Leaps in Innovation: The Bolt versus Bannister Effect in Algorithmic Tournaments

Joel Wooten, Assistant Professor of Management Science, University of South Carolina

Abstract: This paper explores whether innovation breakthroughs stimulate or impede future progress in individual innovation. On the one hand, one could argue that substantial improvements to the status quo might inspire advances through competition. On the other hand, one could claim such improvements might have the opposite effect, stifling motivation or creativity in rivals. Using a unique data set of predictive modeling contests from Kaggle we analyze 25,898 distinct attempts at innovation. We address two related questions to frame our central theme: (1) What effect do discontinuous leaps (as opposed to incremental steps) in innovation contests have on future progress? (2) What predicts such discontinuous leaps in innovation contests? The answers to these questions are as follows. Behavior after discontinuous leaps differs from behavior after continuous steps in innovation tournaments. We find that leaps result in increased rates of entry submission early in contests and improved scores going forward. For the second question, the entrant characteristics that predict leaps turn out to be somewhat different than those that predict steps. While more entries and bigger teams increase the likelihood of any improvement, steps are more likely with bigger teams, bigger prizes, and higher prior scores. Discontinuous leaps are not predicted by prize amount or prior success. This paper contributes new understanding to the literature on innovation tournaments and offers managers guidance about how to foster leaps in innovation.

Read the full paper here.