How does UPSC normalize marks across different optional subjects?

UPSC officially uses statistical moderation by linear transformation, NOT subject-vs-subject scaling. The head examiner reviews sample scripts to standardise strictness across examiners within a paper — there is no public formula adjusting tough vs easy optionals against each other.

The Official Position

UPSC has stated in multiple Annual Reports and the Supreme Court (in UPSC v. Angesh Kumar, 2018 and earlier PILs) that it follows statistical moderation by linear transformation wherever considered necessary — not subject-vs-subject scaling like some state PSCs (RPSC, UPPSC). The 73rd Annual Report (2022-23) is explicit: moderation is an examiner-side tool, not a subject-side equaliser.

What Moderation Actually Does — The Mechanism

Step	Purpose	Outcome
1. Head examiner sets standards	Defines what a 10/10, 7/10, 4/10 answer looks like	Anchor scoring band
2. Sample script review	Re-checks 5-10% of each examiner's bundle	Detects strict/lenient outliers
3. Linear correction	Adjusts strict examiner's scores upward, lenient examiner's downward	Means converge across bundles
4. Borderline review	Scripts close to qualifying line re-checked	Catches edge cases
5. Final tally	Marks released after moderation	What appears on your marksheet

This fixes examiner subjectivity — not subject difficulty. Two candidates writing identical answers but graded by different examiners should now get the same mark.

What Moderation Does NOT Do

UPSC has never published a scaling formula between optionals and has denied using one in multiple RTI replies (CIC orders, 2010s). Specifically:

It does not convert raw Mathematics scores to a normalised band against Anthropology.
It does not apply Z-score formulas across optional subjects.
It does not use the Normalised Equi-Percentile method (that's a JEE/CAT technique, sometimes erroneously attributed to UPSC).

What coaching circles call "scaling" at UPSC is actually examiner-pool moderation within the same optional, not cross-optional equalisation.

Recent Optional Performance Data (CSE 2015-2024)

Optional	Avg. Toppers in Top 50	Typical Best Score	Volatility
Anthropology	4-7	320+	Low — predictable
PSIR	5-8	320+	Moderate
Sociology	3-6	310+	Low
Geography	4-6	315+	Moderate
Public Administration	1-3	300+	High since 2013 reform
History	2-4	310+	Moderate
Mathematics	1-2 (high years)	370+	Very high
Physics	0-1	340+	Very high

Maths and Physics produce very high scores in good years (one examiner-pool, internal consistency, objective evaluation) but suffer in years with strict head examiners. They are not "scaled down" — they are simply less forgiving.

Worked Scenario — Choosing an Optional

A candidate weighing PSIR vs Mathematics:

PSIR: Expected score band 260-310, low variance, 80% GS-II/GS-IV overlap.
Mathematics: Expected 280-370 if strong, but 230-260 if a strict year. High variance.
Expected value: PSIR ~285 (low risk), Maths ~305 (high risk + reward).
Decision rule: if you're risk-averse or new to the exam, PSIR. If you're a strong mathematician and targeting top-100, Mathematics offers ceiling.
Mythical scaling boost: zero. Pick on syllabus fit and interest, not on a phantom multiplier.

Mentor Note

Choose an optional based on three real factors: (i) genuine interest (you'll spend 600+ hours on it), (ii) syllabus length and resource availability, (iii) GS overlap — not on a phantom scaling advantage that does not exist in UPSC's official documents or RTI disclosures. The toppers' optional mix every year reflects this — Anthropology, PSIR, Sociology, Geography dominate not because they're scaled up, but because they're predictable, finite, and GS-overlapping.