Square One

    How to Analyze Your Chess Games (So It Actually Changes Them)

    The Square One team · August 13, 2026 · 8 min read

    The three questions that make analysis real

    One: where did the game turn? Not every inaccuracy, THE moment. Almost every game has one or two. Two: what were you thinking when you played it? This is the question engines cannot ask, and it is where the learning lives, because the fix for "I never considered his reply" is different from the fix for "I saw it and misjudged the trade." Three: have I made this exact kind of mistake before? A mistake that repeats is not a mistake; it is a habit, and habits are where rating actually hides.

    The free method, step by step

    • ·Before the engine: replay the game and mark where YOU think it turned. Guessing first is the exercise; the engine is just the answer key.
    • ·Then run analysis (Lichess server analysis is free and unlimited) and compare. Where the engine disagrees with your guess is exactly where your evaluation needs work.
    • ·At each real turning point, write one honest line about what you were thinking. "Assumed the knight was pinned." "Never looked at checks."
    • ·Keep the lines in one running list per month. Count the repeats. The repeat with the highest count is your next training target.
    • ·Skip deep opening analysis below about 1600. The game rarely turned on move 8; it turned on the fork you missed on move 23.

    What the engine line is for (and not for)

    The engine tells you the truth about the position and nothing about you. Its best line is often a sequence no human would play, and memorizing it teaches almost nothing transferable. Use the engine to locate the turning points and to check concrete tactics; use your own answers to the "what was I thinking" question to change future games. The gap between those two jobs is the whole difference between a review tool and a coach.

    When to automate the counting

    The notebook method has one weakness: the counting. Being honest and consistent about your own mistake categories across fifty games is genuinely hard, and it is precisely the part software does better. Square One reads your recent games and does the turning-point finding and pattern-counting automatically, in plain language, free to try on your existing Chess.com or Lichess account. The fifteen-minute personal review is still worth doing on your most interesting games; the point of automating the rest is that the pattern-level diagnosis stays current without costing you every evening.

    Questions people actually ask

    How long should I spend analyzing a chess game?
    Ten to twenty minutes for a personal review: find the turning points, answer what you were thinking, log the mistake type. Past that, returns fall fast. One well-analyzed game beats five scrolled eval graphs.
    Should I analyze wins too?
    Yes, at least occasionally. Wins hide mistakes your opponent failed to punish, and those habits resurface against stronger players. A quick turning-point pass on wins catches them cheap.
    What is the best free tool for analyzing chess games?
    Lichess server analysis for single-game truth: free, unlimited, excellent. For pattern-level review across many games, Square One's free scouting report names your recurring leak from your existing account in about a minute.
    Is analyzing games better than doing puzzles?
    They do different jobs. Analysis finds what to fix; puzzles are one way to fix it. Analysis without training changes nothing; training without analysis aims at nothing. The loop needs both.