data set

Training data consists of NFL statistics and draft results 2005-2020. A holdout sample of such in 2021, 2022, and 2023 does not contribute to building the model, but is set aside for validation after the fact.

Fantasy_points(fp) were added for each week of 1-17, subsequently fantasy_value(fv) was also calculated for each week of the season. A binary variable, game_played, is True if a player recorded any stat during the game, otherwise is False.

Player_age(age) ignores month/day but is simply defined as: year-year_of_birth. Tho it is noted, the last regular season game is played in January of the following year. This fact is ignored in the age stratification for simplification, whereby age goes by the season's year instead of the game date.

In the last step, data is aligned by ages 21-45. For example, Brock Purdy's 2023 season would share with Matt Stafford's of 2012 in the column age 24.

<     >