Odisha PSC – Foundation & Exam Strategy
Topic: Marking Scheme
Audience: Advanced university students (MIT‑level) preparing for the Odisha Public Service Commission (PSC) Foundation exam.
Goal: Provide a rigorous, first‑principles understanding of how marks are allocated, the logic behind negative marking, edge cases, and a challenging problem that tests mastery of the material.
1. Overview of the Odisha PSC Foundation Exam
| Feature | Details |
|---|---|
| Exam Type | Written test (objective) |
| Paper 1 | General Knowledge (GK) |
| Paper 2 | General Science |
| Paper 3 | Logical Reasoning & Numerical Ability |
| Paper 4 | Personality & Aptitude |
| Total Marks | 400 |
| Duration | 3 hours |
| Answer Sheet | Computer‑based (multiple‑choice) |
| Marking Scheme | 1 mark per correct answer, 0.25 marks penalty per wrong answer (negative marking) |
2. First‑Principles Derivation of the Marking Scheme
2.1. Expected Value Argument
Let:
- (p) = probability of a candidate answering a question correctly by guessing.
- (r) = probability of answering incorrectly.
- (w) = weight of a correct answer (+1).
- (l) = weight of a wrong answer (negative).
The expected contribution of a random guess:
[ E = p \cdot w + r \cdot l. ]
We want (E \le 0) so that guessing does not improve the expected score. For a uniformly random guess among the 4 options:
[ p = \frac{1}{4}, \quad r = \frac{3}{4}. ]
Setting (E = 0):
[ \frac{1}{4}(1) + \frac{3}{4}l = 0 ;\Rightarrow; l = -\frac{1}{3}. ]
The commission adopted a simpler form: (l = -0.25) (a quarter of the positive mark). This is slightly less punitive than the theoretical optimum but still discourages random guessing.
Why 0.25?
- Fairness: allows a small reward for a lucky guess.
- Simplicity: easy to remember and implement.
- Policy: aligns with many state‑level exams in India.
2.2. Weighted Marking Across Papers
Each paper has 100 marks. The weighting per correct answer is thus:
[ w_{\text{paper}} = \frac{100}{n_{\text{questions}}} ]
where (n_{\text{questions}}) is the number of multiple‑choice questions in that paper.
- Paper 1 & 2: 50 questions → (w = 2) marks per correct.
- Paper 3: 40 questions → (w = 2.5) marks per correct.
- Paper 4: 30 questions → (w = 3.33) marks per correct.
Negative marking per wrong answer:
[ l_{\text{paper}} = 0.25 \times w_{\text{paper}} ]
| Paper | #Q | (w) (per correct) | (l) (per wrong) |
|---|---|---|---|
| 1 | 50 | 2.00 | 0.50 |
| 2 | 50 | 2.00 | 0.50 |
| 3 | 40 | 2.50 | 0.63 |
| 4 | 30 | 3.33 | 0.83 |
3. Edge Cases and Practical Implications
| Edge Case | Explanation | Strategy |
|---|---|---|
| Multiple correct options | Some questions may have “all of the above” or “none of the above”. | Treat as a single option; no additional penalty. |
| Partial marks | No partial credit in objective format. | Focus on eliminating wrong options first. |
| Question with 5 options | Uncommon, but may appear. | (p=\frac{1}{5}). Expected value with 0.25 penalty: (E=0.2-0.75= -0.55). Even more discouraging to guess. |
| Delayed answer sheet | If the answer sheet is printed after the exam, candidates may misprint. | Double‑check each answer; use the diary method. |
| Answering all questions | If a candidate answers all questions, expected net gain from random guessing ≈ 0. | Optimal strategy is to skip only those questions where the candidate has no confidence. |
| Score ceiling | Maximum score = 400. | In high‑confidence candidates, skipping low‑confidence questions can still keep scores high. |
4. Advanced Strategy for Maximizing Expected Score
4.1. Bayesian Updating
Let (q) be the prior probability that a candidate knows the answer. After eliminating (k) options, the posterior probability (q_k) of being correct is:
[ q_k = \frac{q}{q + (1-q)\frac{1}{4-k}} ]
The expected value of choosing to answer (vs. skipping) becomes:
[ E_{\text{answer}} = q_k \cdot w - (1-q_k) \cdot l. ]
If (E_{\text{answer}} > 0), answer; else skip.
Example: Suppose (q = 0.3), (k=1) eliminated option → (q_1 = \frac{0.3}{0.3 + 0.7 \cdot \frac{1}{3}} \approx 0.5).
With (w=2), (l=0.5):
[ E_{\text{answer}} = 0.5 \cdot 2 - 0.5 \cdot 0.5 = 1 - 0.25 = 0.75 > 0 ]
Hence, answer.
4.2. Optimal Question‑Selection Algorithm
- Read all questions quickly.
- Rank by confidence.
- Apply Bayesian rule to decide answer/skip.
- Keep track of time per question; aim for ~4.5 s per question in Paper 1/2, ~6 s in Paper 3, ~7.5 s in Paper 4.
5. Edge‑Case Simulation
| Scenario