We select data-driven prognostic covariates for adjustment to broaden the eligibility criteria without sacrificing power of randomized control trials