Insights64

Search every position from 90 million games.

Build precise queries with piece counts, placements, sliding patterns, empty squares, ELO bands, and result. See what was actually played from any position and step through the real games.

Knight outpost on d5sliding horizontally
Total matching positions across all files36,629
0
games indexed
0
positions
0
curated patterns
< 1s
typical query
Query toolkit

Six tools, freely composable.

Five filter types narrow the result set; sort orders pick what surfaces first. Stack as many as you need to find exactly the position shape, material balance, or game context you're studying.

Position

Place specific pieces on specific squares.

Drop a knight on f5, pin a king behind unmoved kingside pawns. Searches every position that contains exactly that arrangement — across millions of games, every variation.

Counts

Constrain how many of each piece are on the board.

White rook = 1, black queen = 0, both bishops alive. Find pure rook endgames, queens-traded middlegames, or any specific material balance — no piece-square work needed.

Sliding patterns

Match a structural shape across every file.

A knight outpost on e5 is the same idea as one on f5. Slide horizontally, vertically, or both — empty-square constraints slide too. The engine evaluates 49 board offsets in seconds.

Empty squares

Require certain squares to be empty.

Open files, isolated pawns, back-rank escape squares — empty marks express the negative space patterns chess players actually think about.

Metadata

Filter by who, when, and how the game ended.

ELO range per side, time control, result, move-number window. Compose with structural filters to find exactly the positions you mean — 2400+ classical games where White won, after move 30.

Sort

Rank results the way you want to study them.

Game count for the most-played ideas, average ELO for the strongest examples, recency for the modern picture, piece count ascending for endgames. Multi-key, ascending or descending.

Filters that compose

Patterns plus material plus ELO plus result.

Each filter narrows the matching set. Stack them to find positions that are exactly what you're studying.

    White knight on e50
    + Supported by pawn on d40
    + Both ELOs over 22000
    + Result: White wins0
Final result set
0
real master games featuring exactly this scenario
52% white24% draw24% black
Move explorer

From any position, see what the world played.

Land on a position from a search and the explorer takes over. Every continuation is ranked by frequency, with the win/draw/loss split for each move from real games at every level.

After 1.e4 e5 2.Nf3Black to move

What players actually do

6,141,514 games
  1. 1. ...Nc65,412,08776%
    47% W26% D27% B
  2. 2. ...Nf6583,2048%
    41% W32% D27% B
  3. 3. ...d689,4171.3%
    51% W24% D25% B
  4. 4. ...Bc541,2900.6%
    39% W30% D31% B
  5. 5. ...f512,1040.17%
    38% W18% D44% B
  6. 6. ...f63,4120.05%
    26% W15% D59% B
Game viewer

Step through the actual games.

Not just an aggregate — every match is a real game with names, ELOs, and full move history. Step move by move. Jump to any ply. Open the original on Lichess.

Magnus C. (2882) vs Hikaru N. (2812)1-0
Move 2. Nf3ply 3 / 8
Move history
  1. 1.e4e5
  2. 2.Nf3Nc6
  3. 3.Bb5a6
  4. 4.Bxc6dxc6
Step through every game that hits a position you searched for — click a move to jump there, or open the full game on Lichess.
What it's used for

Three concrete questions, three concrete queries.

The filters compose to answer the questions chess coaches and players actually ask — not just "has this exact position been reached."

Coach

Drilling rook endgame technique

Pull pure rook endgames at master level, sorted from sparsest to most piece-rich. Start the lesson with 3-piece positions; build toward complex middlegame-into-endgame transitions.

Counts: rooks ≥ 1 each side, no minors, no queensELO > 2200Time control: classicalSort: piece count ascending
178,402 games found
Player

Preparing for an opponent who plays the Caro-Kann

From the Caro-Kann main line position, look at what their rating band typically plays — and the win rate for each response. Study the lines you're most likely to face.

Position: Caro-Kann main line FENBlack ELO: 1800 – 2000Then open the explorer for next-move stats
12,407 games · 6 distinct continuations
Student

Learning the Stonewall as Black

Find positions matching the Stonewall pawn skeleton (d5, e6, f5) where Black won, sorted by average ELO descending. Step through the strongest examples to learn the typical plans.

Position: black pawns on d5, e6, f5Result: Black winsSort: avg ELO descending
23,841 games · top one is 2782 vs 2731
Common questions

Answers to what coaches and players ask first.

  • How is Insights64 different from a chess opening explorer?

    An opening explorer searches for the exact position you reached, walking move by move from the starting position. Insights64 lets you search for any pattern that recurs across millions of games — a knight outpost on the 5th rank, a Stonewall pawn structure, a rook on the seventh rank — regardless of which move number it appears on or what opening got you there.

  • Can I search chess games by pawn structure or pattern?

    Yes — that's the core idea. Place specific pieces on specific squares, optionally require certain squares to be empty, and find every position across 90 million games that contains exactly that arrangement. Classical structures like Carlsbad, Stonewall, Hedgehog, and the kingside fianchetto are saved as one-click presets, but you can build any structure you can describe on a board.

  • Can I search by an exact FEN string?

    Direct FEN lookup is the next feature shipping — track it on the roadmap. In the meantime, you can recreate the position on the search board and use empty-square requirements to demand "nothing else on the rest of the board," which makes the match exact.

  • What does "sliding pattern" search mean?

    A knight outpost on e5 is the same chess idea as one on f5 — same shape, different file. Sliding-pattern mode tells the search to evaluate the structure across every file (and rank, if you choose), so you build the pattern once and find every offset of it that occurred in real games.

  • Where does the game data come from, and how recent is it?

    Every Lichess game from 2013 onward — about 90 million in total. The corpus is re-indexed on a regular cadence, so recently played games show up in search results.

  • Can I filter by player ELO, time control, or result?

    Yes. Set ELO bands per side, pick a result (white wins / black wins / draw), choose a time-control bucket (bullet, blitz, rapid, classical), and constrain the move-number window. All of these compose with the structural pattern filters, so you can ask things like "2400+ classical games where White won, with a knight on f5 after move 20."

  • Is Insights64 free?

    Yes — fully free, no signup or login required. Sort orders, filter composition, the move explorer, and the game viewer are all open to anyone with the URL.

Search, explore, and step through 90 million games.

Free. No login. Built for chess coaches and players who want to study real examples, not dictionary positions.

Or jump straight into a named structure: Stonewall · Hedgehog · Carlsbad · see all 13

Hear when new features ship

One short email per release. Unsubscribe in one click.