mrcaseb
commented
11 months ago with
dplyr::summarise(
# OFFENSE
off_scr = weighted.mean(.data$off_scr, .data$off_n, na.rm = TRUE),
off_scr_1st = weighted.mean(.data$off_scr_1st, .data$off_n, na.rm = TRUE),
off_scr_2nd = weighted.mean(.data$off_scr_2nd, .data$off_n, na.rm = TRUE),
off_scr_3rd = weighted.mean(.data$off_scr_3rd, .data$off_n, na.rm = TRUE),
off_scr_4th = weighted.mean(.data$off_scr_4th, .data$off_n, na.rm = TRUE),
off_1st = weighted.mean(.data$off_1st, .data$off_n, na.rm = TRUE),
off_td = weighted.mean(.data$off_td, .data$off_n, na.rm = TRUE),
off_fg = weighted.mean(.data$off_fg, .data$off_n, na.rm = TRUE),
off_punt = weighted.mean(.data$off_punt, .data$off_n, na.rm = TRUE),
off_to = weighted.mean(.data$off_to, .data$off_n, na.rm = TRUE),
off_n = sum(.data$off_n, na.rm = TRUE),
# DEFENSE
def_scr = weighted.mean(.data$def_scr, .data$def_n, na.rm = TRUE),
def_scr_1st = weighted.mean(.data$def_scr_1st, .data$def_n, na.rm = TRUE),
def_scr_2nd = weighted.mean(.data$def_scr_2nd, .data$def_n, na.rm = TRUE),
def_scr_3rd = weighted.mean(.data$def_scr_3rd, .data$def_n, na.rm = TRUE),
def_scr_4th = weighted.mean(.data$def_scr_4th, .data$def_n, na.rm = TRUE),
def_1st = weighted.mean(.data$def_1st, .data$def_n, na.rm = TRUE),
def_td = weighted.mean(.data$def_td, .data$def_n, na.rm = TRUE),
def_fg = weighted.mean(.data$def_fg, .data$def_n, na.rm = TRUE),
def_punt = weighted.mean(.data$def_punt, .data$def_n, na.rm = TRUE),
def_to = weighted.mean(.data$def_to, .data$def_n, na.rm = TRUE),
def_n = sum(.data$def_n, na.rm = TRUE),
.groups = "drop"
) %>%
dplyr::relocate("off_n", .after = "team") %>%
dplyr::relocate("def_n", .after = "off_to")
https://github.com/nflverse/nflfastR/blob/c1ca9f234f459fa79330af0ba32bcf43242fb210/R/calculate_series_conversion_rates.R#L170
We summarize the week level data in the above line to the season level. This calculates a mean over means without weights and therefore falsely summarizes the percentages.