How ChessBlunders.org Identifies Blunders in Your Games

Updated

ChessBlunders.org finds blunders by running your games through Stockfish 16 and flagging moves where the engine evaluation drops sharply. The site then turns those flagged positions into puzzles so you can practice the exact mistakes you made. This works best if you have complete game records (PGN) and want feedback tied to your own play rather than generic tactics.

The identification process

The site's stated workflow is: analyze your games with Stockfish 16 to find blunders, then practice them as puzzles.

In practice, that means:

  1. You provide a game — typically a PGN of a game you played.
  2. Stockfish 16 evaluates the positions — the engine assigns a numerical assessment to each position, reflecting who stands better and by how much.
  3. The site compares evaluations across moves — when the evaluation swings heavily against the player who just moved, that move is a candidate blunder.
  4. Flagged positions become puzzles — you replay the critical moment and try to find the better move.

The key mechanism is the evaluation drop. A move that turns a roughly equal position into a clearly losing one is the classic blunder pattern. A move that only slightly worsens your standing is usually classified as an inaccuracy rather than a blunder.

What counts as a blunder vs. an inaccuracy

ChessBlunders.org does not publish a fixed centipawn threshold in the material available, so treat the distinction as directional rather than exact:

Category Typical signal Practical meaning
Blunder Large evaluation drop The move likely decides the game or throws away a decisive advantage
Inaccuracy Smaller evaluation drop The move is suboptimal but not immediately punishing
Good/accurate move Little or no drop The engine broadly agrees with your choice

Because the cutoff isn't documented, the most reliable way to understand the site's judgment is to look at which of your moves it converts into puzzles. Those are the ones it considers worth retraining.

Why the engine choice matters

Stockfish 16 is a strong, current engine, so its evaluations are a reasonable benchmark for tactical and positional mistakes. Two caveats are worth keeping in mind:

  • Engine evaluation is not the same as human difficulty. A move Stockfish calls a blunder may have been hard to see under time pressure.
  • Evaluation swings depend on depth and settings. Different analysis depths can shift borderline cases between "inaccuracy" and "blunder."

How the identification feeds training

The point of flagging blunders is not just to label them — it's to convert them into practice material. By turning your own losing moments into puzzles, the site targets the specific patterns you actually get wrong, rather than random tactics. If you repeatedly blunder in similar structures (for example, missing a back-rank tactic or a fork), those positions will keep surfacing in your puzzle set.

To get useful output, make sure your input games are complete and correctly recorded; an incomplete PGN can distort the evaluation sequence and mislabel moves.