Scores on graduate school admissions exams like the Graduate Record Examination (GRE) do not predict successful completion of Ph.D.'s in physics, a new study shows. The data call into question the effectiveness of typical admissions criteria for physics Ph.D. programs, which often require GRE test scores from applicants. GRE scores have large performance gaps based on race, gender and citizenship, driven by factors like stereotype threat, test anxiety and unequal access to expensive coaching or resources. However, studies suggest that up to 40% of U.S. physics programs use cutoff GRE scores in practice - a selection strategy that hinders the growth of diversity and equity in physics, which is already the least diverse of all sciences. To prevent the misuse of metrics that reflect inequality across various demographics, it is critical to evaluate the validity of these metrics. Casey Miller and colleagues conducted a statistical analysis of a sample of roughly one in eight students (3,962 students in total) who entered physics Ph.D. programs of various rankings from 2000 to 2010. Modelling Ph.D. completion as a function of the student's undergraduate grade point average (GPA), GRE scores, citizenship, race and ethnicity, and program ranking, the researchers found that across all correlation models, GRE Physics Subject Test Scores, gender and citizenship all had no bearing on Ph.D. completion. Undergraduate GPA was the only and most robust numerical predictor of Ph.D. completion. Probability of completing the Ph.D. changed by less than 10 percentage points for physics major students scoring in the 10th versus 90th percentile of U.S. GRE test takers. Based on their results, the authors advocate for a more holistic approach to graduate admission that incorporate factors like research experience in addition to academic standing to select for the next generation of research physicists.