SSC CGL Exam Pattern and Normalization: Marks, Timing and How Shift Scores Are Adjusted

You wrote the exam in the third shift, the paper felt heavy, and your friend in the first shift said it was a breeze. Now the result comes and your marks are lower than his. Was that fair? SSC's answer is normalization. This post has two parts: the pattern (marks, questions, time, negative marking) and then normalization, with a small worked example so you can see what actually happens to a raw score.

Sources: the SSC notices for CGL 2025 and CGL 2026, SSC's normalization notice dated 2 June 2025, and several exam-prep sites that summarise them. I flag the points where the sources disagreed or where I could not verify a detail on the official site. Check ssc.gov.in before you rely on any figure for an application or a bet on strategy.

Tier 1 pattern

Tier 1 is a computer-based test. It is qualifying. The marks do not count toward your final merit, and you only need to clear the cut-off to be shortlisted for Tier 2.

Section	Questions	Marks
General Intelligence and Reasoning	25	50
General Awareness	25	50
Quantitative Aptitude	25	50
English Comprehension	25	50
Total	100	200

Each question carries 2 marks. Each wrong answer costs 0.50 marks. Unattempted questions score zero. Total time is 60 minutes (80 for candidates with a scribe, as the notice specifies).

Sectional timing: what changed

Until the 2025 cycle, you had one 60-minute clock and could move between sections freely. For the 2026 cycle, the reports based on the notice say each section now has its own 15-minute timer and locks when it ends. One secondary source also said sectional timing has been introduced for Tier 2 but that the minutes had not been published. I could not verify the Tier 2 part from the official page, so treat that as unconfirmed.

Some concrete effects:

• You cannot bank time. If you finish Reasoning in 10 minutes, those 5 minutes are gone.
• Marked-for-review questions can only be revisited within the same section's window.
• General Awareness now arrives in a fixed slot, and you cannot skip it for later.
• The old habit of "start with English, finish with Maths" or any custom order no longer works. The order is fixed by the interface.

The maths of it: 25 questions in 15 minutes is 36 seconds each. In practice that means some questions get 15 seconds and a few get 60. Your mock practice must reflect that. If you have been taking 60-minute mocks with free movement, switch to timing each section separately. On Pareeksha's full-length mocks, practise with a stopwatch beside you for each 15-minute block, because the timer on any given mock may follow the older format.

How to think about attempts in Tier 1

With 0.50 negative marking on a 2-mark question, a wrong answer costs a quarter of what a right answer earns. Let me put numbers on it. Suppose you can narrow a question to two options and pick one at random. Expected value is 0.5 x 2 minus 0.5 x 0.5 = 1 - 0.25 = 0.75 marks. Positive. Suppose you cannot eliminate anything and there are four options: 0.25 x 2 minus 0.75 x 0.5 = 0.5 - 0.375 = 0.125. Still slightly positive, but barely worth the time. The trade is time, not marks. See negative marking, when to skip for the section-wise version.

The sensible rule under sectional timing: attempt everything you can eliminate down to two, and skip questions where you have no clue. Do not spend 90 seconds on a question you will guess anyway.

Tier 2 pattern

Tier 2 is where your final merit comes from. It has up to three papers, depending on the post you are eligible for.

Paper I (compulsory)

Section	Module	Questions	Marks	Time
I	Mathematical Abilities	30	90	30 min
I	Reasoning and General Intelligence	30	90	30 min
II	English Language and Comprehension	45	135	40 min
II	General Awareness	25	75	20 min
III	Computer Knowledge Test	20	60	15 min
III	Data Entry Speed Test	one typing task	qualifying	15 min

Every question in the scored modules carries 3 marks. So Section I is 60 questions and 180 marks, Section II is 70 questions and 210 marks, and Section III's computer test is 20 questions and 60 marks. Total: 150 questions, 450 marks.

A heads up about data quality. When I compared five prep sites, two of them listed the Maths module as 180 marks, which is actually the Section I total, and one listed a Paper I total of 620 or a different figure. The pattern above is the one that adds up: 150 questions at 3 marks each is 450. If you see a different number, check the notice.

