The Written Test as a Scoring Problem: Pattern, Marking and Time
This chapter sets out the scoring and time arithmetic that every later chapter in this kit depends on. It covers the pattern of the written Common Entrance Examination (CEE) for Soldier General Duty, which recruitment notices call Agniveer General Duty; the three marking readings that sources report; what a guess is worth; and how much time each item can take. Every figure is labelled as recorded in the official table or as reported by a named source, so you can see how firm each number is.
KarmSakha preparation topics covered
- The Soldier General Duty pattern table and the two readings that circulate
- Marks per item and the three marking readings for a wrong answer
- The arithmetic of a blind guess and the break-even confidence
- Time per item: sixty minutes, fifty items and a review reserve
- The three-pass routine for a timed sitting
- How the written score sits inside the General Duty merit total
- The notice numbers to confirm before any practice is scored
Two pattern readings and why the sitting works under both
The spine of this chapter is the Soldier General Duty table in the Indian Army's CEE syllabus. As recorded in that table, it gives General Knowledge 15 questions for 30 marks, General Science 20 for 40 and Mathematics 15 for 30: 50 questions and 100 marks, with every question worth 2 marks. Every figure from the table is labelled as recorded; none has been confirmed against the live notice for this cycle, which decides. A third-party copy gives subject weights of 30, 40 and 30 per cent, which match the marks and serve only as corroboration.
The second reading is the one that several guides report. Physics Wallah's syllabus page gives General Knowledge 15 questions for 30 marks, General Science 15 for 30 and Mathematics 15 for 30, plus a separate logical reasoning block of 5 questions worth 10 marks. An SSB Crack Exams table dated 2024 shows the same 15, 15, 15 and 5 split. Examsuchana reports 45 questions and 90 marks with no reasoning block. Its total matches neither reading, so it is not used.
The sitting in chapter 6 works under both readings. It uses 15 General Knowledge items, 15 science items, 5 numeral and logical ability items and 15 Mathematics items: 50 items and 100 marks. Under the official reading, as recorded, the table places a numeral and ability heading inside General Science, which carries 20 questions, so the five ability items sit within that subject. Under the variant reading, General Science has 15 questions and the five logical items form their own block worth 10 marks. Either way the item count, the mark total and the time are the same. The kit claims no separate reasoning score unless a live notice lists one.
The Recruiting Year 2025-26 notice of Army Recruiting Office Charkhi Dadri is the newest official notice this chapter cites. A later notice for the 2026-27 intake, which one February 2026 news report labels recruiting year 2027, is reported by secondary sources to have opened registration on 13 February 2026 and to have held the June 2026 written test. That later notice is not cited here, so the figures in this chapter are the 2025-26 pattern as recorded. Recompute the arithmetic from the figures in the notice you hold.
| Subject | Official table, as recorded | Variant reported by secondary guides | This kit's sitting |
|---|---|---|---|
| General Knowledge | 15 questions, 30 marks | 15 questions, 30 marks | 15 items |
| General Science | 20 questions, 40 marks | 15 questions, 30 marks | 15 items |
| Numeral and logical ability | Inside General Science | 5 questions, 10 marks (separate block) | 5 items |
| Mathematics | 15 questions, 30 marks | 15 questions, 30 marks | 15 items |
| Total | 50 questions, 100 marks | 50 questions, 100 marks | 50 items, 100 marks |
Marks per item and what each answer type does under three marking readings
Each item is worth 2 marks. Under the official table as recorded, the subject totals are 30, 40 and 30 marks, which at 2 marks an item means 15, 20 and 15 items. The table as recorded states no deduction for a wrong answer, a point not yet re-verified against the notice. On that reading a correct answer earns 2 marks, a wrong answer earns nothing and a skipped question scores nil. Every reading shares one shape: 2 marks for each correct answer, minus the deduction for each wrong answer, with skipped questions contributing nothing. Only the deduction changes from one reading to the next.
Physics Wallah's syllabus page states 0.5 marks for each wrong answer, and KollegeApply's exam-pattern page and an Examsuchana page dated 31 March 2026 give the same figure. A June 2025 Business Standard report gives 0.25 marks, so the 0.25 reading rests on that single news report. Other posts and trades may be marked differently, so the General Duty arithmetic here must not be carried across without checking the notice for that post. The official table as recorded shows no deduction, but that line has not been re-verified, while every secondary page that states a penalty and could be read gives 0.5 marks. This chapter does not choose between the readings. Every calculation here, and every scoring table in chapter 6, is carried through all three.
Read deduction wording with care, because a penalty can be worded as a share of the question's marks. On a 2-mark item, 25 per cent of the marks is 0.5 marks, which belongs to the 0.5 reading, not the 0.25 reading. Under the 0.5 reading, one wrong answer cancels a quarter of a correct answer; under the 0.25 reading, it cancels one-eighth. Reading the penalty as 0.25 halves it, and on a sitting with nine wrong answers that overstates the score by 2.25 marks. Convert any percentage penalty into marks before you score a single practice set.
