Juno can report scores as Raw Points (like 17/20), Raw Percents (like 85%), or Normalized Percents. The Normalized Percent is adjusted for the probability of guessing, so if you guess randomly, you'd likely get a 0, the same as if you left it blank.
For example, if a student got 15 out of 20 true-false questions, the raw percent is 75%. However, just by guessing randomly, even a monkey would likely get 10/20, so 10 are ignored, and the resulting Normalized Percent is calculated as:
(15-10) / (20-10) = 5/10 = 50%
Likewise, if the student got 15 out of 20 multiple-choice questions with 4 choices each, they are likely to get 5/20 just by guessing, so 5 are ignored and the Normalized Percent is:
(15-5) / (20-5) = 10/15 = 66.7%
The formulas get much more complex for multiple-answer, matching, sorting, and partial credit. (For write-in answers, it assumes the chance of guessing is negligible, so it's the same as the raw percent.)
Normalized Percents give a more accurate measurement of what the student actually knows. This is how standardized tests are scored, like the SAT. This is a powerful statistical tool for high-stakes exams and professional settings. But for K-12 classrooms, it can be confusing to students, because the formulas can get complex, and the Normalized Percent is sometimes much lower than the Raw Percent. To use this for traditional grades, you would need to grade "on a curve" (or more technically, scale the percents), e.g. to count 60% as a C-.
The Normalized Percent has a minimum of 0% (so even if the student guesses worse than average, it won't produce a negative score). If the student gets all the questions correct, the Normalized Percent is 100%, same as the raw percent. (It may be higher than 100% if the test has extra credit.)