Deep game stats

See Where Your Games Go Wrong, Pattern by Pattern

Pick the games. Stockfish checks every move you made for fifty patterns across the opening, quiet play, tactics, defense, converting, the endgame and the clock. Each one comes back as a rate out of the chances your games actually gave you, with what it cost you per game.

Free No credits used · Chess.com and Lichess

  • 50 patterns in 7 groups
  • Rates out of real chances, not raw counts
  • Free on every account
The "Converting an advantage" table from a Deep game stats report: eight patterns, each with how often it went wrong out of its chances, its cost in points per game and a trend line. "Extra material given back" reads 5 of 33, 15%, 0.12 points per game, worsening.
  • Every rate shows what it is out of: 8 forced mates missed out of 15, not just "8 mistakes".
  • Cost is in points per game, so a rare pattern that loses games ranks above a common one that costs little.
  • The trend compares your earlier games with your later ones, and says "no clear change" when there is none.

From Your Games to One Report

  1. Import

    Bring in your Chess.com or Lichess games with a public username.

  2. Choose

    Filter by date, speed, color, result or opening, and select the games to read.

  3. Analyze

    Stockfish checks every move you made, then checks the key moments again, deeper.

  4. Read

    Fifty patterns, each with a rate, a cost per game, a trend and the positions behind it.

One Report Across Every Game You Pick

The report reads your analyzed games together. It leads with the patterns costing you the most per game, and puts the ones you handle well beside them.

The "Where your games go wrong" panel: Allowed a tactic, 8 of 878 moves, about 0.14 points per game; Allowed counterplay when winning, 8 of 201 chances, about 0.12; Missed the only good move, 12 of 78 chances, about 0.12. Each reads "no clear change".
  • Ranked by what each pattern costs per game, not by how often it happens.
  • The bracket under each bar is the likely range. With few chances, the true rate could sit well above or below the one shown.
The "What you handle well" panel: Gave up an outpost, 0 of 103 chances; Chose the harder way to win, 0 of 98; Moved too fast in a critical position, 0 of 50.
  • A pattern counts as a strength only after ten or more chances, so one good game cannot put it here.
  • Each line says what the chances were: pawn moves that could leave a weak square, winning positions with several winning moves, critical positions reached with time on the clock.

Every Pattern Is a Rate, Not a Count

"Twelve missed pawn breaks" means little on its own. Twelve out of 54 positions where a break was the best move is something you can compare, track and work on. Every pattern is counted that way, against the chances your games actually gave you.

The "Quiet play and planning" table: ten patterns from Missed pawn break (12 of 54, 22%, 0.05 points per game) to Unfavorable trade (0 of 28), each with its cost and its trend.
  • Patterns with only a handful of chances are marked "few", so a 1 in 3 does not read like a habit.
  • A "heuristic" tag marks the four patterns where the label is a judgment on top of the engine measurement, rather than something the engine checks directly.
Missed pawn break, opened: what was observed, a likely range of 13% to 35%, the phases it happened in (early middlegame 5, middlegame 4, opening 3) and the rate across four stretches of games.
  • Open any pattern to see what was observed, which phase of the game it happened in, and how the rate moved across your games.
  • Under this come its costliest examples, each one a position you can open on the board.

Every Number Leads Back to a Position

Each finding is a move from one of your games: the board, what you played and where it led, what the engine preferred, and what the difference cost.

A finding filed under Slow slide in quiet positions: at 22...bxc3, three small losses between moves 15 and 22, none of them a blunder, had taken the position from 75% to 65%. The move played and the engine’s Bxc3 are shown with their lines; cost 15%, checked deeper.
  • Not every bad game has a blunder in it. A slow slide is several small losses in a row, and it is reported as one pattern.
  • "Checked deeper" means the engine searched this moment again at a higher depth before keeping it.
A finding filed under Flagged in a good position: at 34...Bxh4 in the endgame the position was worth 68%, but the game was lost on time with 4 seconds left. Tagged Low on time and Changed the result.
  • The clock is read too: this game was lost on time from a position worth 68%.
  • "Changed the result" marks a move that took the game from one expected result to another, such as from winning to drawn.

Fifty Patterns in Seven Groups

The report groups the patterns by the part of the game they belong to. Each one is counted against its own kind of chance: winning positions reached, threats faced, endgames played.

