If you dramatically redefine random to include choices between useful outcomes (ie. the non random boundary conditions to the decision tree), then Occam’s razor works well here.