Not feeling the buzz: Correction study of mispricing and inefficiency in online sportsbooks

We present a strict replication and correction of results published in a recent article (Ramirez, P., Reade, J.J., Singleton, C., Betting on a buzz: Mispricing and inefficiency in online sportsbooks, International Journal of Forecasting, 2022, doi:10.1016/j.ijforecast.2022.07.011). RRS introduced a novel "buzz factor" metric for tennis players, calculated as the log difference between the number of Wikipedia profile page views a player receives the day before a tennis match and the player's median number of daily profile views. The authors claim that their buzz factor metric is able to predict mispricing by bookmakers and they demonstrate that it can be used to form a profitable strategy for betting on tennis match outcomes. Here, we use the same dataset as RRS to reproduce their results exactly. However, we discover that the published results are significantly affected by a single bet (the "Hercog" bet) that returns substantial outlier profits; and these profits are generated by taking advantage of erroneously long odds in the out-of-sample test data. Once this data quality issue is addressed, we show that the strategy of RRS is no longer profitable in "practical" scenarios. Using an extended and cleaned dataset, we then perform further exploration of the models and show that the "impractical" betting strategy that uses best odds in the market remains profitable (in theory). However, evidence suggests that the vast majority of returns are generated by exploiting individual bookmaker's mispricing of odds relative to the market, and the novel buzz factor metric has negligible contribution to profits. We make all code and data available online.