Negative Marking Strategy: When to Attempt and When to Skip a Question

The Math Behind a Guess

An uncertain question, a ticking clock, four options, and a decision to make in a few seconds: attempt it, or leave it blank. The instinct most candidates fall back on in that moment is how confident the question feels, which is understandable but not especially reliable, since confidence and accuracy don't track each other closely under time pressure. Negative marking is actually a straightforward expected-value calculation once you know the scoring scheme, and working through that math once, in advance, replaces the in-the-moment feeling with a clear rule to apply on exam day.

Many exams use some version of a scheme where a correct answer earns a fixed positive mark and an incorrect answer costs a fraction of that mark, with no penalty for leaving a question unattempted — a common example is +1 for a correct answer and −0.25 for an incorrect one, though the exact scheme varies by exam, and you should always confirm the actual scheme for your specific exam from the official notification rather than assuming a figure from elsewhere applies to you. The logic below works the same way regardless of the exact numbers, as long as you plug in the real scheme for your exam.

The Core Calculation

With a +1 / −0.25 scheme, random guessing among four equally likely options is already slightly favourable, not neutral: on average, across many such guesses, you'd get one correct (earning +1) for every three incorrect (costing −0.25 each, or −0.75 total), netting +0.25 across four attempts — a small positive expected value per question, not zero. The true breakeven point for a +1 / −0.25 scheme is five options, not four: at five options, one correct (+1) offsets four incorrect (−0.25 × 4 = −1) exactly. With four options, a blind guess already tilts slightly in your favour before you eliminate anything, with real variance around that average on any individual question.

The calculation changes meaningfully once you can eliminate even one incorrect option. If you can confidently rule out one of four options, you're now choosing between three, so your odds of guessing correctly rise to one in three. Under the same +1 / −0.25 scheme, the expected value across many such guesses becomes positive: roughly one correct in three (earning +1) against two incorrect (costing −0.25 each, or −0.5 total), averaging out to a small but real positive expected value. Eliminate two of the four options, down to a straight guess between two, and the expected value becomes clearly positive — you'd expect, on average, to gain from attempting rather than skipping.

A Simple Table to Internalise

The general pattern, under a +1 / −0.25 scheme specifically: a pure guess among four options is already mildly positive in expected value, not neutral; a guess among three options (one eliminated) is more positive still; a guess among two options (two eliminated) is clearly positive; and a guess where you've narrowed it to one strong candidate with real uncertainty remaining is close to a near-certain correct answer and should essentially always be attempted. The practical rule that falls out of this: the key decision point isn't 'am I sure,' it's 'can I eliminate at least one option with genuine confidence' — if you can, the math favours attempting; if you genuinely cannot eliminate anything and are choosing blind among all the original options, the decision is close to neutral and reasonably comes down to your personal risk tolerance and how much time the question has already cost you.

It's worth being honest with yourself about the difference between 'I can eliminate an option' and 'I have a vague hunch about which one is right.' The expected-value math above assumes a genuine elimination — ruling an option out because you can identify a specific reason it's wrong, not because it merely feels less likely. Treating a hunch as a confident elimination when it isn't one will make the math look more favourable than your actual odds justify.

Where This Breaks Down: Time Cost

The expected-value calculation above only accounts for the mark gained or lost on that single question — it doesn't account for the time spent getting there, which is often the more binding constraint in a timed exam. A question that takes four minutes to narrow down to a confident two-option guess might have a positive expected mark value in isolation, but if those four minutes meant you didn't get to attempt three other, easier questions later in the section, the real cost is higher than the isolated calculation suggests. This is why the practical rule in a timed exam isn't purely 'calculate the expected value and attempt anything positive' — it's 'calculate the expected value, and weigh it against what else those same minutes could earn you elsewhere in the paper.'

A reasonable practical habit: set a rough time ceiling for how long you'll spend trying to eliminate options on a single difficult question — somewhere in the range of a minute or so for most sections, adjusted to your exam's specific time pressure — and if you haven't narrowed it down within that window, mark it to return to later if time allows, rather than letting one question consume time that easier, unattempted questions elsewhere could use more productively.