Negative marking in Paper I is 1 mark for each wrong answer in Section I, Section II and Module I of Section III (the Computer test). The Data Entry Speed Test is a typing task, so it has no MCQ marking to worry about.

The Computer Knowledge Test and the Data Entry Speed Test are qualifying in nature. The details of how the qualifying threshold applies to different posts are in the notice. One source I read said the DEST is compulsory for all posts in the 2025 notice; older notices linked it to specific posts. I could not confirm the current rule from the official notice, so check it. The DEST requirement commonly quoted is 2,000 key depressions in 15 minutes.

Paper II (Statistics)

Only for Junior Statistical Officer and Statistical Investigator Grade II. 100 questions, 200 marks, 2 hours, 0.50 negative marking per wrong answer.

Paper III (Finance and Economics)

Only for Assistant Audit Officer and Assistant Accounts Officer. 100 questions, 200 marks, 2 hours, 0.50 negative marking.

Qualifying marks

The minimum qualifying marks reported for Tier 2 papers are 30% for UR and EWS, 25% for OBC, and 20% for SC, ST and other reserved groups. Tier 1's minimum qualifying marks are the same, and the reports on CGL 2025 say the unreserved minimum was 30%. These are only the floor. The actual cut-off for shortlisting is decided by the number of vacancies and how the marks are distributed, and it is higher than the floor.

What the pattern means for how you prepare

Tier 2 negative marking changes the arithmetic

At 3 marks per question and a 1-mark penalty, the break-even accuracy for a random guess is 25%, exactly the accuracy of guessing among four options. Look closer. If you eliminate one option and guess among three, your expected value is (1/3 x 3) - (2/3 x 1) = 1 - 0.667 = 0.333 marks per question. If you eliminate two options and guess between the last two, it is (1/2 x 3) - (1/2 x 1) = 1 marks. So in Tier 2 a two-option guess is clearly worth taking. A no-clue guess is worth about zero, so skip and save time.

Compare that with Tier 1. In Tier 2 a wrong answer costs one-third of the value of a right answer, while in Tier 1 it costs one-quarter. Slightly harsher, but not drastically so. The bigger difference is that Tier 2 questions are longer, so guesses cost more time.

Time per question in Tier 2

• Maths: 30 questions in 30 minutes, one minute each, but questions vary hugely. Plan on skipping 3 to 5 and returning.
• Reasoning: 30 questions in 30 minutes.
• English: 45 questions in 40 minutes. About 53 seconds each. Reading comprehension takes the largest chunk, so grammar and vocabulary need to be fast, say 20 seconds each.
• General Awareness: 25 questions in 20 minutes. Either you know it or you do not, so spend at most 20 to 30 seconds per question.

The gap between Section I and Section II in terms of time to reflect on each question is real. Practise each block separately.

Selection uses Tier 2 only

Your final merit is computed from the Tier 2 papers relevant to your post. Tier 1 marks are not added. This is why some people who scored high in Tier 1 fail to get posts, and others who barely cleared it do well in Tier 2.

Why normalization exists

CGL Tier 1 is held over many days and in multiple shifts. Papers are different in each shift. Even with careful question-setting, some are tougher than others. If SSC ranked raw marks across shifts, candidates in the harder shifts would be penalised through no fault of their own. So the Commission converts raw scores into normalized scores.

For years, SSC used a formula based on the mean and standard deviation of each shift, published in a 2019 notice. On 2 June 2025 the Commission issued a new notice adopting the equipercentile method. The CGL 2025 notification stated that marks in multi-shift computer-based exams would be normalized using the procedure published in that 2 June 2025 notice. That is the method that applied to CGL 2025 and will apply to CGL 2026 unless a new notice says otherwise.

The equipercentile method in plain terms

Here is the idea. Instead of asking "what is the average and spread in each shift", the method asks "what fraction of candidates in my shift scored below me, and what score would that fraction correspond to in the other shifts?"

