Every aspirant knows negative marking exists. Almost nobody does the arithmetic of when guessing helps and when it destroys a selection. This post does the maths for the three schemes you'll actually face - SSC's −0.5, Railways' −1/3, and MTS Session-II's −1 - and turns it into three simple rules you can apply mid-exam.
A blind guess among 4 options is right 25% of the time. Its long-run value per question is: (0.25 × marks for correct) − (0.75 × penalty). If that number is positive, guessing gains marks over many questions; if negative, it loses marks. The switch that changes everything is eliminating options - each option you rule out moves the odds from 1-in-4 to 1-in-3 to 1-in-2.
Blind guess: 0.25×2 − 0.75×0.5 = 0.5 − 0.375 = +0.125 per question. Mathematically, even blind guessing is marginally positive at SSC's ratio. Eliminate one option (1-in-3): 0.33×2 − 0.67×0.5 = +0.33. Eliminate two (50–50): 1 − 0.25 = +0.75 - three-quarters of a mark per attempt, which is why 50–50s should ALWAYS be attempted in SSC.
Blind guess: 0.25×1 − 0.75×0.33 = 0.25 − 0.25 = exactly zero. The scheme is engineered so blind guessing neither helps nor hurts on average - but "average" hides variance: 20 blind guesses can easily swing your score ±4 marks, which at Group D cutoff densities is thousands of ranks. Eliminate one option: 0.33 − 0.22 = +0.11; eliminate two: 0.5 − 0.17 = +0.33. Rule: in RRB, guess only after eliminating at least one option.
Blind guess: 0.75 − 0.75 = zero, same engineering as Railways. One elimination: 1.0 − 0.67 = +0.33; two eliminations: 1.5 − 0.5 = +1.0 per attempt - a full mark, enormous at MTS's 3-mark weighting. Same rule as RRB: never blind, always post-elimination.
Rule 1 - SSC (−0.5): attempt every question where you can eliminate even one option; 50–50s are mandatory attempts. Rule 2 - RRB and MTS-II (breakeven schemes): blind guessing is a coin with your rank on it - only guess after at least one elimination. Rule 3 - everyone: track your personal accuracy in mocks, not the theoretical 25%. If your "eliminated two, then chose" accuracy is 55% in mock analysis (most students' really is, because elimination reflects partial knowledge), the expected values above are underestimates and you should attempt more, not fewer.
Panic-guessing in the last five minutes. Twelve rushed blind guesses at −1/3 have zero expected value but huge variance - students routinely wipe out 3–4 earned marks in the final sprint. The fix is procedural, not mathematical: keep a "review" list during the paper, and spend the last five minutes ONLY on questions where you had eliminated options earlier. Practise this exact discipline in full-length mocks, where your scorecard shows wrong-answer damage separately - watch that number fall week by week.
Blind guessing has exactly zero expected value at +1/−1/3, but it adds score variance that can move your rank by thousands. Guess only after eliminating at least one option, which makes the expected value positive (+0.11 or better per attempt).
At +2/−0.5, even a blind guess has a slightly positive expected value (+0.125 per question), and a 50-50 guess is worth +0.75 - so in SSC, any question where you can eliminate options should be attempted.
Track marks lost, not count: if negative marks exceed roughly 4-5 in a Railway paper or 5-6 in an SSC paper, your attempting strategy - not your knowledge - is the leak. Use mock analysis to separate concept errors from guessing errors.