JEE Advanced Score Calculator

From an answer-key check to a score

It turns your answer-key comparison into a score: the total, the maximum available, the percentage, a breakdown by paper and by subject, and an analysis of what you attempted and how much of it was right. It then does two things most score calculators do not.

The first is an honest account of negative marking. It tells you how many marks your wrong answers cost, what the score would have been if those same questions had been left blank, and — the part that actually matters — the accuracy at which attempting them was worth it in the first place. That last figure is computed from the marking scheme you entered, so it is right for your paper rather than a rule of thumb.

The second is a refusal. It does not estimate your rank. There is a section further down explaining why in detail, but the short version is that a rank depends on how eighty thousand other people did, and this page has no information about that whatsoever.

Everything runs in your browser. Nothing is uploaded, nothing is stored, and every mark used in the calculation is visible on the page and editable by you.

Why the marking scheme is an input

This is the design decision that shapes the whole page, so it is worth explaining rather than just asserting.

JEE Advanced does not use a fixed marking scheme. The number of sections, the question types within them, the positive marks, the negative marks and the partial-credit rules have all changed between years. They also differ between Paper 1 and Paper 2 of the same examination. A section that carried plus four and minus two one year may carry plus three and minus one the next, and a numerical section may have negative marking in one paper and none in the other.

That leaves a score calculator with two options. It can hard-code one scheme, in which case it is correct for one sitting and quietly wrong for every other — and a candidate has no way to tell which case they are in. Or it can ask.

This page asks. The specification it was built to actually agrees: it lists applicable positive marks and applicable negative marks among the inputs, not among the things the calculator is expected to know. So the page pre-fills the fields from common question-type patterns, marks them clearly as unverified, and lets you correct every one of them from your own information brochure.

The same reasoning applies to the year selector. It records which sitting your result refers to, so a printed or copied score is identifiable later. It changes no number, because this page holds no verified per-year scheme, and pretending otherwise would be the exact problem described above wearing a disguise.

Question groups: how to enter a paper

A JEE Advanced paper is not a uniform block of questions. It is a set of sections, each with its own question type and its own marking. The calculator mirrors that: you enter one question group per section, per subject.

A group is defined by four things — which paper it belongs to, which subject, what question type it uses, and what the marks are. Inside that group you enter four counts: how many you got right, how many wrong, how many you left blank, and, where the type supports it, how many were partially correct.

A typical Paper 1 might therefore be six or nine groups: three subjects multiplied by two or three sections each. That sounds like a lot of entry, and it is more than typing one number — but it is the only way to score a paper whose sections are marked differently, and it is the reason the subject and section breakdowns afterwards are real rather than estimated.

The page opens with three groups already set up, one per subject, so the intended shape is visible immediately. Add more with Add question group, remove any with the cross.

What each input means

Exam year

A label. It appears on your result and in anything you copy or print, so a saved score says which sitting it belongs to. It does not alter any mark.

Paper

Choose Both papers and each group gets its own paper selector, with the result reporting Paper 1, Paper 2 and the combined total separately. Choose a single paper and the selector disappears, because everything belongs to that paper.

Switching between the two does not lose anything. Your per-group paper assignments are remembered while the selector is hidden and come back exactly as they were when you switch to both papers again.

Subject

Physics, Chemistry or Mathematics. This drives the subject breakdown, which is usually the most useful part of the result — a total tells you where you stand, but the subject split tells you what to do next.

Question type

Choosing a type fills in the marks that pattern normally carries and, for partial-credit types, reveals two extra fields. It is a convenience, not a constraint: change any mark afterwards and the calculator uses what you entered.

Marks if correct, marks deducted if wrong

Read both from your brochure. The deduction is entered as a positive number — the calculator subtracts it. Entering it as a negative would apply the sign twice and turn your penalty into a bonus, which is exactly the sort of quiet error that makes a score calculator worse than useless, so the field rejects negative values outright.

Correct, wrong, unattempted

Whole numbers of questions. Blank counts as none. A group where all four counts are empty is ignored rather than reported as an error, so the starter rows you do not need cause no trouble.

Worked example: one paper

Worked example: both papers

Partial credit, and how to enter it

Multiple-correct questions are the ones people get wrong when calculating their own score, because they have three outcomes rather than two.

In a partial-credit scheme you can select all the correct options and get full marks; select only correct options but not all of them and get a smaller amount per option identified; or select any wrong option and take the full penalty. A calculator with only correct and wrong boxes cannot represent the middle case, so people force their partial answers into one of the other two and get a wrong total.

