Lichess open database · source period 2026-05

What actuallywins, at yourrating.

We take 49.8 million rated games from the Lichess open database and work out what actually wins at each rating band, not what wins for grandmasters. Every figure on this site carries the band it came from and the number of games behind it.

Now reading
Italian Game, White
Win rate
52.0%50.9%50.3%49.8%
Band
600100014001800
Sample
n = 217,726965,410602,367833,888
1.e4 e5 2.Nf3 Nc6 3.Bc4Italian Game, C50
3,810openings replayed and validated
49.8Mgames analysed across four bands, never averaged together
1,255per band statistic rows
25traps ranked by how often they land

The moat

Three numbers nobody else publishes.

Opening databases tell you what happens when strong players play. That is a different game to yours. These three are computed from the pool you are actually in.

01 / Per band win rate

52.0%50.9%50.3%49.8%

Italian Game, White, band 600, n = 217,726band 1000, n = 965,410band 1400, n = 602,367band 1800, n = 833,888

The same opening scores 52.0% at 600 and 49.8% at 1800. A single average across all ratings hides the only part that applies to you.

60052.0%
100050.9%
140050.3%
180049.8%

02 / The Blunder Rate

49.1%47.6%45.5%38.6%

Italian Game, band 600, n = 57band 1000, n = 143band 1400, n = 66band 1800, n = 101

How often a game reaching this opening swings three or more points of material inside the first 24 half-moves after leaving book. A proxy for sharpness in practice, measured against the band's own median, not an engine verdict.

60049.1%
100047.6%
140045.5%
180038.6%

Band 600 median 44.6%Band 1000 median 42.9%Band 1400 median 40.0%Band 1800 median 36.4%

03 / Per band tier list

AABB

Italian Game, White, band 600band 1000band 1400band 1800

Win rate bounded for sample size, adjusted for how safe the opening is relative to its own band, then weighted by how often you will actually reach it over the board.

A600
A1000
B1400
B1800

Tier list preview

The ladder rearranges itself
at every band.

Switch the band in the header and watch the order change. An opening can sit at the top of one band and the bottom of another. That movement is the whole point.

Band 600, the 16 most played graded openings, sample sizes from 127,956 to 2,849,881 games. Source period 2026-05. Full tier list at 600

Band 1000, the 16 most played graded openings, sample sizes from 433,246 to 9,343,706 games. Source period 2026-05. Full tier list at 1000

Band 1400, the 16 most played graded openings, sample sizes from 339,514 to 6,813,943 games. Source period 2026-05. Full tier list at 1400

Band 1800, the 16 most played graded openings, sample sizes from 736,226 to 12,493,284 games. Source period 2026-05. Full tier list at 1800

Band by band

Where the bands disagree.

The three openings whose grades move furthest between 600 and 1800. Read top to bottom and you are watching the advice invert.

Center Game Accepted

C21

600A
50.0% for Black, n = 42,251
1000C
47.8% for Black, n = 190,895
1400C
45.5% for Black, n = 132,349
1800D
43.4% for Black, n = 207,511

Graded A at 600 and D at 1800. Whatever this opening relies on works less often as the opposition improves.

Italian Game: Two Knights Defense, Knight Attack

C57

600S
56.6% for White, n = 37,919
1000S
54.4% for White, n = 127,363
1400B
51.8% for White, n = 59,351
1800C
50.2% for White, n = 72,587

Graded S at 600 and C at 1800. Whatever this opening relies on works less often as the opposition improves.

Center Game: Normal Variation

C22

600A
54.2% for Black, n = 20,972
1000B
51.5% for Black, n = 87,038
1400B
49.7% for Black, n = 41,768
1800D
45.4% for Black, n = 42,286

Graded A at 600 and D at 1800. Whatever this opening relies on works less often as the opposition improves.

Method

How the numbers are made.

Rated rapid and blitz games are streamed from the monthly Lichess dump. A game is counted in a band when both players sit inside it and their ratings are within 300 points of each other. Openings are matched by longest move prefix, then every line is replayed for legality before anything is stored.

The Blunder Rate is a material swing proxy, not an engine verdict, and it is computed from a smaller sample than the win rates. It is labelled as such everywhere it appears.

Read the full method
  1. 1
    Filter

    Rated rapid and blitz, both players in band, rating gap 300 or less.

  2. 2
    Match

    Longest move prefix against 3,810 validated opening lines.

  3. 3
    Replay

    First 24 plies replayed move by move. Anything illegal is dropped, loudly.

  4. 4
    Compute

    Win rate, blunder rate and trap hits per band, then the tier formula.

Pick the opening that wins
in the pool you are in.

Build my repertoire