The Graduation Success Rate (GSR) plays a critical role supporting the NCAA's Collegiate Model of amateur college athletics. The NCAA created the GSR to correct a statistical bias in the legally mandated Federal Graduation Rate (FGR) that causes it to underestimate rates. But the GSR's attempted correction causes it to overestimate rates. This paper reports the first estimate of the size of this statistical bias. The focus is on the big revenue sport of men's basketball in the so-called Power Conferences. The small size of basketball squads allows a reasonably accurate estimate of GSR cohort sizes based on publically available data. This in turn enables the calculation of a "corrected" GSR. The results indicate that the GSR exaggeration is large, perhaps as much as 20 percentage points. This raises fundamental questions about the success of the GSR as a useful graduation rate metric.