GPSC answer key analysis: where the coaching keys disagree
We compared 9 coaching institutes’ GPSC answer keys against the official final key, question by question. Of 200 questions, 145 drew the same answer from every institute and 55 did not. The official answer changed between the provisional and final keys on 6 questions, and the commission cancelled 11.
How close each institute came to the official final key
| Institute | Agreement | Matched / compared |
|---|---|---|
| Nirmit Academy | 96.8% | 180 / 186 |
| Praajasv Foundation(updated) | 94.1% | 177 / 188 |
| Dr. Amit Academy GPSC | 93.1% | 176 / 189 |
| Phoenix Mentorship | 93.1% | 176 / 189 |
| Web Sankul | 92.9% | 170 / 183 |
| Shubh Academy | 92% | 172 / 187 |
| Team Dark Horse | 92% | 172 / 187 |
| DICS | 91.5% | 173 / 189 |
| Sunya IAS | 89.9% | 170 / 189 |
Questions where the official answer changed
The commission’s final key differs from its own provisional key on 6 questions. These are the questions objections actually moved.
- Q 36
- Q 42
- Q 48
- Q 52
- Q 55
- Q 59
Where the coaching majority disagreed with the official key
On 8 questions the most common coaching answer was not the official final answer.
| Q | Official final | Coaching majority | Coaching answers |
|---|---|---|---|
| 3 | C | D | D ×5, C ×2 |
| 48 | B | A | A ×4, D ×3, B ×1, B/D ×1 |
| 59 | D | C | C ×5, D ×3 |
| 61 | B/D | B | B ×7, B/D ×2 |
| 67 | C | B | B ×5, C ×4 |
| 76 | A | C | C ×6, A ×1, B ×1 |
| 96 | C | A | A ×4, C ×3, A/C ×2 |
| 97 | B | C | C ×5, A ×4 |
The most contested questions
Questions where the coaching institutes were furthest from agreement, most divided first.
| Q | Official final | Coaching answers |
|---|---|---|
| 20 | not covered | D ×4, C ×2, A ×1, B ×1, C/D ×1 |
| 48 | B | A ×4, D ×3, B ×1, B/D ×1 |
| 96 | C | A ×4, C ×3, A/C ×2 |
| 127 | C | B ×4, C ×4, D ×1 |
| 51 | not covered | D ×3, C ×2, B ×1 |
| 168 | B | B ×4, D ×4 |
| 55 | A | A ×5, B ×3, D ×1 |
| 67 | C | B ×5, C ×4 |
| 97 | B | C ×5, A ×4 |
| 137 | B | B ×5, C ×2, B/D ×1, D ×1 |
| 59 | D | C ×5, D ×3 |
| 100 | not covered | C ×5, A ×2, A/C ×1 |
| 147 | D | D ×5, A/D ×2, A ×1 |
| 1 | not covered | B ×6, C ×3 |
| 18 | B | B ×6, D ×3 |
| 38 | B | B ×6, A ×3 |
| 52 | A | A ×6, C ×2, A/C ×1 |
| 99 | not covered | C ×6, A ×3 |
| 141 | C | C ×6, D ×3 |
| 143 | A | A ×6, D ×3 |
How we did this
- Every key is translated into a single question series before anything is compared, using the same series mapping the score calculator uses, so question 12 in one series is matched to the same question in another.
- A question marked Z is a cancellation, not an answer, and is never counted as a disagreement.
- A question a key does not cover is excluded from that key’s comparison rather than scored as a mistake. A partial key is measured only on what it actually claims.
- Where institutes split evenly there is no majority, and we report no majority rather than picking one.
- Accuracy is agreement with the official final key, over the questions where both that key and the institute’s key give a real answer.
- Some institutes publish a combined answer such as C/D where they judged a question ambiguous. We treat that as its own distinct answer: it is counted as a disagreement with a key that gives a single letter, and it does not count as a match when the official answer is one of the two. So an institute that flagged ambiguity scores lower here than one that committed to the official letter, and that is a limitation of this table rather than a judgement about the institute.