Opening and transition (6)
Worse out of the opening, ground lost after a good opening, development left unfinished, castling delayed at a cost, the same piece moved again, attacked before you were ready.
Quiet play and planning (10)
Pawn breaks missed or played too early, a slow slide in quiet positions, aimless shuffling, a bad piece left bad, unnecessary pawn weaknesses, outposts given up, open files left unused, unfavorable trades, an old plan kept after the position changed.
Tactics and calculation (8)
Material left hanging, a tactic allowed, a winning tactic or a quiet killer move missed, a combination started but not finished, an unsound sacrifice, a missed in-between move, the only good move missed.
Defense and recovery (8)
A threat ignored, attacking while forgetting to defend, defense that was too passive, an active defense or a defensive trade missed, a necessary concession refused, a chance to recover missed, one mistake leading to more.
Converting an advantage (8)
A forced mate missed, a winning attack let go, extra material given back, not simplifying when ahead, trading into a worse ending, counterplay allowed when winning, an edge never turned into anything concrete, the harder way to win.
Endgame technique (6)
The king left out of play, a passed pawn mishandled, opposition, triangulation or zugzwang missed, poor piece placement, a win that slipped into a draw by rule, a drawing resource missed.
Clock and time pressure (4)
Errors under time pressure, moving too fast in a critical position, long thinks followed by a time scramble, flagging in a good position.

Where in the Game, and How the Clock Changes It

Two more views of the same games: which phase your losses come from, and whether you make more errors when you are short of time.

The "Where in the game" panel: the average share of a point given away per move and errors per 100 moves for the opening (2.1%, 4.5), early middlegame (4.6%, 15), middlegame (4.9%, 18.3) and endgame (4.6%, 9.5).
  • Phases are set by material and development, not by move number, so a short opening and a long one are compared fairly.
  • Each phase shows the average share of a point given away per move, and errors per 100 moves.
The "Under time pressure" panel: 11.4 errors per 100 moves at normal time, 4.8 when low on time, and 0.42 times the usual rate for positions of the same difficulty, from 29 games with clock times.
  • Your error rate at normal time beside your rate when low on time, from the games that carry clock times.
  • The last figure compares positions of the same difficulty, so hard late-game positions are not blamed on the clock.

Nothing Runs Until You Press Analyze

Nothing is analyzed when you import. Filter your games, select the ones you want, choose a depth and press Analyze. Games you have already analyzed are skipped.

The Analyze panel: depth choices of 14 Fast, 16 Standard and 18 Thorough with key moments re-checked at 18, an "Analyze 1 game" button, and the note that 1 of 3 selected games needs analysis, under a minute.
  • Three depths: Fast, Standard and Thorough. Whichever you pick, the key moments are checked again deeper.
  • This account runs its searches on the ChessLogix server. On a free account the same searches run in your browser, at no cost.
  • It keeps running while you use the rest of ChessLogix. Each game is saved as it finishes, so a stopped run picks up where it left off.

Frequently Asked Questions

Is Deep game stats free?

Yes. It is open to every account and uses no credits. On a free account Stockfish runs in your own browser. The only paid features on ChessLogix are AI Move Explanation and Meta-Coach, and Deep game stats needs neither.

Which games does it look at?

Only the ones you choose, from the games you have imported from Chess.com or Lichess. Nothing is analyzed on import. The report then covers whichever analyzed games your filters match, so you can read your blitz games apart from your rapid games, or this month apart from last month.

How long does it take?

That depends on how many games you pick, how long they are, the depth you choose and, on a free account, your device. The page estimates the time before you start and shows the time left while it runs. Finished games are saved as they complete, so you can stop and come back later.

How is this different from a game review?

A game review tells you what happened in one game. Deep game stats reads many games together and asks which situations keep going wrong, how often compared with how often they came up, and what they cost. A blunder count cannot tell a slow slide in quiet positions from a missed mate; this can.

How reliable are the labels?

Forty-six of the fifty patterns are checked directly with Stockfish. The other four carry a "heuristic" tag, because the label is a judgment on top of the engine measurement. The moves show what went wrong more reliably than why, so the report keeps what was observed apart from any possible explanation.

How many games do I need?

One game already gives you its findings. Rates mean more with more games: patterns with only a few chances are marked "few", and nothing is called a strength until it has had ten or more chances.

Find the Patterns Behind Your Results

Import your games, pick the ones to read together and press Analyze. Each game is saved as it finishes, so the report fills in while you carry on using ChessLogix.

Related