Choosing a partial-credit question type here reveals two extra fields: the marks awarded per correctly identified option, and a count of the questions where you got partial credit. Those are scored on their own line and reported separately in the breakdown, so you can see exactly what the partials contributed.

If your scheme has no partial credit, choose the all-or-nothing multiple-correct type instead and the extra fields disappear. Anything you had typed into them is cleared rather than silently retained, so a hidden field can never contribute to a total you cannot see.

Worked example: with partial marking

Accuracy and attempt rate are different questions

The analysis panel reports both, and the distinction is not cosmetic.

Attempt rate is how much of the paper you answered: attempted questions divided by total questions. Accuracy is how much of what you answered was right: fully correct answers divided by attempted. In example 1 they were 88.9% and 75.0% respectively.

Under negative marking these two pull against each other. Attempting more raises the attempt rate and, almost always, lowers accuracy — because the extra questions you attempt are by definition the ones you were less sure about. A candidate with 100% accuracy and a 40% attempt rate and a candidate with 60% accuracy and a 95% attempt rate can finish on very similar scores by completely different routes.

This is why neither number alone is a useful target. "Improve your accuracy" is trivially achieved by attempting less. "Attempt more" is trivially achieved by guessing. The score depends on the combination, and the break-even figure below is the thing that actually ties them together.

One note on how accuracy is defined here: it counts fully correct answers only. Partially correct answers are reported on their own line rather than folded into either side, because they are genuinely a third outcome and averaging them into "right" or "wrong" would make the figure mean less, not more.

What your wrong answers cost

This panel exists because most candidates know their score and have no idea what negative marking actually did to it.

It reports three things:

• Marks lost to wrong answers. The total deducted, across every group, at whatever penalty each carried.
• Score if those were left blank. What you would have scored had you not attempted the questions you got wrong. Since a blank earns nothing and costs nothing, this is simply your score plus the deductions.
• Wrong answers. How many there were.

In example 1 the four wrong answers cost 4 marks, and the score would have been 36 instead of 32 with them left blank.

That is not an argument for attempting less, and the page is careful not to present it as one. The comparison is deliberately incomplete: it shows what the wrong answers cost without crediting the attempts that succeeded, and any strategy that avoided those four wrong answers would almost certainly have avoided some correct ones too.

It is a diagnostic, not a verdict. The verdict needs the next section.

Break-even accuracy

Here is the figure that actually answers "should I have attempted that".

Suppose a question is worth p marks if right and costs n marks if wrong, and suppose your chance of getting it right is r. Attempting it is worth, on average:

Expected marks = r × p − (1 − r) × n

That is positive when r exceeds n ÷ (p + n). That fraction is the break-even accuracy: the success rate at which attempting stops costing you marks.

The calculator computes this from the marks you entered, for each distinct scheme in your paper, and puts your actual accuracy on that scheme beside it. In example 1 the break-even for +3/−1 is 25.0% and the candidate's accuracy on it was 75.0% — attempting paid off, by a wide margin.

Sections with no negative marking do not appear in that table. There is nothing to weigh up: a blank and a wrong answer are worth the same nothing, so attempting is free and the break-even figure of zero would only take up a row.

Two things this figure is not. It is not a prediction about a single question: expected value is a statement about many attempts, not about the one in front of you. And it is not a licence to guess, because your real accuracy on questions you are unsure about is lower than your overall accuracy — the overall figure includes all the ones you knew.

Used properly it answers a narrower and more useful question: given how this paper was marked, was my attempting strategy roughly right, or was I systematically attempting things I should have left?

Worked example: when guessing goes badly

The what-if simulator

The simulator answers "what would a few more correct answers actually be worth". Choose which group's marking to apply, enter a number of additional correct answers and additional wrong ones, and it shows the projected score and the difference.

Two design choices are worth knowing about.

The extra answers come out of your unattempted questions. They are questions you left blank and might have attempted, not new questions appearing in the paper. That is why the maximum score does not change, and why the simulator will not let you add more than that group actually has left blank — it caps the numbers and says so.

It applies one group's marking at a time. Adding three correct answers is worth nine marks in a +3 section and twelve in a +4 section, so a simulator that averaged across your whole paper would give a number that applied to nothing. Choosing the group makes the answer specific.

