TacticBarTL;DR: The mate tag appears on 59.12% of puzzles below 1,200, but only 1.02% above 2,400. Quiet moves rise from 0.76% to 20.06%, and defensive moves from 1.08% to 22.38%.
The mix changes a lot
These percentages show how often each theme appears inside a rating band. Rows do not add to 100% because Lichess assigns several themes to one puzzle.
The mate tag almost disappears
Only 1.02% of puzzles above 2,400 carry the broad mate tag. Because puzzles can have several tags, this does not mean that none of the other lines eventually end in mate. The aggregate tags are not enough to tell us how often that happens.
Quiet moves rise twenty-six-fold
The quiet-move tag appears in 0.76% of puzzles below 1,200 and 20.06% above 2,400—a 26-fold increase. Defensive moves rise from 1.08% to 22.38%, or roughly twenty-one-fold.
Sacrifices rise, then level off
Sacrifices climb from 3.84% below 1,200 to 10.42% at 1,600–1,999, but only reach 11.39% above 2,400. Pins follow a similar pattern: 2.88% at the bottom and 8.42% at the top. They matter, but they are not the feature that changes most dramatically.
What I would take from this
If I were choosing puzzles manually, I would still spend plenty of time on mates and forks at lower ratings. At higher ratings, this data suggests adding more positions where there is no immediate attack. That is a training suggestion, though, not something the dataset itself tests.
Method
I analysed all 6,100,952 rows in the official Lichess puzzle export dated 10 September 2026. I divided ratings into five fixed bands and counted a theme once per puzzle. Theme labels overlap and describe the generated dataset, not every chess position ever played.
The derived JSON contains every theme and band, including those omitted here for clarity.