A web tool called ChessCheaterDetector is offering players and tournament organizers a way to quantify how engine-like a chess game looks from a statistical perspective. The site at chesscheaterdetector.com takes a PGN file, runs every position through Stockfish, and produces a suspicion profile based on measurable patterns in the moves.
What the Tool Actually Measures
The analysis pipeline has four stages. First, the uploaded PGN is parsed into individual positions, along with player names, ratings, and results. Second, every position is scored by Stockfish running on the server. Third, the system extracts three features from each move: centipawn loss (how far the played move deviates from the engine's top choice), engine agreement (how often the played move matches the engine's recommendation), and position complexity (how many plausible options exist at each point). Fourth, a statistical model combines those features into a suspicion profile with an overall likelihood percentage.
The example output on the site shows a game with 78% engine assistance likelihood, 94% engine agreement across the game, an average centipawn loss of 18, and observed accuracy of 93% against an expected accuracy of 76% for the player's rating. That gap between expected and observed accuracy is one of the signals the model weighs most heavily, because consistent overperformance relative to rating is harder to explain through natural variation than isolated brilliant moves.
Statistical Anomalies, Not Proof
The site is explicit about its limitations. It identifies statistical anomalies, not cheating. A player having a good day, a well-prepared opening line, or a game where the opponent made it easy to find good moves can all produce engine-like statistics without any actual assistance. The tool does not account for game context, preparation quality, or psychological factors that influence human play.
That distinction matters. Chess cheating accusations carry real consequences for players, and a percentage on a website should not be treated as definitive evidence. The site frames its output as a starting point for investigation, not a conclusion.
How It Differs From Existing Analysis
Most chess analysis tools focus on move quality for improvement purposes. They tell you what the best move was and how far you deviated. ChessCheaterDetector adds a statistical layer on top of that analysis, comparing the observed pattern of play against what would be expected from a human at a given rating level. The difference is between "you played well" and "you played in a way that is statistically unusual for your strength."
The position complexity metric is a useful addition. A player finding the best move in a position with one clear continuation is less suspicious than finding it in a position where the engine itself considers multiple options roughly equivalent. The tool weighs those situations differently rather than treating every engine match as equally notable.
What This Means for Chess
Online chess has struggled with cheating detection since the pandemic accelerated the shift to remote play. Platforms like Chess.com and Lichess use their own internal systems, but those systems are not transparent about their methodology. ChessCheaterDetector offers an open alternative where the features and pipeline are visible, even if the statistical model itself is proprietary.
For tournament organizers and federations dealing with cheating allegations, the tool provides an additional data point that is independent of the platforms where the game was played. For individual players, it offers a way to check whether an opponent's performance in a specific game raises statistical flags, though the site correctly notes that flags are not proof.
The tool is free to use and requires no account. Upload a PGN, wait for the analysis, and review the output with the understanding that you are looking at statistical patterns, not a verdict.