Negative Marking, Time Management and Mock Strategy for APPSC, DSC and AP Police

Most aspirants prepare the syllabus and leave the paper itself to chance. They do not decide in advance how many questions they will attempt, how long each block of the paper deserves, or what they will do with a question where two options look right. Those decisions are worth a surprising number of marks, and unlike the syllabus, they can be learned in a few weeks of deliberate practice. This guide gives the arithmetic behind them. It starts with what the official documents of the Andhra Pradesh Public Service Commission (APPSC), the School Education department (AP DSC) and the State Level Police Recruitment Board (SLPRB) actually say about negative marking, because a strategy built on the wrong penalty is worse than no strategy.

Every exam fact below is labelled "As on October 2026" and names the document it comes from. Where a notification does not say something, this guide says "not mentioned" and does not fill the gap with a guess. The examples with invented numbers are labelled illustrative. None of them describes a real exam, a real cohort or a real result.

What the official documents say about negative marking

As on October 2026, the position differs by recruiting body. Read the table first, then the notes under it.

Recruiting body and document	Negative marking	Paper pattern printed in the document
APPSC Group-I Notification 07/2026 (dated 06.10.2026), Screening Test	One third of the marks for the question, per G.O.Ms.No.235, Finance (HR-I, Plg and Policy), 06.12.2016	Two objective papers, 120 questions, 120 minutes, 120 marks each
APPSC Assistant Statistical Officer Notification 20/2026 (15.09.2026)	One third, same G.O.	Two objective papers, 150 questions, 150 minutes, 150 marks each
APPSC Forest Beat Officer Notification 18/2026 (15.09.2026)	One third, same G.O.	Qualifying essay (50 marks, 45 minutes), then two objective papers of 100 questions, 100 minutes, 100 marks each
APPSC Executive Officer Grade-III 23/2026 and Supervisor Grade-I 13/2026	One third, same G.O. printed in the notification	See the notification of the post
SLPRB Sub-Inspector Notification (Rc.No.81, 16.09.2026)	Not mentioned. The notification says full marks for one correctly darkened bubble and zero if no bubble is darkened	Preliminary: two papers of 100 questions, three hours each. Final objective papers: 200 questions each
SLPRB Police Constable Notification (Rc.No.91, 16.09.2026)	Not mentioned. Same awarding-of-marks wording	Preliminary: one paper, 200 questions, 200 marks, three hours. Final: 200 questions
AP Mega DSC 2025 Information Bulletin (School Education)	Not mentioned in the bulletin	SA and SGT: 160 MCQs, 80 marks, 2 hours 30 minutes
APTET June 2026 Information Bulletin	None. The bulletin says there shall be no negative marking	MCQs of one mark each, four options

Three points need emphasis.

First, the APPSC rule is not a custom of one exam. The same sentence, citing G.O.Ms.No.235 of 6 December 2016, appears in each of the APPSC notifications opened for this guide. For any APPSC objective paper, assume a deduction of one third of the marks for the question, and confirm it in the notification of your post. For a one-mark question, a wrong answer costs 0.33 marks. For a question worth two marks it would cost 0.67. An unanswered question costs nothing.

Second, "not mentioned" does not mean "no penalty". The SLPRB notifications describe how marks are awarded and say nothing about wrong answers. The Mega DSC 2025 bulletin was searched for any penalty language and none was found. That is a statement about the documents, not about the exam. The 2026 notification, the OMR or on-screen instructions and the hall ticket are the final word. Until you have read them, the safe habit is to record two scores in every mock test: one with no penalty and one with a one-third penalty. If the real paper turns out to have no penalty, you attempt everything. If it has one, your habits are already tuned for it. The APTET June 2026 bulletin is the clear exception, because it says in plain words that there is no negative marking.

Third, the marks per question are not always one. In the SLPRB Sub-Inspector final written examination, the 200-question papers carry 200 marks for one post code and 100 marks for two others, so a question is worth one mark in the first case and half a mark in the second. In Mega DSC 2025 the 160 questions of a School Assistant or SGT paper carry 80 marks, which is half a mark each. If a penalty applies, it is a fraction of that question's marks, so the arithmetic below scales proportionally.

