The impact of uncertainty on regularized learning in games