The honest answer from my own log is that there is no single right session length, because a study session is three different jobs. Across 690 sessions logged between 25 March and 30 October 2026, my median session was 45 minutes and my mean 59. Split by what I was doing, plain study ran a median of 32 minutes, mock analysis 60, and the mocks themselves 120, because a CAT paper is two hours. Asking "how long should I study for" without asking "doing what" gets you an average of three different things.
This updates an earlier post written at 340 sessions. The mean has barely moved (59 then, 59.3 now). What the larger log adds is the shape of the distribution and the split by session type.
The distribution
| Length | Sessions | Share of sessions | Share of hours | Mean mood (1-5) |
|---|---|---|---|---|
| Under 15 min | 10 | 1.4% | 0.2% | 3.20 |
| 15 - 29 min | 102 | 14.8% | 4.9% | 3.75 |
| 30 - 44 min | 194 | 28.1% | 16.1% | 3.63 |
| 45 - 59 min | 81 | 11.7% | 9.5% | 3.80 |
| 60 - 89 min | 155 | 22.5% | 25.9% | 3.90 |
| 90 - 119 min | 53 | 7.7% | 12.7% | 3.94 |
| 120 min or more | 95 | 13.8% | 30.6% | 3.46 |
The table carries the main finding. Sessions under 45 minutes are 44% of my sessions but only 21% of my hours. Sessions of 60 minutes or more are 44% of my sessions and 69% of my hours. Counting sessions makes me look like a short-session studier. Counting time says most of my preparation happened in blocks of an hour or longer.
Three jobs, three lengths
| Session type | Sessions | Median | Mean | Mean mood |
|---|---|---|---|---|
| Study (learning, practice) | 293 | 32 min | 45.1 min | 3.73 |
| Analysis (reviewing a mock) | 332 | 60 min | 65.3 min | 3.86 |
| Test (sitting a mock or sectional) | 65 | 120 min | 92.2 min | 3.00 |
Plain study is the short one: a quarter of my study sessions ran 60 minutes or more, and half ran 32 or less. Analysis runs about an hour. Tests are fixed by the paper. The low mood in the 120-plus band, 3.46, is mostly a test effect: 42 of those 95 sessions were mocks, and tests average a mood of 3.00. Long sessions that were not tests average 3.83, no lower than shorter ones.
What I take from it
- Pick length by the job, not by a rule. A 25-minute timer suits a block of formula drills. It is too short to analyse a CAT mock properly, which in my log takes about an hour.
- Count hours, not sessions. A day of six 25-minute sessions feels busy and adds up to 2.5 hours. If your tracker shows sessions, also look at the hours.
- Long is not automatically worse. In my log, non-test sessions of two hours or more were rated as well as 60-90 minute ones. That is a mood rating, not a measure of learning, so it only says they did not feel worse.
- Breaks still matter. The research summarised in the time-of-day post found a short break did more for test performance than the hour of day. A long session with a break inside it is not the same as one unbroken block.
Where this is weak
- One person's log. These are my habits, shaped by working hours and a CAT paper that is two hours long.
- Mood is self-rated after the session, on a 1-5 scale, and says nothing directly about how much was learned.
- Session length is a choice, not an experiment. I chose longer sessions on days I had more time, so length is mixed up with what kind of day it was.
- Timer sessions and hand-logged sessions are mixed. A hand-logged session may round its length.
Every row here is a query over the session log in Karma Yogi, which stores the length, type and mood of each session, so the same split can be run on anyone's own data.
End of essay
- Anish Guruvelli