Most people practise sudoku by solving whatever appears at whatever level they already handle. That maintains a skill rather than building one. A twenty-minute routine with some structure moves you up a level in weeks.
The four-part session
- Warm-up (3 min) — one easy grid, scanning only, no candidates. The goal is speed of recognition, not difficulty.
- Technique focus (7 min) — a medium or hard grid where you deliberately hunt for one pattern: locked candidates one week, hidden pairs the next.
- Main solve (8 min) — one grid at your current level, played properly and timed.
- Review (2 min) — if you got stuck, use a hint and read the technique name. That is your study topic for next session.
The review step is the one people skip and the one that does the work. The hint here names the technique rather than just filling a cell, which turns a stuck moment into a diagnosis.
Choosing your working level
Solve at the level where you succeed roughly four times out of five. Below that you are drilling; above it you are guessing. When you complete five in a row without a hint, move up.
Expect the step up to feel bad for about a week. That is normal and it is the only signal that you are actually learning something rather than repeating something.
Drilling a single technique
Take a hard grid and refuse to place anything by scanning after the first two minutes. Fill candidates, then find only locked candidates — pointing and claiming — until the grid opens up. Artificially restricting yourself forces recognition of a pattern you would otherwise skip past.
Do this for one technique at a time, for about a week each. The order that works: locked candidates, then naked pairs, then hidden pairs, then triples.
Use a puzzle you have already solved
Re-solving a grid you have finished, but restricted to one technique, removes the pressure of the unknown and lets you concentrate entirely on the pattern.
Tracking progress honestly
Best times are stored per difficulty on this device, which gives a like-for-like comparison. Two other measures are more useful than raw time: how often you need a hint, and how often you break a grid and have to repair it. Both should fall before your times do.
| Signal | What it means |
|---|---|
| Times falling, hints steady | Getting faster at what you know |
| Hints falling, times steady | Learning new techniques — the better sign |
| Errors falling | Bookkeeping is improving; harder grids are now viable |
| Everything flat for weeks | Time to move up a level or drill a new technique |
How often, and for how long
Daily beats weekly by a wide margin, because pattern recognition is built by frequency. Twenty minutes a day for a fortnight will do more than three hours on a Sunday. If you only have five minutes, do the warm-up and the technique focus and skip the rest.
Avoiding the plateau trap
The usual plateau looks like this: comfortable at medium, frustrated at hard, so you stay at medium. The way through is not more medium puzzles. It is one hard puzzle, with candidates filled, hunting a single named technique, with a hint used the moment you stall. Two weeks of that and hard becomes the comfortable level.