The purpose of this repository is to provide all research related to the paper entitled Judgement Under Uncertainty: An Empirical Evaluation of NHL Draft Picks.
The Draft Trades Dataset was webscraped from ProSportsTransactions and includes 356 trades from 1980-2020 involving only draft picks. Each row of the csv file includes relevant information on the trade, including picks exchanged and picks received, as well as the year of each draft pick. The 356 draft trades can be separated into two groups: 191 involving picks from the current year, and 165 involving both current and future picks.
The Draft Trades Analysis script was used to develop the results found in Section 3.1 of the paper and the discussion of Section 4.1. The first half of the notebook develops the NHL Trade Value Chart using the Weibull Distribution proposed by Thaler and Massey. This notebook first processes the data and analyzes the trades which involve picks from the current year. Using the MPE Minimization function, the notebook determines the parameters lambda and beta of the Weibull distribution. Figure 1 of the manuscript can be seen in the notebook, along with statistics of goodness of fit. The remainder of the notebook applies these parameters to determine the discounting applied to future draft picks, calculated as the percentage difference between the difference in draft value in the absence of future picks and the estimated value of the future picks.
The Point Shares of players from the 1980-2010 draft classes were then webscraped from hockey-reference.com. This dataset includes 7751 players: 4451 forwards, 2532 defensemen, and 768 goalies. Each row of the csv file includes the draft pick, the name of the player, position, Games Played, Point Shares, and Time on Ice. Drafted players whose information was not given on the hockey-reference.com website were referred to by the name of the player drafted before them whose information was known. However, these players position was assigned based on their designation when they were drafted.
The Point Shares Analysis script was used to develop the results found in Section 3.2 and the discussion of Section 4.2. The notebook preprocesses the Point Shares Dataset and completes the process described in Section 3.2 to create the Point Shares Value Chart, using the Weibull Distribution proposed by Thaler and Massey. For all players, along with forwards, defensemen, and goalies seperately, the notebook determines the difference in draft valuation and how it compares to the overall Draft Trade Value Chart. Figures 2-4 in the manuscript were developed using this notebook.