About IQ WORLD RANKING

A service that visualizes traits such as logical thinking and reasoning and lets people compare them across borders.

How we measure

Each question has a difficulty and a discrimination value. We estimate ability θ with a two-parameter logistic item response theory (IRT) model, so results from different question sets can be compared on one scale. Estimated IQ = 100 + 15 × θ (standardized against the reference group), shown with a 95% confidence interval from the measurement error. Response time is not used to estimate ability; it is used only for tie-breaking and the separate speed board.

Current measurement basis

  • Scale version: provisional-2026.1
  • Model: irt_2pl_eap
  • Standardization: Not standardized (provisional)
  • Reference group: 未標準化(暫定):問題の難易度は制作時の見積もり。受験データによる校正前。

Limitations

About the questions

All questions are original (no commercial test content is reproduced). Generated questions are checked by a separate procedure to confirm exactly one correct answer, and only those that pass are used. Variants that only change numbers or figures are capped per structure, and one test never contains two questions of the same structure. Our goal is 10,000 questions; current counts are below.

24,041Registered
24,041Verified (machine + human)
0Human-reviewed
15,156Distinct structures
10,000Target

How ranks are decided

Equal estimated IQs (shown rounded) are not tied; they are compared in this order:

  1. Ability estimate on the common scale (unrounded)
  2. Number of correct answers
  3. Difficulty-weighted score
  4. Accuracy
  5. Total answer time measured by the server (shorter is better)
  6. Average of domain scores
  7. Completion time (earlier is better)

Fairness

Correct answers are never sent to the browser; scoring and timing happen on the server. Results with anomalies such as extremely fast answers are not judged as cheating automatically; they are held from the rankings until an administrator reviews them.