The arithmetic of guessing

Suppose a question carries one mark, has four options, and the penalty for a wrong answer is a fraction p of a mark. If you pick one option at random, the chance of being right is one in four, and the chance of being wrong is three in four. The expected change in your score is:

expected value = (1/4 x 1) minus (3/4 x p)

With a penalty of one third, that is 0.25 minus 0.25, which equals zero. With a penalty of one quarter, it is 0.25 minus 0.1875, which equals plus 0.0625. A blind guess at one third is therefore exactly neutral on average, and a blind guess at one quarter is very slightly favourable. The word "average" matters, and it is explained below.

The useful generalisation is the break-even probability. If your chance of being right is q, the expected value of attempting is q x 1 minus (1 minus q) x p. Setting that to zero gives q = p / (1 + p). For p = 1/3 the break-even is 1/4, which is 25 percent. For p = 1/4 it is 1/5, which is 20 percent. The rule is simple to carry into the exam hall: if you honestly believe you have at least a one in four chance of being right, the attempt does not lose marks on average under the APPSC penalty.

Now add elimination. If you can remove options with confidence, you are choosing among fewer. For n remaining options, the expected value is (1 minus (n minus 1) x p) divided by n. The table gives the values for a one-mark question.

Options left after elimination	Chance of being right	Expected marks, penalty 1/3	Expected marks, penalty 1/4
4 (blind guess)	25.0 percent	0.000	+0.063
3	33.3 percent	+0.111	+0.167
2	50.0 percent	+0.333	+0.375
1 (certain)	100 percent	+1.000	+1.000

The table shows where the real gain from elimination lies. Removing one option out of four turns a neutral guess into a gain of about 0.11 marks. Removing two turns it into a gain of one third of a mark. Over thirty questions where you can remove one option, the expected gain is about 3.3 marks, and where you can remove two it is 10 marks. That is why a revision habit of asking "which two options can I prove wrong" pays more than a habit of memorising a few extra facts.

A warning about the word "average". A single guess is not an average. Under a one-third penalty, a blind guess gives either plus 1 with probability 0.25 or minus one third with probability 0.75. The variance of one such guess is 0.25 x 1 plus 0.75 x (1/9), which is 0.333, so the standard deviation is about 0.58 marks. If you guess forty questions blindly, the expected result is zero, but the spread is the square root of 40 x 0.333, about 3.65 marks. You could finish three or four marks above zero or three or four marks below it purely on luck. In a competition where the gap between two candidates is often a mark or two, that spread is not trivial, and it is a reason to prefer eliminated guesses to blind ones.

Notice also that the arithmetic changes if the question's marks are not one. For a half-mark question, every figure in the table is halved. The break-even probability does not change, because both the reward and the penalty scale together.

Attempt rate and accuracy under a one-third penalty

Your score under the APPSC rule is the number correct minus one third of the number wrong. If you attempt a questions with accuracy r, the score is a x r minus a x (1 minus r) / 3, which simplifies to a x (4r minus 1) / 3. The table applies it to a 150-question, 150-mark paper such as the objective papers of the Assistant Statistical Officer notification.

Questions attempted	Accuracy 90 percent	Accuracy 80 percent	Accuracy 70 percent	Accuracy 60 percent
100	86.67	73.33	60.00	46.67
120	104.00	88.00	72.00	56.00
140	121.33	102.67	84.00	65.33
150	130.00	110.00	90.00	70.00

Read across a row and the penalty is visible: dropping from 90 to 60 percent accuracy at 120 attempts costs 48 marks. Read down a column and attempting more helps only if accuracy holds, and accuracy usually does not hold, because the extra questions are the harder ones. The second table follows one student with a realistic profile, in which accuracy falls as the attempt count rises. The figures are illustrative.

Attempts	Accuracy	Correct	Wrong	Net score
100	85 percent	85	15	80.00
120	80 percent	96	24	88.00
140	70 percent	98	42	84.00
150	60 percent	90	60	70.00

For this student, the best attempt count is about 120. Going from 100 to 120 adds 11 correct and 9 wrong answers, a net gain of 8 marks. Going from 120 to 140 adds only 2 correct and 18 wrong answers, a net loss of 4 marks. The marginal question is the one that matters: the last twenty questions you attempt are right 10 percent of the time, which is well below the 25 percent break-even, so they cost marks. A rule that follows from this is to find your own marginal accuracy in mock tests. Look at the last block of questions you attempted in each mock, compute how many were right, and stop attempting at the point where that block drops below one in four.

