30 Build a Session
Every source structures a session the same way: warm up, work on one thing, then play for fun or for scores. What they argue about is length.
- Warm up (10 to 20 minutes). Korilak's method: a fixed set-list of maps, in order, ending with one you can barely pass. A long low-BPM stream map, then an aim and reading map slightly faster, then your "final boss". His rule is not to start real practice until you've passed all of them. 31
- Focus block (40 to 60 minutes). One skill, from Ch. 27, run as one of the programs below. Cap retries at two or three per map, or Korilak's three strikes, then move on and come back later. 42
- Variety (15 minutes). Something else entirely, so one-tricking doesn't narrow you. Tournament mappools, published for every rank range and built to test every skill, are the easiest ready-made variety: download one as a collection and rotate through it. 4,44
- Finish (10 to 20 minutes). Comfortable full combo attempts or maps you simply enjoy. Don't chase fast maps right after stamina work; tired fingers rehearse sloppy movement. 31,132
- Stop at pain, not at strain. Burning forearms are ordinary; anything sharp, or pain that persists after rest, ends the session. 5,163
- Research. Skill dips after a break and returns within a short block of practice, an effect called warm-up decrement. Field studies found that task-specific activity, practicing the skill itself, removed it best, which supports Korilak's in-game set-list over generic hand exercises. 144,145
- Real-world data. In a 2025 study of 16 Counter-Strike 2 and Valorant players, physical warm-ups produced no statistically significant change in aiming performance, though the no-warm-up condition produced the fewest best results. Warm up with the game. 146
- Nuance. Korilak recommends an hour or more of warm-ups before serious play; Tonehh uses 10 to 13 minutes on a work night and about 30 when derusting. The evidence supports a warm-up, not any particular length. 6,16,31
- Nuance. Two of the three top players interviewed barely warm up at all, and one argues that an elaborate routine buys 1 to 2% while costing playing time; the third says players who skip it spend the first stretch of every session wondering what's wrong. Warm up enough to clear the dip, then stop optimizing it. 40
Two kinds of session
A session that produces no new top play can be a complete success, as long as you know which kind of session it was. Deciding before you start is what stops you judging one by the other's measure.
| Skill session | Score session | |
|---|---|---|
| Objective | Adaptation: change what you can do | Execution: get the most out of what you can already do |
| What you play | Uncomfortable BPM, unfamiliar maps, low AR, your weakest patterns, adjusted difficulty | Maps inside or slightly above your established range |
| Expected scores | Bad, and that's fine | Your best available |
| Success looks like | Fewer finger locks, a longer clean stream, cleaner transitions, one fewer miss on the target pattern | A full combo, an accuracy record, a new top play |
| Failure looks like | Playing it safe and scoring well | Grinding a map far above you and learning nothing |
Both are legitimate and both are needed. The mistake is mixing them without noticing: farming comfortable maps all week and calling it training, or playing nothing but your skill ceiling and wondering why your scores went backwards. The plateau in Ch. 23 is usually one of those two, run for a month.
Session length. Tonehh's floor for real improvement is about two hours, with shorter sessions spent on your weakest skill where gains come fastest; Korilak warns against playing at all on days when nothing works. Spacing research favors regular moderate sessions over marathons, and sleep does part of the work for you. 6,31,102,103
How much should you retry?
Three pieces of advice have circulated in osu! for over a decade, usually stated together: play more, push your limits on hard maps, and play variety without spamming retry. The first two are in every source in this book. The third is the shakiest, and it is worth knowing why. 65
The no-retry rule is newer than it sounds. The two most-read improvement guides of the early 2010s don't mention retrying at all; the rule spread around 2016 from one top player's advice and streams, and other top players repeated it. Two arguments then collided: that retrying costs you the variety that builds consistency, and the opposite, that moving on after a mistake leaves the mistake in your muscle memory while fixing it requires playing the thing again. 65,66,72
- Real-world data. A 2017 analysis of the top 10,000 players plotted a rough improvement rate (pp per hit) against average play length (hits per play, standing in for how much people retry). The trend ran the other way from the rule: players with shorter average plays gained pp slightly faster. A 2020 reanalysis with a better-defined improvement measure found the same direction at r = −0.117 across 10,000 players: statistically significant at that sample size, and practically tiny. 65
- Nuance. Both analyses measure pp, which rewards farm maps, so they arguably show that retrying farm maps farms pp faster, which nobody disputes. Average play length is also confounded by map choice: a player who prefers marathons looks like someone who never retries. The strong version of the anti-retry claim, that heavy retrying narrows you to a small set of maps, isn't measurable in this data at all. 65
- Nuance. The advice is normally given to newer players, while the data covers the top 10,000, whose skill sets are already broad. Critics in the same thread make both points, and the author's own conclusion is modest: whichever side is right, the effect is probably small.
- Research. What does hold up is spacing: repetitions spread out beat repetitions massed together, which is an argument for returning to a map later rather than for never replaying it. 103
So the honest version of the rule is narrower than the slogan. Nothing shows that retrying as such slows you down. What the coaches in this book actually object to is a specific pattern: restarting the first twenty seconds of a farm map, never submitting a score, never finishing a map, and hammering one section until you're playing it from memory. Cap your retries if the cap keeps you moving through varied maps; don't adopt a retry rule that makes you miserable in the hope of a benefit the data can't find. 31,42,65