Drills#

Drills are the core of the app. Everything else — the library, the assessment, the scoring — exists so that a drill session can answer one question well: what are the three things I should repeat right now?

How the app picks cases#

When you press Start session with no focus set, the selection is:

  1. Take every case in the phases you have enabled for the session.
  2. Sort by score, None first (never assessed, never timed), then ascending.
  3. Apply the star boost — a starred case moves up relative to unstarred cases with a similar score.
  4. Take the first drills.weakest_case_count (default 3).
  5. Give each case drills.default_reps repetitions (default 5).

Both numbers are configurable per instance in config.yml and can be overridden per session on the drills page.

If you have focus items set (see the library), step 1 is restricted to the focused cases, groups or phases and everything else is ignored. The drills page tells you when a focus is active, because a forgotten focus silently disables the weakest-first logic.

Three is not a small number

Three cases × five reps is fifteen timed executions and takes about four minutes. The point of the small set is repetition density: you meet the same case again while the last attempt is still in your fingers. If you want volume, run more sessions rather than a bigger session.

Running a session#

Each repetition looks the same:

  1. The scramble is shown. It is a real scramble that leaves the cube in this case — not just the setup moves — with a random AUF and, where it does not destroy recognition, a random cube rotation. Apply it to your cube. See Scrambling.
  2. Optional inspection. If drills.inspection_seconds is non-zero you get a countdown before the timer starts. It ships as 0.
  3. The timer runs. Press and hold the spacebar (or touch and hold the big button on a phone) to arm, release to start, press again to stop. This is the standard stackmat-style interaction every cuber already has in their hands.
  4. You record the result. Keep the time, mark it +2, or mark it DNF. A DNF is stored as an attempt with is_dnf set — it counts as a rep and is excluded from your time statistics.
  5. You can re-rate. After each rep the six-point assessment scale is one tap away. Use it. The moment right after you fumble a case is when your rating is most accurate.

Between reps you can also:

  • open the case in the 3D viewer to re-watch the algorithm;
  • see the inverse of the algorithm to undo the case by hand;
  • star the case;
  • skip a rep, which records nothing;
  • end the session early — everything already recorded is kept.

Endless repetition#

Any case (or any group) can be repeated an unlimited number of times. From a case page choose Drill this case and set the rep count, or leave it open — the session simply keeps handing you the same case with a fresh scramble until you stop it.

This is the mode to use when you are learning a new algorithm: twenty reps of one OLL in a row builds the muscle memory that five reps spread over a week does not.

After the session#

The summary shows, per case:

  • every attempt with its time, DNFs marked;
  • your best and your trimmed average for the session;
  • the change in the case's score;
  • how the session average compares with the split target for your goal.

Sessions are stored, so the drills page also lists your recent sessions and you can reopen any of them.

Session kinds#

Kind Cases come from Use it when
weakest The weakest-first ranking (the default) You want the app to decide
focus Your focus selection You are deliberately learning a group
starred Your todo list You want to clear your stars
case One case, endless reps You are learning a single algorithm
plan A block of a training plan You are following a plan

Recording an attempt outside a session#

Attempts made anywhere — a one-off timed rep on a case page, a drill, or the phase splits from a full solve — all land in the same attempt table with a source field saying where they came from. Score calculations treat them the same, so a case you timed once on its own page still counts toward the three attempts needed for a measured score.

Why times, not counts

The trainer never asks "how many times have you practised this". Repetition counts reward grinding; times reward improvement. The only thing that moves a case's score up is executing it faster relative to the target for your goal.

Source: docs/user/drills.md