Another way to see the same thing is to ask what a block of twenty extra questions is worth at different marginal accuracies. At 50 percent accuracy the block adds 6.67 marks. At 40 percent it adds 4.00. At 30 percent it adds 1.33. At 25 percent it adds nothing. Below that it subtracts. This is a table you can reproduce on the back of a rough sheet in the exam, and it explains why advice such as "attempt everything" and "attempt only what you know" are both wrong in general: the right answer depends on your marginal accuracy and on the penalty.

A three-tier rule for deciding what to attempt

A mock test gives you the data to build a personal rule. A workable one has three tiers.

• Tier 1: certain. You can state the reason the answer is right and the reason the other options are wrong. Mark it and move on. These should make up most of your attempts.
• Tier 2: narrowed. You are sure of at least one wrong option, and you have a reason to prefer one remaining option. Under a one-third penalty, a two-option or three-option question is worth attempting. Mark it in the question paper so that you can return to it.
• Tier 3: blind. You cannot remove any option, or the question is unfamiliar. Under a one-third penalty, skip it on the first pass. If time remains at the end and you have no better use for it, guessing is neutral on average, and you can decide on a tolerance for variance. If the notification says there is no penalty, or if it does not mention one and you decide to accept the risk, the position reverses and you should answer every Tier 3 question before time runs out.

Elimination is a skill, and it works differently by subject. In history and polity, look for the options that mix periods or institutions, such as a Satavahana fact attached to a Kakatiya ruler. In geography and economy, absolute words such as "all", "never" and "only" are often wrong. In arithmetic and reasoning, the options often differ in the last digit or the units, and a quick estimate removes two of them. In statement-type questions, one statement you know is false removes every option that includes it.

Time arithmetic for the papers

Time per question is the simplest number to compute, and the most useful. The table divides the printed duration by the printed number of questions. It uses the figures in the documents listed above.

Paper	Questions	Duration	Seconds per question
APPSC Group-I Screening, Paper I or II (Notification 07/2026)	120	120 minutes	60
APPSC Assistant Statistical Officer, Paper I or II (20/2026)	150	150 minutes	60
APPSC Forest Beat Officer, Paper I or II (18/2026)	100	100 minutes	60
SLPRB SI Preliminary, each paper (16.09.2026)	100	3 hours	108
SLPRB SI Final, Paper III or IV (16.09.2026)	200	3 hours	54
SLPRB Constable, preliminary or final (16.09.2026)	200	3 hours	54
Mega DSC 2025, SA, SGT and TGT (Special Education)	160	2 hours 30 minutes	56
Mega DSC 2025, PET and SA (Physical Education)	200	3 hours	54

Two patterns stand out. The APPSC objective papers give exactly one minute per question, which makes timing easy to track, and leaves no slack at all, so every minute spent on a Tier 3 question must be recovered elsewhere. The police and DSC papers give less than a minute per question, and the SI preliminary papers, with 100 questions in three hours each, give nearly two minutes per question. One of them is an arithmetic and reasoning paper, where questions take time. The same discipline cannot serve both. In the first kind, speed on factual questions is the lever. In the second, accuracy and method are.

A section-wise time-split model for a 150-question paper

The model below is for a 150-question, 150-minute paper in the style of the Assistant Statistical Officer notification, whose syllabus covers current events, science, history, geography, polity, economy and mental ability. The notification does not fix the number of questions in each subject, so the split is an assumption for planning, and you should adjust it to the pattern of the mock or previous papers you use. The model is not an official distribution.

Block	Questions	Minutes	Seconds per question
Current affairs (international, national, Andhra Pradesh)	30	24	48
History (India and Andhra Pradesh)	20	18	54
Geography (India and Andhra Pradesh)	15	13	52
Polity and governance	15	12	48
Economy and planning	15	13	52
Science and technology	20	17	51
Reasoning, mental ability and numeracy	35	42	72
Reserve for review and marking check	0	11	not applicable
Total	150	150	60 average

The logic is simple. Factual blocks are quick when you know the answer and impossible to improve by thinking when you do not, so they get about 50 seconds per question. Reasoning and numeracy repay careful work, so they get about 70. The reserve of 11 minutes is not a luxury. It covers the return to Tier 2 questions, a check that the answers you marked match the question numbers, and an unexpected slow patch. A candidate who spends the reserve early has no margin for a sudden hard block.

