The Punishment score is a scoring model with a penalty: how to get a score on a multiple-choice test by subtracting the score for wrong answers. Penalties are given to educate students, so they only guess the answers to questions they understand, and these guesses can produce wrong and correct answers. To improve the quality of learning, it is necessary to prevent students from guessing answers, including scoring with a penalty. Based on these assumptions, the problem arises 'How is the application of the punishment score model to the mathematics learning outcomes of high school students in Jakarta?" This study aims to find empirical evidence about applying the punishment score model to the mathematics learning outcomes of high school students in Jakarta. The study targets are: (1) to determine the most appropriate number of options on a multiple choice test using the punishment score model and (2) to form an honest character in students. This study uses a quasi-experimental method. The research instrument was a multiple-choice test in Mathematics. There are 30 test items. Data were obtained from a three-option multiple-choice test in the first group and a five-option multiple-choice test in the second group with a punishment score model. Furthermore, the average score fairness index was calculated for each group using the Donlon and Fischer method. The higher the fairness index, the more good the student scores. The results showed that the fairness index of the scores on the five-option multiple-choice test was higher than the three-option multiple-choice test. So the five-option multiple choice test in the punishment score model is more appropriate to use to improve the quality of learning. In implementing the research, the Donlon and Fischer method can be used to analyze the results of mental measurements by calculating the item difficulty level on the delta scale.

punishment score, fairness index score, penalty, multiple choice test