How to Self-Evaluate a UPSC Mains Answer When No One's Grading It

There are thirty or forty answers in that folder. You wrote them over four months, most on paper, under something close to real time pressure. You read each one back a few minutes after finishing, decided it was roughly fine, and moved on to the next.

Nobody has read them. Not properly — not the way an examiner would.

And you already know that reading an answer back ten minutes after writing it isn’t evaluation. But the alternative on offer seems to be inventing a mark — deciding this one was probably a 9 out of 15 — and you know that number isn’t real either. So the folder grows, and the question underneath it stays unanswered: are these getting better, or am I just getting faster at the same mistakes?

You can't mark your own answer. You can audit it — and you can catch many of the most fixable problems yourself.

01 · The open loop

The answers nobody has read

The loop most aspirants are running looks complete from the outside. You study a topic. You attempt a question on it. You write the answer with the clock going. That is three of the four steps, and already more than many people manage — if you are writing at all, you are doing the hard thing most aspirants defer.

The fourth step is the one that turns the first three into preparation: something reads what you wrote and tells you what it did and didn’t do. Without it you aren’t practising so much as rehearsing — answers written without correction are repetitions of the same habits, some working, some not, and nothing in the stack tells you which is which.

A test series closes part of that loop for some aspirants — a number comes back later with a couple of lines of comment. Many have no external reader at all. Either way, the answer in front of you tonight will not be read by anyone but you.

So the useful question isn’t whether self-evaluation matches an examiner. It plainly doesn’t. It’s what you can reliably do with an answer nobody else will read.

02 · The wrong first move

Stop guessing the number

The instinct is to grade. You finish, you look at the page, you try to decide what it would have scored.

That instinct fails for two reasons, and they compound.

The first is calibration. A mark out of 15 is a comparative judgement — this answer, set against the range of answers that exist. Producing one needs a reference set you don’t have. You’ve never watched your own script being marked, and you can’t hold it against the others at known scores that would give the number meaning. You aren’t being lazy when you can’t do this. You’re being asked to measure without a scale.

The second reason matters more, and it’s harder to notice. You know what you meant. When you read your own sentence, you read it carrying everything that was in your head while you wrote it — the connection you intended, the example you almost added, the qualification you assumed was implied. The examiner reads the words. Which means your own reading of your own answer runs generous in a way you cannot feel from the inside, and the more thinking that went into the answer, the more generous it runs.

And a number doesn’t tell you what to change. “That was probably a 7” is not an instruction — even if the 7 were accurate, there’s nothing in it you can do differently tomorrow.

03 · Two different jobs

Grading needs an examiner. Checking doesn’t.

There are two separate jobs hiding inside the phrase “evaluate this answer.”

One is judgement — how good is this, set against what a strong answer looks like? That needs an examiner, or something standing in for one, because it needs a reference outside your own head. It’s the job you’re poorly positioned to do alone.

The other is inspection — did this answer do what the question asked it to do? That’s a different species of question. It doesn’t require you to know how good the answer is; it requires you to hold the answer against the question, and both are on the table in front of you.

The second job is the one you can do well. It’s also the one most aspirants skip past on the way to attempting the first.

04 · Five checks

What you can reliably inspect on your own

So here is the audit: five things, in this order. Each is scoped so that you can make a useful first assessment without needing to know exactly what a stronger answer would have looked like — working from what’s written and what was asked.

01 · Sub-demands

Break the question into every distinct thing it asked for, then find where each is answered. A question with three demands and an answer addressing two isn't a matter of opinion — it's a gap you can see, and often one of the most consequential gaps on the page.

02 · Directive

The directive named an intellectual task — examine, discuss, critically examine. Find the sentences that perform it. If "critically examine" produced description and no judgement, that shows up plainly on a re-read.

03 · Shape & proportion

Is there a visible opening, body and close? And did one demand quietly eat most of the word budget while another got a single line? Paragraph lengths show you this before you read a word.

04 · Evidence & examples

Take the major claims and ask which have something concrete behind them — an example, a scheme, a committee, a case, a data point, a constitutional provision. Bare claims are among the most commonly recoverable weaknesses in a Mains answer.

05 · Conclusion

Read only the last few lines, on their own. Do they add a final judgement, a direction, a consequence — or restate what's already above in slightly different words? A conclusion that only summarises spends your last thirty words without buying anything.