Run the paper in two passes. In the first pass, attempt Tier 1 and Tier 2 questions, marking the Tier 2 ones. Skip anything blind. Keep to the block minutes in the table, and if a block overruns, move on and make up the time in the reserve rather than in the next block. In the second pass, return to Tier 2 and decide on blind questions. Before the final minute, check the sheet.

For the APPSC Group-I screening test, the paper has an official structure: Paper I of 120 questions has four parts of 30 marks each, and Paper II has Part A of 60 marks and Part B of two sections of 30 marks each, per Annexure II of the notification. That gives a ready-made split of 30 minutes per 30-question part in Paper I, and for Paper II, 60 minutes for the 60-question Part A with 30 minutes each for the two 30-question sections. If your strongest part is History and Culture, you may finish it in 22 minutes and bank eight minutes for the part you find hardest, which is the reason to know your own speeds from mock tests.

For Mega DSC 2025 School Assistant papers, the bulletin lists 20 questions in General Knowledge and Current Affairs, 10 in Perspectives in Education, 10 in Classroom Implications of Educational Psychology, 80 in the subject content and 40 in methodology, which is 160 in all. A planning split for the 150 minutes is 15, 8, 8, 70 and 35 minutes, which is 136, with 14 minutes held in reserve. For SGT, the 160 questions are 16 in General Knowledge and Current Affairs, 8 in Perspectives, 16 in Psychology, and 24 each in the five language and subject blocks, where each block is 16 content and 8 methodology questions. A planning split is about 12, 6 and 12 minutes for the first three and 22 minutes for each of the five blocks, which is 140 minutes, with 10 held back. As in the APPSC model, these splits are planning aids, and the bulletin does not prescribe them.

For the SLPRB papers, the Constable paper of 200 questions in three hours is the tightest, at 54 seconds per question. The notification does not fix how many questions come from each subject, so build the split from the mock you use: give the quick factual blocks about 40 seconds a question, and bank the saved minutes for arithmetic and reasoning questions, which may need more than a minute. With no negative marking mentioned, the main trade-off is between speed and accuracy, not between attempting and skipping, but see the point above about practising with both scores.

Discipline on the OMR sheet and the screen

Several rules in the notifications bear on time and marks, and they are easy to overlook.

For OMR papers, the SLPRB notifications say candidates must answer on an OMR sheet using a blue or black ball point pen only. The APPSC Group-I notification says that if a candidate fails to bubble, or wrongly bubbles, the booklet series on the OMR sheet, the sheet is invalidated, because it is read by an optical mark reader, and that tampering with the sheet by whitener, eraser, blade or chalk powder leads to invalidation. No request for reconsideration of such cases is entertained. Bubbling the booklet series is therefore the first act in the hall, and should be checked twice. A plan for transferring answers is also part of time management. Marking the answer in the booklet and transferring in batches of ten or twenty saves switching time but leaves a risk at the end. Marking directly saves the end rush but costs a few seconds per question. Try both in mock tests and settle on one.

For computer-based tests, the Mega DSC 2025 press note on the examinations says the CBT records the option the candidate selects, and it addresses complaints from candidates who did not save an answer after selecting it, or did not save again after changing it. The practical lesson is to learn the on-screen routine on a practice test before the exam: select, save, and confirm the question palette status. It is a few seconds per question and avoids a loss that has nothing to do with knowledge.

If questions are deleted after the key is released, APPSC Group-I notification 07/2026 says that scaling, a proportionate increase, is applied to that part of the paper up to the maximum marks for the paper, rounded to two decimals. For example (illustrative), if a part is worth 30 marks and two of its 30 questions are deleted, a candidate who scored 18 on the remaining 28 would be scaled to 18 x 30 / 28, which is 19.29. This is one reason the final key, and not the initial key, is the document to use for your own estimate.

Why fixed-paper mocks matter

A fixed-paper mock is one set of questions that every candidate attempts, as opposed to a test generated at random from a bank. The property matters for three reasons.

