A raw score is the direct result of a scoring rule, such as the number of correct answers. A converted score places that total on another scale. Percentage correct is a third representation. Keeping these separate makes a report much easier to read.
One performance, three descriptions
Suppose you answer 22 of starp's 30 questions correctly. Your correct-answer total is 22. Your accuracy is 22 divided by 30, or about 73.3%. The current starp conversion maps that total to an indicative score of 100.
The numbers 22, 73.3 and 100 describe the same performance in different ways. None is a population percentile simply because it appears in a report.
Conversion does not have to be linear
A fixed table can assign unequal score gaps to adjacent raw totals. In starp's current mapping, 20 correct gives 90, 21 gives 95, 22 gives 100 and 23 gives 104. Each extra correct answer raises the score, but the increase is not identical at every step.
It would therefore be wrong to take a percentage, multiply it by a constant and assume you have reproduced the scoring method. Read the actual conversion rules instead.
What happens to the skill breakdown?
The total and the profile summarize different aspects of the result. Two people can both answer 22 questions correctly but distribute those correct answers differently across patterns, spatial, logic and numerical questions.
They receive the same overall score under starp's method, while their skill profiles can differ. That is useful: an identical total does not require an identical practice plan.
Compare like with like
When keeping notes, write the test name, date, correct-answer total and score together. Avoid comparing 80% on one practice sheet with a converted score of 110 on another assessment. First check question count, difficulty, timing and scoring. A conversion changes the representation of performance; it does not make unrelated tests interchangeable.
The individual IQ score guides show whether starp can produce each number and, where it can, the corresponding number of correct answers. This helps keep raw accuracy separate from a modeled percentile.