Loading...
Loading...
We read 92 topper answer copies and counted what each one contained. Here is how the averages shift across rank bands — questions attempted, and examples, quotes and diagrams used per copy.
Read this as a pattern, not a formula. Publicly shared copies skew heavily to the top ranks, so the bands have very different sample sizes (shown in the first column). This is correlation across an uneven sample — the relevance of an example always matters more than the count.
| Rank band | Copies read | Avg questions | Avg examples | Avg quotes | Avg diagrams | Evaluated |
|---|---|---|---|---|---|---|
| AIR 1–10 | 37 | 14 | 12.8 | 3.8 | 3.6 | 1 |
| 11–50 | 17 | 12.9 | 12.1 | 5.4 | 3.9 | 0 |
| 51–100 | 13 | 9.7 | 11.4 | 3.8 | 2.8 | 1 |
| 101–300 | 8 | 10.6 | 14 | 4.9 | 4 | 0 |
| 300+ | 17 | 13 | 13.2 | 4.3 | 3.6 | 4 |
Averages are per copy within each band, computed live from the dataset. “Evaluated” is the number of copies in that band carrying real examiner marks or comments.
It’s not about quantity. Notice that example and diagram density is broadly similar across the bands — top-rank copies don’t simply pack in more. What separates a strong copy is how relevant and well-placed each example is, not how many there are or how long the answer runs.
Weight every row by its copy count. A band built from a handful of copies tells you little; the top bands, with more copies, are the more reliable signal.
Then go to the source. See exactly which examples and quotes recur in the examples bank and the quotes bank, and study the evaluated copies to see what an examiner actually rewarded.
For every topper copy we read, we counted the questions attempted and the examples, quotations and diagrams used, then averaged those per copy within each rank band (AIR 1–10, 11–50, and so on). It shows the density patterns in how higher- vs lower-ranked candidates filled their answers.
No — this is correlation across a small, uneven sample, not proof of cause. Copy availability is biased: top rankers share far more booklets than mid-rankers, so some bands have many more copies than others, and within a band a few prolific sharers can pull the averages toward their own style. Read the numbers as patterns to learn from, not a formula. Quality and relevance of an example matter more than raw count.
Because publicly shared answer copies skew heavily toward the top ranks. We show the copy count per band precisely so you can weigh each row by its sample size — a band with three copies is indicative at best.
Knowing the patterns is one thing; hitting them under exam pressure is another. The Sherlocking Mains programme trains it with daily answer writing, AWE Bot evaluation and mentorship. Or get one Mains answer evaluated free.