The steps as summarised by prep sites and the official document:

1. Convert your raw marks into a percentile within your shift. In effect, the percentile is the proportion of candidates in your shift who scored at or below you. If 80,000 candidates appeared in your shift and 64,000 scored at or below your marks, your percentile is 80.
2. Line up all shifts on the percentile scale. For every shift, you have a table: percentile against raw score. Because a given percentile may not appear exactly in each shift's table, the gaps are filled by interpolation, so the mapping from percentile to marks is smooth.
3. Assign the normalized score. For your percentile, look up the raw score at that percentile in every shift. The normalized score is then a combined value of those, described in secondary sources as their average.

The intuition is that a person at the 80th percentile in a hard shift is "as good as" a person at the 80th percentile in an easy shift, even though the raw scores differ. The method assumes that the candidate pool in each shift is roughly similar in ability, which is reasonable when there are many thousands of candidates per shift and shifts are randomly assigned.

A worked example with made-up numbers

These numbers are invented to show the mechanics. They are not from any real CGL shift.

Say there are three shifts. At the 80th percentile, the raw marks are:

Shift	Raw marks at 80th percentile
Shift A (harder paper)	120
Shift B	132
Shift C (easier paper)	144

Take a candidate in Shift A who scored 120, which is exactly the 80th percentile there. The candidate would get a normalized score equal to the combined value across shifts at the 80th percentile, here (120 + 132 + 144) / 3 = 132.

A candidate in Shift C who scored 144, also at the 80th percentile in their shift, gets the same normalized score of 132. Their raw marks are 24 apart, but the normalized scores are equal, because they beat the same share of their own shift.

Now a candidate whose raw score lies between table entries. Say the Shift A table says 116 is the 76th percentile and 122 is the 82nd. A candidate with 120 is between. Interpolating: (120 - 116) / (122 - 116) = 0.667, so the percentile is 76 + 0.667 x 6 = 80. That is where the interpolation step comes from. Then the same process gives the normalized value from the other shifts' tables.

This is why the normalized score can be higher or lower than your raw score. It is higher if your shift was harder than average, lower if easier.

Things I could not verify

Secondary sources describe the procedure, but I did not find the official document's full text, so a few details are not confirmed:

• How ties (many candidates with the same raw marks) are handled in the percentile calculation.
• Whether normalization is applied at the level of the total score or separately by section.
• The precise formula for combining scores across shifts, beyond "average" as described by secondary sources.
• How the method works when the candidate count in a shift is small, such as special-batch shifts.

If any of these matter to you, read SSC's 2 June 2025 notice on the official site. It is short.

What you can and cannot do about normalization

You cannot game it. Guessing that your shift was hard and slacking off accomplishes nothing. The method looks at your rank inside your shift, not at how hard you felt the paper was.

Do not compare raw scores across shifts. A 140 in one shift and a 135 in another tell you nothing about who did better until normalized. Pareeksha's percentile after each mock uses the same idea of rank among test-takers, so use it as a rough guide to your standing, not a prediction of the real result.

Focus on rank within a large pool. A percentile-based method rewards being better than most candidates at the same time, however hard or easy the paper. So the relevant question in a mock is "where do I rank among takers", not only "what did I score". This is why a mock platform with a good number of takers is worth more than one with thin data.

Expect some noise. Normalization does not eliminate difficulty differences perfectly. There will be people who feel it hurt them. Accept it and use your energy on the raw score.

A 15-minute plan for each Tier 1 section

Since each section is now its own small exam, it helps to have a script for each. These are starting points, not rules.

Reasoning (25 questions). Do the quick ones first: series, coding-decoding, analogy, classification, blood relations. Those should take 20 to 30 seconds each, which buys you about 4 to 5 minutes for the slower questions, like a seating arrangement or figure counting. If a question is still unclear after 45 seconds, move on. You can come back within the section.

General Awareness (25 questions). This is the section where time is least useful. If you know the answer, you know it inside 10 seconds. Read each question, answer, move. You will probably finish in 6 to 8 minutes. Use the remaining time to go back over the questions where you were between two options. Do not go looking for time to transfer to Maths, since it is not possible any more.

Quantitative Aptitude (25 questions). The tightest section. A workable split: 12 to 14 questions at 25 to 30 seconds each (percentage, ratio, averages, simple algebra, quick mensuration), then 8 to 10 questions at 50 to 60 seconds (time and work, geometry, data questions), and skip the last few which need long calculations. Leaving 3 to 5 questions is normal and correct, not a failure.