Comparison. When thousands of people attempt the same paper, the distribution of their scores is a measuring instrument. Your score on its own tells you little, because a paper can be easy or hard. Your score against the distribution of the same paper tells you where you stand, and that is the idea behind the percentile discussed below.

Repeatability. Because the paper does not change, you can attempt it, analyse it, and return after a fortnight to see what you retained. The second attempt measures learning, not luck. A mock that regenerates itself cannot give you that.

Trend. A series of fixed papers of similar difficulty lets you plot net score, attempt count and accuracy across weeks. A real improvement shows as a rising net score with stable or rising accuracy. A false improvement shows as rising attempts with falling accuracy, which the tables above show can lower the net score.

There are two cautions. First, a fixed paper is not a prediction of the real paper, so do not memorise it, and do not read a mock percentile as a rank in the real exam. The group of people who take mock tests is not the group that takes the exam, so the comparison is indicative only. Second, because the questions are fixed, a second attempt inflates your score by memory, so record the first attempt as your baseline and treat the second as a retention check.

A mock-test analysis routine

A mock is worth only as much as the analysis that follows it. Plan two to three times the length of the test for analysis in the early weeks. The routine has five steps.

1. Record the headline numbers. Attempts, correct, wrong, unattempted, net score under a one-third penalty, raw score without penalty, and time used. Add the number of negative marks lost, which is one third of the wrong answers.
2. Go through every wrong and every unattempted question. Do not review only the wrong ones, because an unattempted question that you could have answered is a hidden loss.
3. Tag each with one cause. Use the categories in the table below, and use only one tag per question, so that the counts add up.
4. Write the corrected fact or method in one line in a notebook arranged by subject, not by test. This notebook is your revision material for the last fortnight.
5. Choose one change for the next mock. One change, such as "no blind guesses in history" or "leave the reasoning block to the end", is testable. Five changes are not.

Error category	What happened	Typical fix
Concept gap	You did not know the topic	Study the topic from the source, then add practice questions on it
Recall failure	You studied it and could not retrieve it	Spaced repetition: revisit after one day, one week and one month
Careless or misread	You knew the answer and marked the wrong option, or misread the question	Underline the key word in each stem, such as "not", "except", "first"
Calculation slip	Method right, arithmetic wrong	Practise written methods and shortcuts, check units and the last digit
Blind guess	No option removed and the answer was wrong	Apply the Tier 3 rule: skip under a penalty
Trap or near-miss	You narrowed to two and chose the wrong one	Note the trap type, such as date confusion, and review the distractor
Time pressure	Rushed or left unattempted because time ran out	Apply the block minutes and the reserve, and cut slow blocks
Language or translation	Misunderstood a term in the second language of the paper	Learn the English and Telugu terms together for the subject

An illustrative example shows how the numbers work. Suppose a student attempts a 150-question paper with a one-third penalty and gets this outcome (the numbers are invented): 126 attempted, 94 correct, 32 wrong, 24 unattempted. The net score is 94 minus 32/3, which is 83.33. Accuracy is 94/126, or 74.6 percent. The wrong answers are tagged as 10 concept gaps, 8 recall failures, 7 careless errors, 5 blind guesses and 2 calculation slips. Of the 24 unattempted questions, 14 were left because time ran out and 10 were skipped on purpose.

The analysis shows what to do. Each careless error converted to a correct answer is worth 1 plus 1/3, that is 1.33 marks, so fixing the 7 careless errors is worth 9.33 marks. Skipping the 5 blind guesses is worth 1.67 marks. Together those are 11 marks without learning a single new fact. Fourteen questions lost to time suggest that the reasoning block is taking too long, which is a time-split problem. Only the 10 concept gaps and 8 recall failures need study. A student who looked only at the score of 83.33 would have studied everything. A student who read the error log would study a fifth of it and spend the rest of the time on method.

Repeat the routine for four or five mocks and look at the error counts across tests. If careless errors fall from 7 to 3 and attempts rise, the habits are working. If concept gaps stay constant in one subject, that subject needs a different kind of study.

Percentile: what it is and how it is computed

A percentile is a statement about position in a group. The percentile rank of a score tells you the share of the group that scored below you. It is not a percentage of marks. A candidate with 62 percent marks may be at the 90th percentile of a hard paper and at the 40th percentile of an easy one.

