How the Soul Virtues Extractor Scores You: The Model Behind the Seven Virtues

By the maker of Soul Virtues Extractor · 9 September 2026 · about 7 minutes

Most soul quizzes ask ten questions and drop you into one box. This one asks 66 and gives you seven numbers. People keep asking how the numbers are made, so here is the whole model: the weights behind every statement, the reason a “no” costs more than a “yes” pays, what happens when you feed it random answers, and what 500 real results look like.

Seven traits the game never defined

Undertale gives each of the six fallen children a color and a pair of objects, and gives the player a seventh color, red. It never says what patience or bravery means as a trait, and it never scores anyone. The lore is covered on the seven souls page and the game itself on the Undertale explainer; this post is about the part the game left blank. To turn seven colors into something a person can be measured on, we had to write definitions, write statements that separate the definitions from each other, and decide how an answer should move the numbers. Every one of those decisions is an opinion, so they are written down here where you can argue with them.

One statement, several virtues

There are 66 scored statements and three story beats that do not score. Each statement carries weights on one to three virtues. Here are four of them exactly as they sit in the question bank:

The second column is the whole idea. Acting first in an emergency is brave, and it is also impatient, so agreeing with that statement adds to Bravery and takes a little from Patience. Reporting a friend is just, and it costs kindness. Across the bank, 21 statements touch a single virtue, 42 touch two, and 3 touch three; 16 are reverse-keyed, meaning agreement lowers the virtue (“When something gets hard, I quickly lose interest” is a Determination statement with a negative weight). Each virtue leads nine or ten statements: Determination, Kindness and Integrity lead ten, the other four lead nine.

The side weights are spread unevenly, and that matters later. Patience is the virtue most often charged for someone else’s statement: six statements subtract Patience when you agree with them. Kindness is charged five times, Determination four. Perseverance is never charged at all.

Why a “no” costs more than a “yes” pays

Each statement has five answers, and they pay unevenly. The multiplier on the weight is:

Agreeing with a flattering sentence is cheap; most people will nod at “I keep the promises I make to myself.” Refusing one is a stronger signal, so the model pays more attention to it. A virtue you actively push away from drops faster than a virtue you politely agree with rises.

Your raw total on each virtue is then divided by the most that virtue could possibly score, curved a little so small leans stay visible, and mapped onto a 0–100 scale where 50 means neutral. One worked example, starting from a sheet of all-neutral answers: strongly agreeing with “If a plan of mine falls apart, part of me refuses to accept it” moves Determination from 50 to 61.4 and Patience from 50 to 42.9. Strongly disagreeing with the same sentence sends Determination to 36.9 and Patience to 56.3.

Three sheets get a named result instead of a spread. Answer neutral to everything and you are The Unreadable; strongly agree with everything, contradictions included, and you are The Mirror-Eater; strongly disagree with everything and you are The Closed Door. The device can tell when it has learned nothing.

What 100,000 random players would get

The first thing to check in any quiz is whether the deck is stacked. We ran 100,000 answer sheets through the real scoring code with every answer chosen at random. If the bank secretly favored one color, it would show up here.

A perfectly flat deck would give each virtue 14.3%. The real spread runs from 12.9% to 16.4%, which is about as flat as 66 hand-written sentences get; Perseverance edges ahead because it is the one virtue nothing subtracts from. Three more things fell out of the same run.

A random sheet averages 44 on every virtue, below the neutral 50, because disagreement costs more than agreement pays. Only 0.12% of random sheets finish with all seven virtues above 50, so a result where everything is above the line is a real statement about the person, and a fairly rare one. And 21% of random sheets are close calls, with the top two virtues inside 2.5 points of each other; that is why the result page prints the margin under your top virtue instead of pretending the ranking is settled.

One sheet is worth naming. Agree with everything, and you get Perseverance 73.5, Bravery and Justice 70.1, Determination 67.8, Integrity 65.2, Kindness 63.6, Patience 52.7. Patience barely moves because it is the virtue the other statements keep charging. The only way to score Patience is to disagree with the rushing sentences, which is exactly what a patient person would do.

What 500 real results look like

Results stay on your device; we do not log them (see the privacy page). The one place a result reaches us is when someone opens a checkout for the optional Deep Reading, because the reading has to be generated from their numbers. Between 16 August and 8 September 2026 that happened 500 times, and 210 of those went on to buy. So this is the population of people who wanted to read more about themselves, and it is the only real distribution we can show honestly.

Two virtues carry more than half of everyone who reaches the checkout. Kindness finishes first for three in ten, Perseverance for one in four, and Patience for fewer than one in thirty. Against a near-flat random baseline, that lean comes from the people, and we see three explanations we cannot separate with this data.

First, self-report flatters. “I notice when someone in the room is being left out” is easy to agree with, and agreeing with kind sentences is cheap by design; the model discounts agreement, and Kindness still comes out ahead. Second, selection: someone who answers 66 statements and then reads three free entries before deciding to buy is, by that very act, a persevering person, so Perseverance is over-represented among readers. Third, structure: Patience is the most-charged virtue in the bank, so a Patience-first result requires a run of deliberate disagreements, and few people play a personality quiz that way. The chart is the honest picture; the reasons are our reading of it.

What the seven-soul system is, and what it is not

In the game the traits are flavor. Each belongs to one child, is announced by two objects, and is never tested. The test treats them as seven continuous dimensions that everyone has some amount of, which is closer to how trait psychology models a person than to how a story assigns a color. Some of the overlap is obvious once you look: Perseverance and Integrity sit near conscientiousness, Kindness near agreeableness, Bravery near the low end of anxiety, and Justice has no clean counterpart in the common five-factor model at all. Determination is the strangest of the seven. In the game it is the power to reload a save; in the test it is the refusal to accept an ending, which reads more like grit than like a trait a questionnaire can separate from Perseverance. That confusion is real, people report it, and the result page addresses it directly under each virtue.

What the test cannot do is also worth stating plainly. It is a one-sitting self-report with no retest data yet, it can be gamed by agreeing with everything (and will tell you so), and it is written for reflection and fun, which is why every result page and the terms carry the line that it is not a psychological assessment. What would make it better is more real data than 500 checkout sheets, gathered with consent; if we add an opt-in to share an anonymous result, it will be announced here first.