding a row shows sample games with opponent, result and a link.

On Endgames we group games by remaining material (white vs black pieces) and theme. Each row summarises results and an estimated Elo for that endgame type, with a link to real games so you can see where you perform best.

Initial import fetches games from the last 6 months from Lichess or Chess.com (for the username you link).

Update fetches only new games since the last stored one (or the initial slice on first run). The process has two steps: import PGN and metadata, then run Stockfish only on games still missing analysis. Then ratings, virtual FIDE and phase Elo are recalculated.

Yes. You can link one username per platform under the same account and filter metrics and lists by site or view them combined, depending on the screen.

In the product’s current form the dashboard and these metrics are free: there are no paid plans to import, run the engine, and view phase Elo, openings and endgames.

No. Virtual FIDE and phase Elo are models and heuristics meant to guide training. They depend on analysis quality, sample of games, conversion tables and phase-cut rules. Treat them as a compass, not a certificate.