There are two common conventions, and you should know which one a report uses.

• Strictly-below plus half the ties: percentile = 100 x (number scoring below you + half the number scoring equal to you) / N, where N is the number of candidates.
• At-or-below: percentile = 100 x (number scoring at or below you) / N.

The two differ only when there are ties, and for large groups they are nearly the same. Here is an illustrative example, with invented scores for twenty candidates on a paper, sorted from lowest to highest: 38, 45, 52, 55, 58, 61, 63, 66, 68, 70, 72, 74, 75, 78, 80, 83, 85, 88, 91, 96. This is not data from any real exam or any real cohort. The mean is 69.9 and the standard deviation is about 15.1.

A candidate with 74 has 11 scores below, and is one of 20 with one score equal. Using the first convention, the percentile is 100 x (11 + 0.5) / 20, which is 57.5. Using the second, it is 100 x 12 / 20, which is 60. A candidate with 91 has 18 scores below, so the first convention gives 100 x 18.5 / 20, which is 92.5, and the second gives 100 x 19 / 20, which is 95. The top score of 96 is at the 97.5th percentile by the first convention and the 100th by the second. In a group of twenty, one candidate is a large share of the distribution, and the two conventions disagree by several points. In a group of fifty thousand, they would agree to a small fraction of a point.

To get from percentile to an approximate rank, reverse the arithmetic: rank is about 1 + N x (100 minus P) / 100. In an invented group of 50,000, a percentile of 96 corresponds to a rank of about 2,001. This is why a percentile of 96 can be very good or not enough, depending on the number of posts. Percentile alone does not say whether you are inside the vacancies. The number of vacancies relative to the number of candidates does.

Three more facts keep percentiles honest. Percentile depends on the group, so a mock percentile compares you with other people who took the mock, who may be more or less prepared than the real candidates. Percentile is not additive: you cannot average percentiles across tests as you average marks. And percentile hides the gap between scores: in the example, the gap between 85 and 88 is only three marks but crosses a percentile step of 5, while the gap between 38 and 45 is seven marks and crosses the same step.

As on October 2026, the notifications opened for this guide (APPSC Group-I 07/2026, the SLPRB SI and Constable notifications, and the Mega DSC 2025 notification and bulletin) select candidates on marks, and the word percentile does not appear as a selection measure. The Mega DSC 2025 merit list prints a total score and a rank. Percentile is therefore a practice tool for you, and not a figure that the recruiters publish. The companion guide on percentile, cut-off and normalisation explains what the official lists do print, and how to read them.

A weekly rhythm for mocks

The plan below is a suggestion for the last twelve weeks before an exam, not an official schedule. Adjust it to your preparation. The three phases correspond to what the data shows you need at each stage.

Phase	Weeks	Test type	Aim
Foundation	1 to 4	Untimed sectional tests, one subject at a time	Accuracy, and building the error notebook
Speed	5 to 8	Timed sectional tests, then one full paper a week	Block minutes, elimination, finding your marginal accuracy
Full simulation	9 to 12	Two full-length papers a week in exam conditions, with one rest day	Stamina, the two-pass method, final attempt count

In the speed phase, stop looking at the clock only at the end. Note the time at the end of each block and compare it to the model. In the simulation phase, sit the paper at the hour of the real exam where the timetable is known, with the same stationery and the same rules, and take the analysis seriously the next day. Keep one rest day a week, because accuracy falls with fatigue, and a tired mock gives misleading data.

Applying this to APPSC, DSC and police papers

APPSC objective papers. Take the one-third penalty as given. Practise with it in every mock. Build your attempt count from your marginal accuracy. For Group-I screening, remember that the screening test is for shortlisting: the Group-I notification says the number of candidates admitted to the mains is 50 times the number of vacancies, so the screening score matters only as far as it places you within that group. The marks of the mains and interview decide selection. A cautious, high-accuracy approach that secures a place in the shortlist is better than a risky one that gambles for a high score.

