“I attempted 85” tells you very little without the number of wrong answers. Calculate the score before deciding whether an attempt strategy worked.
For CGL 2026 Tier-I, a correct answer earns 2 marks and a wrong answer loses 0.50. Tier-II Paper-I objective questions use +3 and −1. Official scheme, paragraphs 13.8–13.9.
Two candidates, different attempt counts
Here is an illustrative Tier-I calculation, not an actual result:
| Attempt | Correct | Wrong | Raw marks |
|---|---|---|---|
| A: 80 questions | 70 | 10 | 140 − 5 = 135 |
| B: 85 questions | 71 | 14 | 142 − 7 = 135 |
| C: 80 questions | 74 | 6 | 148 − 3 = 145 |
Candidate B answers one more question correctly than A but gains nothing after the additional mistakes. Candidate C earns ten more marks with the same attempt count as A.
Your review question becomes specific: which wrong answers were avoidable? Reading “incorrect” as “correct”, copying the wrong number, or selecting a different option from your working are errors you can investigate immediately.
What does the guessing calculation say?
Let p be your probability of answering correctly. The expected Tier-I score from one attempt is 2p − 0.5(1 − p). It becomes positive when p exceeds 0.20.
With four equally likely options, a genuinely random guess has an expected value of 0.125 marks under this simple model. That is a mathematical average over many trials, not a guarantee for one paper. A short sequence can lose marks. Time also has value: spending a minute on a weak guess can stop you solving a question you understand.
For the +3/−1 model, break-even is 0.25. Do not transfer the Tier-I calculation to a different paper without changing the scoring inputs.
The useful habit is calibration. In mocks, label uncertain attempts before checking the key. Over several papers, calculate how often those answers were actually correct. “I narrowed it down” is only helpful if your elimination is reliable.
Raw score is a review tool
Save the response sheet and distinguish the tentative answer key from the final key. A corrected or discarded question can change the calculation. SSC's 2026 challenge-management guidelines explain the different treatments; a challenged question does not automatically give everyone a bonus.
A raw total also does not reproduce a normalised result. Use it to assess your work, then use the official marks and result notice for the recruitment decision. The result, marks and cutoff guide explains that distinction.
Review your sectional timing alongside accuracy. Better question selection and fewer careless errors can matter more than forcing a headline attempt count.
Source checked: 20 September 2026. Evidence cycle: CGL 2026. All candidate scores and probability examples here are illustrative calculations, not predicted cutoffs.