This report presents prediction equations for 12 variables that are indicative of metallurgical coal quality. The variables include six for coke strength and six for carbonization product yields. All variables were predicted for coking coals having volatile matter contents from 16.3 to 42.1 percent, dry basis. The prediction equations, developed by mathematical-statistical techniques, were based on information from the proximate and ultimate analyses, the size consist, and the petrographic composition of bituminous coals carbonized in bm-aga retorts at 800 deg., 900 Deg., and 1,000 deg. C. The equations were evaluated for validity with independent data derived from the 900 deg. C carbonization of 33 binary blends of various coals and 22 binary blends of Pocahontas No. 3 Low-volatile bituminous coal with 11 high-volatile bituminous coals. Results show that metallurgical coal quality can be predicted ahead of mining because data used to derive the prediction equations can be obtained at the time of exploration. Capability to forecast coal quality is useful in mine planning and should provide basic information for optimization of blending and crushing techniques and thus lead to the production of uniform coke.