Police papers. The notifications as read do not mention negative marking, and the preliminary written test is qualifying: marks obtained in it have no relevance in the final selection, according to both the SI and Constable notifications. The minimum is 40 percent for open and EWS candidates, 35 percent for backward classes and 30 percent for SC and ST. For a 200-mark Constable preliminary, that means 80, 70 and 60 marks. In the preliminary stage, therefore, the goal is to cross the line safely, and for the final written paper, to maximise marks, as final selection is on relative merit of the final written score for the Constable civil post, or on the aggregate for the SI posts. Practise both scores, and read the exam-day instructions.

DSC papers. The 2025 bulletin did not mention a penalty. Each question in the SA and SGT papers is worth half a mark, and the CBT total of 80 marks is combined with 20 marks for APTET or CTET weightage in the final merit, subject to normalisation across shifts. Because the final score is on a normalised scale, your relative position in your own shift is what the formula works from, and a slow shift or an easy shift does not change the plan: maximise your own raw score. Read the DSC exam pattern guide for the part-wise structure and the 90-day plan for study scheduling.

Common mistakes

• Treating "negative marking" as one number for all exams. Check the notification of your post.
• Attempting every question by default, then attempting nothing by default, instead of finding your marginal accuracy.
• Using blind guesses under a penalty and calling it luck.
• Reviewing only wrong answers, and ignoring the unattempted ones.
• Reading a mock percentile as a prediction of rank.
• Memorising a fixed paper and reading the second attempt as progress.
• Forgetting the OMR rules: the booklet series, the pen, and the ban on whitener.
• Leaving no reserve minutes at the end of the paper.

Key facts for exams

• APPSC objective-type papers: one third of the marks deducted for each wrong answer, G.O.Ms.No.235, Finance (HR-I, Plg and Policy) Department, 6 December 2016, as printed in notifications of September and October 2026.
• Break-even accuracy for guessing: 25 percent at a one-third penalty, 20 percent at a one-quarter penalty.
• A blind four-option guess has expected value zero at a one-third penalty.
• APPSC Group-I screening test, as on October 2026: two papers, each 120 questions, 120 minutes and 120 marks; mains candidates are 50 times the vacancies.
• SLPRB preliminary written tests are qualifying only; minimum 40 percent OC and EWS, 35 percent BC, 30 percent SC and ST.
• SLPRB notifications of 16.09.2026 do not mention negative marking and award zero for an unmarked question.
• Mega DSC 2025 SA and SGT papers: 160 MCQs, 80 marks, 2 hours 30 minutes; the bulletin does not mention negative marking.
• APTET June 2026: no negative marking.
• Percentile is the share of the group scoring below you, and is not a percentage of marks.

Ten practice questions

1. Under a one-third penalty, the expected value of a blind guess among four options on a one-mark question is: (a) minus 0.25 (b) zero (c) plus 0.25 (d) plus 0.33 Answer: (b). One quarter of 1 equals three quarters of one third.

2. The break-even probability of being right under a one-third penalty is: (a) 20 percent (b) 25 percent (c) 33 percent (d) 50 percent Answer: (b), from p / (1 + p) with p = 1/3.

3. A candidate attempts 120 questions in a 150-mark paper, 96 correctly and 24 wrongly, with a one-third penalty. The net score is: (a) 96 (b) 90 (c) 88 (d) 72 Answer: (c). 96 minus 8.

4. Under a one-third penalty, if you can eliminate one option of four with certainty, the expected value of guessing among the remaining three is: (a) 0.000 (b) 0.111 (c) 0.333 (d) 0.500 Answer: (b), which is 1/9.

5. An APPSC objective paper has 150 questions in 150 minutes. The average time per question is: (a) 54 seconds (b) 56 seconds (c) 60 seconds (d) 108 seconds Answer: (c).

6. In the invented group of twenty scores in this guide, how many scores lie strictly below 74? (a) 10 (b) 11 (c) 12 (d) 13 Answer: (b).

7. A candidate's percentile is 90. This means: (a) the candidate got 90 percent marks (b) about 90 percent of the group scored below the candidate (c) the candidate answered 90 questions (d) the candidate is ranked 90th Answer: (b).

8. As on October 2026, which statement about the SLPRB Constable notification of 16.09.2026 is accurate? (a) It prescribes a one-third penalty (b) It prescribes a one-quarter penalty (c) It does not mention negative marking and awards zero for an unmarked question (d) It prescribes no marks for correct answers Answer: (c).

