Our Ranking & Normalization Methodology

When a user provides their answer sheet URL or HTML code, our headless parser inspects the table layout:

• Candidate Metadata Extraction: Extracts Roll Number, Candidate Name, Exam Center, Date, Shift Time, and Subject.
• Option Matching: Maps candidate's chosen option against the commission's correct option marked in the response markup.
• Negative Marking Evaluation: Applies precise exam-specific penalties (-0.50 marks for SSC CGL Tier 1; -1.00 marks for Tier 2; -0.333 marks for RRB NTPC).

Because exam difficulty varies across shifts, raw marks cannot be directly compared across different days. We apply the official Mean-Standard Deviation normalization model:

This model scales raw marks based on whether your shift's top 0.1% mean was higher or lower than the overall mean, ensuring complete statistical fairness.

To prevent artificially inflated or deflated rank benchmarks:

• Roll Number & Shift Hash De-duplication: Multiple submissions of the identical response sheet update the single entry rather than creating duplicates.
• Outlier Filtering: Submissions with physically impossible completion patterns are excluded from the normalization pool.

Candidate privacy is paramount. RankCheck.karo adheres to strict ephemeral processing rules:

• No candidate passwords or login credentials are ever requested or stored.
• Response sheet HTML documents are parsed on the fly and immediately discarded from memory.
• Candidate's calculated scorecards are saved locally in their personal browser's localStorage for quick re-inspection.