Risk-taking at the outset of competition: Evidence from tournament chess

Abstract: What drives risk-taking at the outset of competition? This paper explores the role of baseline performance expectations, which predict who is more likely to win (the favorite) or lose (the underdog) in one-on-one competition. We used large-scale data from face-to-face tournament chess (5,676,486 games played by 225,200 players), a competitive context in which performance expectations are exceptionally accurate and salient. To study risk-taking, we developed and empirically validated a measure based on the concept of a gambit, a strategy involving an early material sacrifice that is widely regarded as risky. Our analyses demonstrate that higher expectations are robustly associated with increased risk-taking, including in settings where pairings are assigned in random order and after accounting for player characteristics. These findings would not have been predicted by the accepted view that favorites seek to avoid unexpected losses (implying low risk-taking) and underdogs pursue unexpected wins (implying high risk-taking).

Keywords: risk-taking; expectations; competition; chess; specification-curve analysis