9. A question is worth half a mark and the penalty is one third of its marks. The penalty for a wrong answer is: (a) 0.083 (b) 0.167 (c) 0.333 (d) 0.500 Answer: (b), one third of 0.5.

10. In a mock analysis, a wrong answer caused by marking the wrong option although you knew the answer belongs to which category? (a) Concept gap (b) Recall failure (c) Careless or misread (d) Blind guess Answer: (c).

Frequently asked questions

Is there negative marking in all APPSC exams? In the objective papers of the notifications opened for this guide, yes: one third of the marks for each wrong answer, citing G.O.Ms.No.235 of 6 December 2016. Descriptive papers, such as the Group-I mains, are marked differently. Confirm in the notification of your post.

Is there negative marking in AP Police recruitment? The SLPRB Sub-Inspector and Constable notifications of 16.09.2026 do not mention it. They say that a question with no bubble darkened gets zero. Check the exam-day instructions and do not assume.

Is there negative marking in AP DSC? The Mega DSC 2025 bulletin does not mention it. The APTET June 2026 bulletin says there is none. For a DSC 2026 recruitment, read the new bulletin.

Should I attempt every question? If a penalty applies, attempt a question when your honest chance of being right is above the break-even, which is 25 percent for a one-third penalty. If no penalty applies, attempt every question.

How many mock tests should I take? There is no official number. A common pattern is to start with sectional tests, move to timed papers, and finish with two full-length papers a week, with at least as much time on analysis as on the test.

What is a good percentile in a mock? A percentile shows your position among the mock takers and does not predict your position in the exam. Track your net score, accuracy and attempt count along with it.

Does the official result use percentile? In the documents opened for this guide, no. Selection is by marks and rank, with qualifying marks and, for multi-shift DSC tests, normalisation. The companion guide explains each.

How do I use fixed-paper mocks? Attempt once under exam conditions, analyse the errors, study the gaps, and attempt again after about two weeks as a retention check, keeping the first score as the baseline.

Related guides

• Percentile, cut-off and normalisation explained for AP recruitment exams
• AP DSC exam pattern and marks
• AP DSC 90-day preparation plan
• APPSC Group-II prelims 90-day plan
• APPSC Group-I guide
• AP Police SI guide
• AP Police Constable guide
• AP Police recruitment 2026 update
• AP reservation, roster and local status
• AP TET guide

Practise on pareeksha.in

To apply these ideas, take a timed paper and keep both scores, with and without a penalty. Pareeksha.in offers 10 Lakh+ MCQs, 100+ Exams and 5,000+ Full-Length Mocks, and the mock papers are fixed, so you can compare your result with the distribution of the same paper. Start with the free mini mock, move to full-length papers in the online test series, and use practice sets for daily topic work. These tools measure preparation and do not promise any result.

Sources and verification

Official documents opened for this guide, all as on October 2026:

• APPSC Notification No.07/2026 for Group-I Services (psc.ap.gov.in), for the one-third rule, the screening test scheme, the 50-times shortlisting rule, minimum qualifying marks, OMR invalidation rules and the scaling rule for deleted questions.
• APPSC Notification No.18/2026 (Forest Beat Officer), No.20/2026 (Assistant Statistical Officer), No.23/2026 (Executive Officer Grade-III) and No.13/2026 (Supervisor Grade-I), for the printed negative-marking sentence and the paper patterns.
• SLPRB Notification Rc.No.81/SLPRB/Rect.1/2026 (Sub-Inspector) and Rc.No.91/SLPRB/Rect.2/2026 (Constable and other posts), both dated 16.09.2026 (slprb.ap.gov.in).
• AP Mega DSC 2025 Notification 01 and Information Bulletin (School Education), and the Mega DSC 2025 press note on examination conduct and normalisation (tet2dsc.apcfss.in).
• APTET June 2026 Information Bulletin (tet2dsc.apcfss.in), for the statement that there is no negative marking.

Not verified: the Government Order text of G.O.Ms.No.235 itself, which was read only as printed in the notifications; the 2026 DSC notification and bulletin, which were not available; the number of questions per subject in any APPSC paper, which the notifications do not fix; and the exam-day instruction sheets of the police recruitment. The time-split models, attempt tables, error categories and the 12-week plan are study aids written for this guide and are not official. The example scores and percentile data are invented for illustration.