The increasing availability of real-world data allows the agricultural value chain to account for quality at the variety level. Cotton presents diverse quality endpoints, with varieties often traced from field to bale. In this study, 79 cotton varieties from 300,000 bales were evaluated across 12 gin-years to determine ginning performance (gin speed), stability (field versus controlled phenotypic data), and potential for probabilistic variety identification. This study demonstrates that varieties can be associated with downstream processing efficiency, that continuous variety properties can be correlated to controlled trial data, and that variety names can be probabilistically imputed, even on tough, real-world holdout samples.