Exam-Day Rules of Thumb

Translating the math into something usable under real time pressure: first, complete a full pass through the section attempting every question you're genuinely confident about, skipping anything uncertain on the first pass entirely — this protects your easy and medium-confidence marks before time pressure forces rushed decisions on harder questions. Second, use remaining time for a second pass through skipped questions, applying the elimination-based rule above: attempt anything where you can confidently rule out at least one option, and be honest about whether a 'leaning' is a genuine elimination or just a guess with extra steps. Third, in the final minute or two, make a conscious decision about any remaining untouched questions rather than running out of time mid-decision — a last-second, half-considered guess with no time to actually think it through is the lowest-value use of your final seconds, even under a scheme where guessing is technically neutral in expectation.

A Worked Example

Suppose a reasoning question has four options, and you immediately recognise one option violates a constraint stated in the question — a genuine elimination, not a hunch. You're left choosing among three, with no further information distinguishing them. Under a +1 / −0.25 scheme, this is worth attempting: the expected value is mildly positive, and skipping forfeits that small positive edge entirely in exchange for a guaranteed zero. Compare this with a question where you have no real foothold at all — none of the four options can be ruled out with any specific reasoning — and you're genuinely torn between all four. Here the expected value is close to neutral, and whether to guess becomes much more a matter of remaining time and personal risk tolerance than a clear mathematical edge either way.

Why This Framework Matters Beyond the Exam Itself

Beyond the direct scoring benefit, practising this kind of elimination-based decision-making during preparation — rather than only during the real exam — builds a habit of looking for a reason to rule out an option before committing to a guess, which tends to improve accuracy generally, not just under negative marking specifically. You can build this habit directly through timed practice with ThinkZene's free practice quizzes, which don't require signing up before you start, and reviewing afterward which attempts were genuine eliminations versus hunches that happened to pay off.

Frequently Asked Questions

Does this math change if the negative marking scheme is different from +1 / −0.25?

Yes — the specific break-even points shift with the scheme, but the underlying logic doesn't. Work out the same expected-value calculation using your exam's actual scoring scheme: a correct answer's value against an incorrect answer's penalty, multiplied by the relevant probabilities, tells you the break-even point for your specific exam rather than assuming the commonly cited +1 / −0.25 example applies universally.

Is it ever correct to guess with zero information at all?

Under a scheme where blind guessing is close to expected-value neutral, like the common +1 / −0.25 example, it's close to a coin flip in terms of score impact either way — neither clearly smart nor clearly reckless. Many candidates reasonably choose to skip in this specific situation anyway, simply to protect the certainty of a flat zero rather than accept the variance of a roughly neutral gamble, which is a defensible risk-tolerance choice rather than a mathematically wrong one.

Does this strategy change in the last five minutes of an exam?

A little, yes — with very little time left, the opportunity cost of spending another minute hunting for an elimination on one question drops, since there's nothing else left to spend that minute on anyway. This tips the calculation slightly further toward attempting borderline questions in the closing minutes than it would earlier in the paper, when that same minute could have gone toward an easier, unattempted question elsewhere.

A Second Worked Example, With a Different Scheme

Not every exam uses +1 / −0.25. Suppose an exam instead uses +2 for a correct answer and −0.5 for an incorrect one among four options — a steeper penalty relative to the reward than the earlier example. Run the same logic: a pure guess among four options still nets out close to zero on average (one correct in four, averaging +0.5, against three incorrect averaging −0.375, for a small positive edge in this particular scheme — notice this isn't quite neutral the way +1/−0.25 was, which is exactly why checking your own exam's actual numbers matters more than memorising a single break-even rule). Eliminate one option and the arithmetic shifts further in the attempter's favour, the same direction as before, just at different numbers. The mechanism is identical across schemes; only the specific break-even point changes, which is the whole reason this guide works through the calculation rather than handing you a single number to memorise.