A worked run: a candidate on 10 marks from a group of 4 correct, 2 wrong and 4 unattempted at +3/−1. Converting three of those blanks to correct answers takes them to 19, a gain of 9. If one of the four had gone wrong instead — three right and one wrong — they would be on 18, a gain of 8. Attempting all four blanks and getting all four right gives 22.

That last number is the ceiling for that group, and the simulator will not go past it however large a figure you type. The difference between 18 and 22 is what "attempt everything" is worth here, and whether it is achievable is a question about you rather than about the arithmetic.

Why no rank is estimated

Every year, sites appear that convert a JEE Advanced score into a predicted rank. This page does not, and the reason is worth stating properly rather than hiding behind a disclaimer.

A score is a fact about you. A rank is a fact about everyone else. It is your position in a distribution of roughly a hundred and eighty thousand other results, and nothing about your own answer sheet determines it.

That distribution moves every year, for reasons nobody can predict in advance:

• Paper difficulty. A hard paper compresses scores downwards, so the same raw mark buys a much better rank. An easy paper does the opposite, and the shift between two consecutive years can be thousands of ranks at the same score.
• Cohort size and composition. The number qualifying from the screening examination changes, and so does who they are.
• Marking scheme changes. When the sections themselves change, last year's score-to-rank mapping is not measuring the same thing as this year's.
• Normalisation and revisions. Dropped questions and revised answer keys move a lot of candidates at once.

A rank predictor is therefore only as good as the dataset behind it, and it is only honest if it tells you which year's data it used and how it was collected. Most do not. What they usually offer is a curve fitted to whatever they could gather, presented with a precision the underlying data does not support — and a candidate reading "AIR 4,200" will take it far more literally than the person who built it intended.

The specification this calculator was built to says the same thing: a rank estimate is permitted only where it is based on an explicitly supplied historical model or dataset. No such dataset exists in this project, so no rank appears. If one is added later it will arrive with its source and its year stated on the page, and until then the gap stays visible rather than filled with a guess.

What this page gives you instead is a score you can trust, computed from a marking scheme you verified, with the working shown. For the rank, wait for the official result — it is the only one that counts.

Common mistakes

Not entering the unattempted questions

The most common and the most misleading. The maximum score is built from the questions you enter, so leaving out the blanks shrinks the maximum and inflates the percentage. A candidate who enters only their 16 attempted questions in example 1 would see 32 out of 48, or 66.7%, rather than the true 59.3%.

Entering the penalty as a negative number

The deduction field wants a positive figure — the calculator subtracts it. The field rejects negatives outright, which is deliberate: applying the sign twice would turn every penalty into a bonus and produce a score that looks plausible and is badly wrong.

Using one marking scheme for the whole paper

Sections differ, and Paper 1 differs from Paper 2. Entering everything as one group with one scheme is quick and produces a number that belongs to no real examination.

Forcing partial answers into correct or wrong

Worth six marks in example 3, in either direction. Use the partial count.

Trusting the pre-filled marks

They are patterns, not your paper. Open the brochure.

Reading the break-even figure as advice about one question

It is an average over many attempts. On any single question you either get it right or you do not, and the expected value does not apply to a sample of one.

Comparing raw scores across papers with different maximums

Example 2 shows a six-mark gap that is actually a twenty-eight-point gap in percentage terms, because one paper was marked out of 54 and the other out of 72.

Limitations

The marking scheme is yours, not ours. The templates here are common patterns and are labelled unverified. If you enter the wrong marks, the calculator will produce a wrong score confidently and without complaint. This is unavoidable in a tool that has to work across years and papers, which is why the marks are shown on screen at all times rather than hidden.

No rank, percentile or cut-off is produced. See the section above. Nothing on this page tells you where you stand relative to anyone else.

Official adjustments are not modelled. Dropped questions, revised answer keys, bonus marks awarded after review and any normalisation applied by the examining body will move the official figure away from this one. Those decisions are published after the examination and are not knowable in advance.

It cannot check your answer key comparison. The counts you enter are the counts it uses. If you miscounted your correct answers, the score will be wrong by exactly that much, and there is no way for the page to detect it.

Aggregate rules are not applied. Some sittings have used subject-wise minimum marks or other qualifying conditions alongside the aggregate. Those are institutional rules that change, and this page calculates the aggregate only.

Nothing is stored. Every calculation runs in your browser. Nothing is sent to a server and nothing survives closing the tab, so copy or print the result if you want to keep it.

Frequently asked questions