The Amateur Chess Observatory
What real games reveal about chess improvement.
Almost all chess literature is built on master games: you learn what the best players do. The Observatory looks somewhere else — at the games of ordinary players, and not to admire the moves but to count the mistakes: which ones keep coming back, at what level they fade, and which ones hold on.
The result of the first edition
We analyzed 52,005 games from real players, rated 400 to 2399, move by move with an engine — and traced the fate of every piece lost after being left hanging.
Of 13,754 pieces lost that way, 6,486 could still be saved at the moment of the mistake: a move existed that put the piece out of danger. The move was there; the player played something else.
That proportion — 47% — is the same in every rating band. As players improve, they leave three times fewer pieces hanging; faced with a still-savable piece, they do not save it any more often. Defense is the most under-trained skill in amateur chess.
Of 13,754 pieces lost after being left hanging in amateur games, 47% could still be saved at the moment of the mistake — and that proportion is the same from 400 to 2399 rating.
Fall 2026 edition · published September 22, 2026 · 52,005 games analyzed
52,005
games analyzed
302,391
mistakes detected and named
38
mistake mechanisms tracked
5
rating bands, from 400 to 2399
The method
The corpus. Tens of thousands of real amateur games from chess.com and Lichess, engine-analyzed. Our detectors do not merely flag a mistake: they name its mechanism, from the fork handed over to the weakened king.
The weighting. Every rate is computed player by player, over 400-point rating bands, from a minimum of 20 games per player and band. A very prolific player never weighs more than anyone else.
The continuity. An edition’s figures are frozen at publication and never change. Each following edition re-measures the same indicators on a larger corpus — comparability over time is what makes the exercise worth doing.
The limits. The corpus is not a random sample of the platforms: these are the games of players who came to have them analyzed. Bullet is set aside — under three minutes, a share of the mistakes recorded owes more to the clock than to understanding, and bullet is not played equally at every level; the 52,005 games kept are blitz, rapid, classical and correspondence. And our detectors stay silent when the proof is missing: the published rates are floors, never ceilings.
Our in-depth reports
One question worked through in full: what the research says, what our own measurements say, and what neither yet allows us to conclude.
Citing the Observatory
The Observatory’s data is freely reusable with attribution.
The Amateur Chess Observatory, Fall 2026 edition, ChessPivot. chesspivot.com/en/observatory
The 38 mechanisms × 5 rating bands, plus the barometer, the survival index and the race exposure. Every row carries the definition of what it measures. Comma-separated, dot decimals, UTF-8.
Journalists, creators, coaches: for a precise figure, an unpublished cut of the data or an interview — contact@chesspivot.com.
And in YOUR games?
These figures describe a level, not a player. ChessPivot runs the same 38 detectors on your own games and tells you which of these mistakes keep coming back for you — named, counted, and tied to the exercise that trains them.