How the SBTI Test Scores Answers and Chooses a Type

This article explains the rules currently used by this site. It follows an answer from its numeric option value to a dimension total, then to a template comparison and a final result. Being able to repeat that calculation makes the process inspectable. It does not establish that the questionnaire is a validated psychological instrument, or that its categories predict someone’s behavior outside the test.

There are thirty normal questions, assigned in pairs to fifteen dimensions. The dimensions belong to five groups: self, emotion, attitude, action drive and social behavior. Question identifiers determine those assignments. Moving a question to a different screen position does not change the dimension it contributes to. The interest question and conditional drinking question take a separate path; their values are not added to the normal dimension totals.

Treat each answer as a report about a particular situation. A preference about making decisions is not a certificate of competence, and a preference about personal space is not a moral grade. The result compresses a limited set of choices into an entertainment label. Understanding that compression is useful precisely because it shows what the label leaves out, including context, intensity and behaviors the questions never ask about.

The normal questions offer three numeric choices, but position is not a universal scoring key. Questions q14 and q27 use values three, two and one in that order; the other normal questions use one, two and three. Repeatedly pressing the first option therefore does not mean repeatedly contributing one point. A manual check must use the actual value attached to each answer and its assigned dimension.

Two valid normal answers give each dimension a total between two and six. Totals of two or three become L; four becomes M; five or six become H. The implementation expresses this as at most three for L, exactly four for M and otherwise H. Those branches describe the stated bands when all normal answers are complete and valid.

For example, one plus two gives three and therefore L. One plus three gives four and therefore M. Two plus three gives five and therefore H. Notice that moving from three to four crosses a boundary, while moving from two to three does not. This loss of detail explains why a small answer change can sometimes affect matching sharply and sometimes leave the comparison unchanged.

The level vector follows a fixed order: S1 through S3, E1 through E3, A1 through A3, Ac1 through Ac3 and So1 through So3. For matching, L, M and H become one, two and three. Each of the twenty-five normal types supplies its own fifteen-position template. The algorithm takes the absolute difference at each position and adds those differences to obtain a distance.

Matching levels contribute zero. Neighboring levels contribute one; L against H contributes two. Across fifteen positions the maximum possible distance is thirty. Type names and code letters do not enter this arithmetic. DRUNK and HHHH are special outcomes without normal templates, so they are not additional competitors in the distance ranking.

Candidates are ordered by smaller distance, then more exactly matching dimensions, then higher similarity. Exact matches count equal positions, not correct answers: this test has no answer key for a successful personality. A nearby alternative can still be worth reading. Compare the dimensions where its template differs rather than assuming that the first label makes every other description irrelevant. A person’s history contains much more information than any of these templates.

A hand calculation can use the same rule on a few positions before checking the whole vector. If your first level is L and the candidate has M, that position adds one. If your next level is H and the candidate has L, it adds two. Equal levels elsewhere add nothing. You must still inspect the remaining positions before choosing a winner: a candidate that matches your favorite dimension can lose overall. There is no extra weight attached to a memorable question or to a type name you prefer.

Similarity is one minus distance divided by thirty, multiplied by one hundred, rounded to an integer and bounded below by zero. A distance of six gives eighty; twelve gives sixty; thirteen rounds to fifty-seven. The displayed percentage is a conversion of template distance. It is not a probability that you belong to a population category, a confidence interval or an accuracy score.

The fallback to HHHH happens only when the best normal similarity is below sixty. Exactly sixty keeps the normal result. To check an example, apply rounding before the strict comparison. HHHH means that the available templates do not fit the level combination closely enough under this rule; it does not establish exceptional intelligence, psychological abnormality or rarity among people.

Comparing one person’s eighty with another person’s ninety is also misleading. Each may be close to a different template. A higher figure does not repair the questionnaire’s limited coverage or remove the effects of interpretation and current circumstances. Ask which specific dimensions fit your experience, and which do not. Concrete examples make that discussion more useful than treating the percentage as a personal rating.

The interest question has four options. Choosing its third option, drinking, gives value three and inserts the drinking-attitude question in the browser. Changing that interest answer to another option removes the follow-up and clears its previous answer. This prevents an answer to a now-hidden question from continuing to control the outcome.

Seeing the follow-up is not enough to trigger DRUNK. Its second answer, the exaggerated statement about treating spirits in a thermos as drinking water, has value two and sets the drinking flag. The first, moderate-drinking answer does not. Both the entrance choice and the follow-up choice matter in the normal interactive flow.

Final selection checks that flag first. If it is set, DRUNK wins; otherwise the algorithm checks the below-sixty fallback and then uses the closest normal type. The normal dimensions are still calculated. This branch is a joke attached to particular selections, not an estimate of alcohol consumption or an assessment of dependence. Real concerns about drinking need evidence about real behavior and appropriate health advice, regardless of what an entertainment page calls the result.

Start with the fifteen raw totals, convert each to its level, and manually compare one template. Then apply similarity rounding and the special-result branches. This separates a calculation question from disagreement with a description. If the arithmetic differs, inspect answer values and assignments; if the wording feels wrong, investigate the wording rather than repeatedly retesting until it sounds flattering.

The same valid answers and drinking flag under the same rules should repeat the calculation. If a retest changes, examine changed answers, especially totals crossing three to four or four to five. Also check the supplemental choices. Other sites may use different wording, templates or rules despite sharing the SBTI name. This walkthrough makes a claim about this implementation, not every page carrying that label.

You can accept a useful observation and reject a broad characterization. Neither understanding the method nor obtaining a high similarity requires adopting a type as your identity. Use the result as a conversation starter about habits. Persistent distress, health concerns and serious relationship decisions deserve support and information beyond an entertainment classification; employment screening and clinical assessment require very different standards of evidence.