The table below scores one fixed attempt under each reading. The same 36 correct answers are worth 72 marks throughout, so the spread between readings comes entirely from the 9 wrong answers. The lowest total, 67.5, is 4.5 marks below the top score. A candidate who has not confirmed the deduction cannot tell whether a practice score is high or low, so the notice comes before any scoring. Whichever reading applies, the correct answers are the part of the score a reader controls directly, while the penalty depends on how carefully guesses were made.
| Reading | Deduction per wrong answer | Total deducted | Score: 36 correct, 9 wrong, 5 skipped |
|---|---|---|---|
| No deduction (official table as recorded; not re-verified) | 0 | 0 | 72 marks |
| 0.5 per wrong answer (Physics Wallah, KollegeApply, Examsuchana) | 0.5 marks | 4.5 marks | 67.5 marks |
| 0.25 per wrong answer (Business Standard, June 2025; single report) | 0.25 marks | 2.25 marks | 69.75 marks |
The guessing arithmetic and the break-even confidence
A guess is an answer given without knowing the right option. On a four-option item a blind guess is right with chance 1 in 4, earning 2 marks, and wrong with chance 3 in 4, costing the deduction. Its expected value is 2 × 1/4 minus the deduction × 3/4. With no deduction a blind guess is worth 0.5 marks on average; with the 0.5 deduction, 0.125 marks; with the 0.25 deduction, 0.3125 marks. The expected value stays positive in all three readings, but the margin shrinks as the penalty rises.
The break-even confidence is the chance of being right at which a guess is worth exactly nothing. Call that chance p and the deduction d. The guess gains 2 marks with chance p and loses d marks with chance 1 − p. Setting the expected value to zero gives p = d ÷ (2 + d). For d = 0.5 this is 0.5 ÷ 2.5 = 0.2, or 20 per cent. For d = 0.25 it is 0.25 ÷ 2.25, or about 11.1 per cent. With no deduction, any chance is worth taking. Answer when your chance is above the break-even for the reading your notice gives.
Elimination lifts a chance above break-even. If you rule out one of four options, a pick among the remaining three is right with chance 1 in 3, about 33 per cent, which clears both break-even figures. If you rule out two, the chance becomes 1 in 2, and under the 0.5 reading the expected value rises to 0.75 marks. The arithmetic assumes the options left are equally likely to be right. Your real chance may be higher or lower, and the break-even shows only where the line falls.
Consider a reader who believes one option has a 15 per cent chance of being right. Under the 0.5 reading, 15 per cent is below the 20 per cent break-even, so the guess loses marks on average and should be skipped. Under the 0.25 reading, the same 15 per cent clears 11.1 per cent, so the guess is worth making. The same judgement on the same item can be right under one reading and wrong under another. Record the reading beside every guessing decision in your error log.
| Deduction per wrong answer | Expected value of a blind four-option guess | Break-even chance of being right |
|---|---|---|
| None (official table as recorded; not re-verified) | 0.5 marks | 0 per cent |
| 0.5 marks | 0.125 marks | 20 per cent |
| 0.25 marks | 0.3125 marks | 11.1 per cent |
Time budget: sixty minutes, fifty items and a three-pass routine
Two secondary pattern pages checked for this edition give sixty minutes for the General Duty paper: Examsuchana, and KollegeApply, whose post-wise table lists 120 minutes for the Clerk paper, so the figure varies by post. Aglasem's page reports the same sixty minutes but was not re-checked for this edition, and Testbook's one-to-two-hour figure is not confirmed, so neither is relied on. This kit uses sixty minutes as its working protocol, and the notice for your cycle decides the real figure. Sixty minutes is 3,600 seconds, which spread over 50 items gives seventy-two seconds per item if every second goes to answering. A final check also needs time, and the next paragraph sets that aside.
A five-minute review reserve leaves fifty-five minutes, or 3,300 seconds, of working time for both passes: sixty-six seconds per item on average. Treat that as an average, not a limit, because some items settle in ten seconds and others need much longer. The reserve is held back for the final check, and the second pass uses working time for flagged items. If the pace drifts past the average, the reserve can absorb only a little of it, so the checkpoints in chapter 6 matter more than any single item's timing.
The routine has three passes. In the first, answer every item you can settle quickly and flag the rest, rather than guessing at once. In the second, return to the flagged items in their original order, eliminate what you can, and guess only where the chance clears break-even under your notice's reading. In the third, use the reserve to check changed answers and confirm that every answer has been saved. Sure answers bank marks before time runs short, and guesses are made last, when the remaining time is known.
