We present a theory explaining the impact of ability tracking on academic performance based on grading policies. Our model distinguishes between initial ability, which is mainly determined by parental background, and eagerness to learn. We show that achievements of low ability students may be higher in a comprehensive school system, even if there are neither synergy effects nor interdependent preferences among classmates. This arises because the comprehensive school sets a compromise standard which exceeds the standard from the low ability track. Moreover, if students with lower initial ability have a higher eagerness to learn, merging classes will increase the average performance.