Important since PP players have more chances to score a goal. One way to do this would be to assign a 1 for any NHL player that has acquired >= 5-10 PPP's, although this will not be 100% accurate. Scraping the names off of daily faceoff would be good but we won't be able to find this data for every single past game. This would also be really computationally expensive.
However, when we create test sets for each day of games, we will be using the PP names off of daily faceoff. Does it matter that the methods of generating this feature differ between the training and testing?
Important since PP players have more chances to score a goal. One way to do this would be to assign a 1 for any NHL player that has acquired >= 5-10 PPP's, although this will not be 100% accurate. Scraping the names off of daily faceoff would be good but we won't be able to find this data for every single past game. This would also be really computationally expensive.
However, when we create test sets for each day of games, we will be using the PP names off of daily faceoff. Does it matter that the methods of generating this feature differ between the training and testing?