Negative Marking Calculator
Work out your real score in an exam with negative marking – and see how many marks the wrong answers cost you.
0 / 400
Score 0% of the maximum
- Marks gained: 0
- Marks lost: 0
- Unattempted: 100
- Attempted: 0
- Accuracy: –
- Score without negatives: 0
With +4 / −1, an answer is worth attempting on average when you are more than 20% sure it is right (for 4 options a pure guess is 0.25 marks on average).
Formula: score = correct × marks per correct − wrong × marks lost per wrong. Unattempted questions score 0.
Free text tools · No sign-up needed
More free tools
Utilities
Prompts & ideas
Names
Writing
Social media
Fun
More tools
How it works
Step 1
Enter the questions
Type the total number of questions, how many you answered correctly and how many were wrong.
Step 2
Set the marking scheme
Enter the marks per correct answer and the marks lost per wrong answer, or pick a common scheme such as +4 / −1.
Step 3
Get your score
See your final score, the marks lost to negative marking, unattempted questions and your accuracy.
How negative marking works
In an exam with negative marking, every correct answer adds marks and every wrong answer takes marks away; questions you leave blank score zero. Your score is: correct answers × marks per correct − wrong answers × marks lost per wrong. With the common +4 / −1 scheme, 60 correct and 20 wrong answers give 60 × 4 − 20 × 1 = 220 marks. Many objective tests use schemes like +4 / −1, +1 / −⅓, +2 / −⅔ or +1 / −¼; enter the scheme of your exam to calculate your expected score from an answer key.
Should you guess when there is negative marking?
It depends on how sure you are. With +P marks for a right answer and −M for a wrong one, an answer is worth giving on average when your chance of being right is higher than M ÷ (P + M). For +4 / −1 that is 20%, so with four options a pure random guess (25%) is still slightly positive on average, and eliminating even one option makes it clearly worth it. For +1 / −⅓ the break-even is 25%, so a blind guess between four options is worth zero on average. The calculator shows this break-even for the scheme you enter. It is an average – single guesses can still cost you marks.
Accuracy and attempts
Accuracy is correct answers ÷ attempted answers. Two candidates with the same number of correct answers can end up with very different scores if one of them attempted many more questions and got them wrong. Compare 'score without negatives' with your final score to see how much your wrong answers cost you, and use it when planning how many questions to attempt in the next mock test.
Different schemes in one paper
Some exams use different marking for different sections or question types – for example no negative marking for numerical answers. In that case calculate each section separately with its own scheme and add the results. Always check the official notification of your exam for the exact rules.
Frequently asked questions
How is the score calculated with negative marking?
Score = correct × marks per correct − wrong × marks lost per wrong. Unattempted questions give zero.
What does −1/3 negative marking mean?
For every wrong answer, one third of the marks for that question is deducted. With 1 mark per question, three wrong answers cancel one correct answer.
Do unattempted questions get negative marks?
In most exams, no – only wrong answers are penalised. If your exam is different, count those questions as wrong.
Is my data stored?
No. Everything is calculated in your browser and nothing is sent or saved.
Guides
Opus 5.5 vs GPT-6 (Astra & Sol): Benchmarks, Price, Which to Use
Claude Opus 5.5 vs GPT-6 Astra: $4/$20 vs $10/$50 per 1M tokens. Opus leads Terminal-Bench 4.0 and HLE; Astra leads formal math. Where GPT-6 Sol fits.
Kokoro TTS: How to Use It Online, Locally and in the Browser
Kokoro TTS is a small open-weight text-to-speech model (82M parameters, Apache 2.0). How to try it online, run it on your PC or in a browser, voices and tips.
GPT-6 Sol vs GPT-5.6 Sol: Half the Price, What Changed
GPT-6 Sol vs 5.6 Sol: GPT-6 Sol costs $2/$10 per 1M tokens vs $4/$20, is newer (Sept 22, 2026) and wins coding and terminal tests; 5.6 Sol still leads some.
GPT-6.1 Sol vs GPT-6 Astra: 5x Price Gap, Benchmarks, Which to Use
6.1 Sol vs 6 Astra: GPT-6.1 Sol costs $2/$10 per 1M tokens, GPT-6 Astra $10/$50. Shared benchmarks split 3–3. Which OpenAI model fits coding, agents and math.
Claude Fable 5.5: Release Date, Rumors and What's Confirmed
Is Claude Fable 5.5 out? As of October 8, 2026 Anthropic has not announced it. What's confirmed about Fable 5.1, Opus 5.5 and how to spot a real Fable launch.
Claude Startup Program: Credits, Free Claude Team, Who Can Apply
Anthropic's Claude Startups program (expanded Oct 6, 2026): a one-time $1,000 API credit, a free year of Claude Team for 5 seats, up to $45,000 in partner perks.