Although the use of bibliometric indicators for evaluations in science is becoming more and more ubiquitous, little is known about how future publication success can be predicted from past publication success. Here, we investigated how the post-2000 publication success of 85 researchers in oncology could be predicted from their previous publication record. Our main findings are:​ (i) Rates of past achievement were better predictors than measures of cumulative achievement. (ii) A combination of authors’ past productivity and the past citation rate of their average paper was most successful in predicting future publication success (R² ≈ 0.60). (iii) This combination of traditional bibliographic indicators clearly outperformed predictions based on the rate of the h index (R² between 0.37 and 0.52). We discuss implications of our findings for views on creativity and for science evaluation.