English (25 questions). Grammar and vocabulary questions in 15 to 20 seconds each, then the comprehension passage. If reading comprehension is your weak spot, read the questions first and then look for the answers in the passage. Cloze and para jumbles go last.

Two more points. First, practise going in the sequence the interface gives you, whatever it is, so you do not lose the first minute working out where you are. Second, some students freeze when a timer locks a section with questions still unattempted. Rehearse that in mocks so that it is not a surprise on the day.

Reading a mock result properly

After a mock, look at four numbers per section, not just the total.

• Attempts: how many questions you answered.
• Accuracy: right answers divided by attempts.
• Net marks: right answers times 2 minus wrong answers times 0.5.
• Time spent when you finished, if the platform shows it.

Suppose you attempted 20 in Maths, got 15 right and 5 wrong. Net marks are 30 minus 2.5, which is 27.5 out of 50. Accuracy is 75%. A friend attempted 15, got 13 right and 2 wrong. Net is 26 minus 1, so 25 out of 50, accuracy 87%. You scored more with lower accuracy. Neither approach is better in general. It depends on whether your accuracy falls as you attempt more. If it does not, attempt more. If your extra 5 attempts produce 5 wrong, you should stop earlier.

The percentile shown by Pareeksha after each mock is the second thing to check. It only tells you how you did relative to the people who took that mock, which is a useful signal of rank, though it is not the same as the normalized rank in the real exam.

The other pattern details worth knowing

Language

Question papers are in English and Hindi, except the English language sections, which are English only. Regional language options for some posts and exams have been offered in the past. Check the notice.

Scribes and extra time

Candidates eligible under the benchmark-disability rules get additional time as specified. The notice gives 80 minutes for Tier 1 for scribe users, per the reports I read.

Physical documents and biometrics

Aadhaar-linked authentication and biometric checks have been in use for SSC exams. One report I found mentioned the SSC chairman outlining reforms such as a single-shift approach for Tier 1 in one edition, centre allocation within a limited distance, and Aadhaar authentication. That was reported for CGL 2025 and I could not confirm the final form from the official page. In any case, take your originals and follow the admit card exactly.

After Tier 2

Post preference and document verification follow, covered in SSC CGL salary and post preference.

How to practise for this pattern

A four-week block for the pattern:

Week 1. Take one Tier 1 mock with a 15-minute limit per section, stopping at the bell. Do not adjust anything. Record how many questions you attempted in each section and your accuracy.

Week 2. Take two Tier 2 style sets of 30 Maths questions in 30 minutes. Record the number of questions you skipped and how many were errors from rushing.

Week 3. Take a full Tier 1 mock and a 45-question English set. Compare the marks you lost to negative marking against the marks you lost by not attempting. If negative marking costs you more than skipped questions do, you are over-attempting.

Week 4. Repeat a Tier 1 mock. You should see a clear improvement in accuracy or attempts.

For the mocks, the free mini-mocks and sectional tests let you rehearse a 15-minute section without committing to a full paper, which is a good match for sectional timing.

FAQ

Is Tier 1 in one shift or many? It has been held in multiple shifts across many days in most years. Reports for CGL 2025 mentioned an intention of a single-shift Tier 1, but the exam was still spread across dates, and the 2026 cycle covers 30 September to 30 October 2026 as reported. Check the admit card for your own schedule.

Does normalization apply to Tier 2? Tier 2 Paper I was reported as held on two days, 18 and 19 January 2026, for CGL 2025. When an exam is held in multiple shifts, the same procedure applies. Confirm in the Tier 2 notice.

Is the marking really 3 marks per question in Tier 2 Paper I? Yes for the scored modules as per the pattern above: 150 questions and 450 marks.

What is the minimum qualifying percentage? Reported as 30% for UR and EWS, 25% for OBC, 20% for SC, ST and others. It is the floor, and the actual cut-off is higher.

Can I choose the order of sections in Tier 1? With sectional timing, no. The order is fixed.

How does normalization affect my percentile in a mock? Mock percentiles are computed only among that mock's test-takers, with no normalization. They tell you your standing among those takers, not your standing in the real exam.