Spaced revision of my CAT mocks worked for exactly one round. Of 67 tests I logged with a revision schedule, 65 got a first revision. 34 got a second, and only 8 of those were on time. 12 got a third. Two got a fourth. None got a fifth, sixth or seventh. As I write this on 30 October, 51 of the 67 have their next round overdue. The schedule was fine on paper. What failed was coming back to an old paper when a new one was waiting.
This is not an argument against spaced repetition. The evidence for spacing is strong, and the research post on this site covers it. This is about what happened when one person tried to run it on real mocks for seven months, using his own log.
The schedule I was running
Every mock or sectional starts at R0, the sitting itself. R1 is due 3 days later. After that, each round is due a fixed number of days after the previous round was actually done: R2 after 7 days, R3 after 14, R4 after 21, R5 after 30, R6 after 45, R7 after 60. A late round pushes every later round back. That is deliberate, because computing everything from the test date would make one slip look like a permanent backlog.
What actually happened
| Round | Due after previous | Tests that reached it | Median actual gap | On time | Late |
|---|---|---|---|---|---|
| R1 | 3 days | 65 of 67 | 0 days | 64 | 1 |
| R2 | 7 days | 34 | 13 days | 8 | 26 |
| R3 | 14 days | 12 | 20 days | 4 | 8 |
| R4 | 21 days | 2 | 28.5 days | 1 | 1 |
| R5 - R7 | 30 / 45 / 60 days | 0 | - | - | - |
"Gap" is the number of days between the last session of the previous round and the first session of this one. The 67 tests are 43 full mocks and 24 sectionals.
Round 1 was not really spacing
R1 looks like a success: 64 of 65 on time. The median gap, though, is zero days. In practice my R1 was the same-day analysis, done the evening of the mock. That is valuable, and mock analysis is where most of the learning from a paper happens. But it is not spaced retrieval. The first genuinely spaced round was R2, and that is exactly where the schedule broke: about half of the tests never got one, and three quarters of those that did were late, with a median gap of 13 days against a target of 7.
Why it broke
- New papers always won. I took a mock or sectional most weeks. Each one generated a fresh R0 and R1, which felt more urgent than revisiting a paper from three weeks ago.
- The queue compounded. With 67 tests and seven rounds each, the schedule can ask for many revisions in a single week. Once a few slipped, the overdue list grew faster than I could clear it, and a long overdue list is easy to stop looking at.
- Late rounds did not get easier. R3 ran at a median of 20 days against 14, and R4 at 28.5 against 21. A slip at R2 did not recover; it carried forward, which is what the schedule is designed to do.
What I am changing
- Separate analysis from R1. Same-day analysis stays, but the first spaced round should be a genuine return to the paper, a few days later, from memory.
- Cap the open queue. No new sectional until the overdue count drops below a fixed number. Fewer tests revisited properly beats many tests revisited once.
- Watch one number weekly: how many tests have a round overdue. Today it is 51 of 67. If it rises, I am taking tests faster than I can learn from them.
- Revisit the questions, not the paper. A later round only needs the questions I got wrong or guessed, kept in an error log, which makes a 20-minute round realistic.
Where this is weak
- One candidate, one season. This shows how one person's schedule held up, not how spaced revision works in general.
- Logging, not memory, is what is measured. A round only counts if I logged a session for it. I may have looked at an old paper without logging it, so the real follow-through could be slightly better than this table.
- I did not measure the effect on scores. Whether the rounds that did happen improved later mocks is a separate question this data cannot answer cleanly.
Every number here comes from the revision rounds Karma Yogi attaches to each logged mock and sectional, which is also what shows which ones are overdue right now.
End of essay
- Anish Guruvelli