The order matters. Coverage first, because an unaddressed demand tends to cost more than a weak paragraph — then the task, the structure, the substantiation, the close.

None of it needs an hour. It needs a pen and five minutes, and it goes better if you don’t start by reading the answer straight through: a full re-read pulls you back into the argument you were making.

The five-minute pass · one minute each

Minute 1 — Underline every demand in the question. Tick each one where the answer addresses it.
Minute 2 — Circle the directive. Point to the sentences that execute it.
Minute 3 — Mark where your examples, schemes and data actually appear on the page.
Minute 4 — Look at paragraph lengths without reading them. Did one demand dominate?
Minute 5 — Read only the conclusion. Does it add a final judgement?

Do it the next day rather than the same evening, if you can. Distance from your own intentions is the one advantage you can manufacture for free.

05 · The rule that makes it work

The pointing test

Notice how each check is phrased. Not “was my analysis deep?” but “which sentence does the judging?” Not “did I use good examples?” but “where on this page is the example?”

That phrasing is the entire mechanism. “Is my analysis good?” can be answered from memory, from intention, from the feeling you had while writing — all the sources that run generous. “Point to the line” can only be answered from the page. You either put your finger on it or you don’t.

If you can't point at the line that does it, it isn't on the page. And the examiner reads the page.

It’s also why the pen matters more than it looks. The marking is the proof: an underline you couldn’t place is a demand you didn’t answer, and it stops being arguable the moment you go looking for it.

06 · The honest boundary

What you can know yourself — and what you can’t

An article that stopped here would be quietly dishonest. The audit closes part of the feedback gap. It doesn’t close all of it, and it’s worth being precise about where the line falls.

What you can establish on your own:

Whether every demand in the question got an answer
Whether the directive’s task was actually performed
Whether the answer is proportionate across its demands
Whether your claims have concrete support attached to them
Whether the conclusion does more than restate

What you can’t reliably establish on your own:

Whether your argument is genuinely insightful, or merely competent
Whether the examples you chose are the strongest ones available for that question
Whether you’ve missed an important dimension the question invited
How your answer reads next to a strong answer on the same question
How your articulation lands for a reader who doesn’t already know what you meant

Every item on the second list needs a reference outside your own head. And one of them is the real ceiling on self-evaluation — you cannot inspect the angle you never thought of. A missing sub-demand announces itself, because the question names it. A missing dimension doesn’t: nothing on your page or in the question points at the thing you didn’t consider. It isn’t a gap you overlooked. It’s a gap you had no way to see.

07 · The counter-question

“Isn’t this just marking your own homework?”

Partly, yes. That’s exactly why the scope matters.

If self-evaluation meant deciding how good your answer was, the objection would be fatal — you’d be the least qualified person in the room to make that call, and the most motivated to be kind about it. That’s the version this article argues against.

What’s left is narrower and much sturdier. Checking whether a three-part question got three-part treatment isn’t a judgement about your ability; it’s a comparison between two documents. You can do it badly by rushing, but the check is less vulnerable to the generosity that affects judgements of depth or quality: you can point to where each demand was addressed and then ask whether the response actually engages with it.

The audit is bounded on purpose. Don’t attempt the part you’re poorly positioned to do. Do the parts you can genuinely inspect. And keep looking for an outside reader for the rest.

08 · Where Swādhyāya fits

The second reader

The audit is yours, and it doesn’t need a platform. What a preparation system can add is a second perspective on the part the previous section said you can’t reach alone:

Know what the question was actually asking. Question Intelligence lays out the directive, the examiner’s intent and the sub-demands — the reference you can’t generate yourself, because your reading of the question is one of the things under inspection.
Know how the answer reads to someone who isn’t you. Evaluation provides a second perspective across the dimensions — structure, coverage, substantiation, interlinkages, the close — with specific feedback on each. Not a verdict on the answer; another reading to hold yours against.
Know what keeps recurring. Dimension-wise scores across your attempts show whether the same weakness turns up answer after answer. One answer can never tell you that. Twenty can.

None of this replaces the pen-and-paper pass. It works better after it.

You don’t need to wait for someone to grade every answer before you can learn from it. But you also don’t need to pretend you can be your own examiner.

Inspect what you can. Get another perspective where you can’t.

One re-read at a time.