The economic value of cotton to the farmer greatly depends on the quality as determined by the United States Department of Agriculture (USDA), Agricultural Marketing Service (AMS), Cotton Division, Cotton Classing Office (CO). Optimization of gin processing depends on the accurate prediction of the AMS CO quality measurements in real time in the gin. Cotton samples with a wide variety of trash content were generated with a variety of gin cleaning machinery and sent to the Dumas CO for classing. Data were collected while ginning the samples with equipment similar to that used in gin process control in the Microgin facility at the U.S. Cotton Ginning Laboratory, Stoneville, MS. About 17 readings of samples of each of the 122 lots were made. Similar data from a commercial gin were also analyzed. Several models were evaluated and one model was chosen to predict the AMS CO leaf grade based on readings made while ginning. This model predicted leaf grade correctly 64.5% of the time and was within one grade 99% of the time. This was an improvement over the 40-60% correct predictions experienced in 1994-1996 based on AMS CO data from 1991-1992.