Much of the recent research on homework focuses on using online, web-based, or computerized homework systems. These systems have many reported capabilities and benefits, including the ability to randomize values, which enables multiple attempts by a student or to reduce academic dishonesty. This study reports on the impact of using randomized versus static homework values on student learning as measured by student performance on similar exam questions. Results indicate randomization can have a significant practical and statistical impact on improving exam scores, can reduce the rate of cheating, and can make homework a more accurate measure of student ability as the difference between homework scores and exam scores is reduced.