Wrong posteam on kickoffs with penalties

[ConnerEvans]

The posteam of line 3 in the photo below should be JAX since they were the receiving team when the penalty was incurred. Because of this, the epa in line 3 should be (0.02529127) - (0.09028234) = -0.06499107 instead of
-(0.02529127) - (0.09028234) = -0.11557361

library(tidyverse)
pbp <- readRDS(url(glue::glue("https://raw.githubusercontent.com/guga31bb/nflfastR-data/master/data/play_by_play_2017.rds")))
pbp %>% filter(game_id == '2017_12_JAX_ARI', between(play_id, 4550, 4700)) %>% select(posteam, down, ydstogo, desc, ep, epa) %>% View

I'm pretty sure this is the same bug, there is a penalty during a kickoff but the posteam doesn't change for the line where the penalty was incurred causing the epa to be -2*(the ep of the kickoff).

library(tidyverse)
pbp <- readRDS(url(glue::glue("https://raw.githubusercontent.com/guga31bb/nflfastR-data/master/data/play_by_play_2018.rds")))
pbp %>% filter(game_id == '2018_07_NE_CHI', between(play_id, 1450, 1550)) %>% select(posteam, down, ydstogo, desc, ep, epa) %>% View

So it seems like the posteam has a bug where it says it is the kicking team if there is a penalty when it should be the receiving team. This in turn affects the epa.

[ConnerEvans]

I just realized that I think the only penalty that causes this bug is 'delay of game'. You might want to use str_detect(desc,"Delay of Game") in the fix.

[guga31bb]

In the most recent example, the problem is that posteam needs to be flipped since it's a kickoff, but there's no indicators for posteam so it's hard to do this programmatically. I'm not sure if there's a way to fix this other than hard coding a list of plays with wrong posteam on kickoffs with penalty.

[ConnerEvans]

I just ran the code below today: 3/7/21 9:30am PST and verified that there is still one more instance of this happening in the current data because it was a different penalty (Unnecessary Roughness).

pbp <- nflfastR::load_pbp(2011:2020, qs = TRUE)

pbp %>% filter(game_id == '2014_02_ATL_CIN', between(play_id, 3450, 3600)) %>% select(home_team, away_team, posteam, desc, ep, epa) %>% View

This is how I'm currently hardcoding a fix:

epapp_game <- pbp %>%
  filter(!is.na(posteam),!is.na(ep),quarter_end==0) %>%
  group_by(game_id,game_half) %>%
  
  mutate(
    epa = if_else(is.na(epa), -ep, epa),
    
    posteam = if_else((game_id == '2014_02_ATL_CIN')&(play_id==3498),home_team,posteam),
    kickoff_attempt = if_else((game_id == '2014_02_ATL_CIN')&(play_id==3498),1,kickoff_attempt),
    epa = if_else((game_id == '2014_02_ATL_CIN')&(play_id==3498),lead(ep)-ep,epa),
  )