A data-driven personality typology built from real ratings of 2,113 fictional characters — not from theory.
This system is built on the Statistical "Which Character" Personality Quiz (SWCPQ) dataset from Open Psychometrics. Hundreds of thousands of people rated 2,113 fictional characters on 500 binary personality trait oppositions using a 0–100 slider.
Each character's profile is the average rating across all raters — giving a robust, noise-averaged personality vector that represents how the collective sees that character.
The axes aren't invented — they're discovered. Principal Component Analysis (PCA) on the 500-dimensional trait space reveals the directions along which characters actually vary most. We then select the components that are neutral (neither pole is "better") to build the 16-type system.
Extract the principal components — the axes of maximum variation in how characters differ from each other.
From the top components, select 4 where neither pole carries a moral or competence judgment. These become the type-defining axes.
The remaining major components (PC0 and PC2) carry moral and competence valence. These create 4 sub-quadrants within each type.
Fit a 10-factor model on the trait space. This latent space drives the adaptive Bayesian quiz engine, which selects the most informative next question in real time.
Project factor scores to PCs, threshold on the 4 neutral axes for the type code, and threshold on the 2 valence axes for the quadrant.
Together these 4 axes explain 40.9% of the total variance in the 500-trait space. Each creates a genuine division — no bad poles.
The single largest dimension of personality variation. How you approach action and decision-making.
Spontaneous, reactive, goes with the flow. Comfortable with uncertainty and chaos. Makes decisions in the moment.
Planned, methodical, thinks before acting. Values preparation and strategy. Seeks to control outcomes.
Your style of engagement with the world and other people.
Refined, diplomatic, socially graceful. Communicates with nuance and tact. Values aesthetics and presentation.
Blunt, straightforward, no-nonsense. Values authenticity over polish. Communicates plainly.
Your orientation toward convention and ideas.
Unconventional, creative, challenges norms. Driven by ideas, principles, or vision. May be eccentric or rebellious.
Conventional, grounded, follows established paths. Values what works over what's novel. Reliable and down-to-earth.
Your emotional style and intensity.
Emotionally intense, driven by feeling. Wears heart on sleeve. High emotional stakes in relationships and goals.
Emotionally even, cool under pressure. Controls emotional expression. Appears calm or detached.
These axes carry moral and competence judgment — they're deliberately not used to define the 16 types. Instead, they create 4 sub-quadrants within each type, so every type contains heroes, villains, and everything in between.
Devoted: Loyal, selfless, team-oriented. Follows a cause or person.
Independent: Self-interested, autonomous. Follows their own agenda.
Decisive: Confident, competent, in control. Gets things done.
Striving: Struggling, developing, uncertain. Still figuring it out.
The hero variant. Loyal AND competent. Aragorn, Princess Leia, Mufasa.
The loyal underdog. Good-hearted but developing. Frodo, Forrest Gump, C-3PO.
The antihero. Capable but self-serving. Cersei Lannister, Tyler Durden, Loki.
The troubled soul. Self-interested AND struggling. Sheldon Cooper, Voldemort.
The quiz doesn't use simple scoring. It uses Bayesian inference with a full factor model to classify you in real time.
A 10-factor model is fit on the 500-item trait space via factor analysis. Each question is a trait opposition (e.g., "playful vs serious") answered on a slider.
After each answer, the engine performs a Bayesian update — a Kalman filter step that adjusts the belief about where you sit in the latent personality space. Your uncertainty shrinks with every answer.
The next question is selected by maximum Expected Information Gain — the question that will reduce uncertainty the most, given what's already known. This means the quiz never wastes a question on dimensions it's already pinned down.
Unlike MBTI or similar systems, every axis has two equally valid poles. "Improvisational" is not better or worse than "Deliberate." The valence axes (good/evil, competent/striving) are tracked separately so every type contains heroes and villains.
The axes come from PCA on real response data, not from theory. They represent the actual dimensions along which characters (and by extension, people) vary most. The total variance explained by the system is approximately 55–60%.
The quiz provides probability estimates, not binary classifications. You see how confident the system is about each axis, and your full probability distribution across all 16 types updates after every answer.
25 questions. Bayesian inference. Your closest character matches.
Take the Quiz