ABSTRACT Studies of financial markets have shown markets inefficiently underreacting to surprise information, discounting it too heavily in the short term. This paper explores this phenomenon in a new but analogous setting: Major League Baseball. I analyze how surprise information – the deviation between a team’s realized win total and the preseason expectations of betting markets – is incorporated into the teams’ personnel investments and into betting markets. Using fixed effect panel regressions, I find that higher win totals are associated with increased investment in the following season (as measured by team payroll and investment in free agents) and that, conditional on win totals, underperforming (overperforming) expectations in the prior season is associated with more (less) investment. This is consistent with teams discounting some of the surprise information. The effect is more pronounced for teams near the threshold of a playoff appearance. I find that some discounting is rational since roughly 27% of the gap between realized performance and expected performance does not persist into the following season. I find no evidence that betting markets are inefficient in their incorporation of surprise information. However, estimates suggest that teams significantly underreact to surprise information, an inefficiency analogous to those observed in financial markets.