Offense safety fix

[mrcaseb]

This should fix the score calculation for home_team and away_team in case the posteam scores a safety

[mrcaseb]

The following games have got wrong scores in the pbp data compared to Lee's games data.

Using the following code with this branch

library(dplyr)
wrong_scores <- c(
  "1999_04_PHI_NYG",
  "2000_03_ATL_CAR",
  "2000_11_OAK_DEN",
  "2000_16_OAK_SEA",
  "2001_14_DAL_SEA",
  "2003_03_NO_TEN",
  "2006_05_STL_GB",
  "2009_08_STL_DET",
  "2009_17_IND_BUF",
  "2011_14_TB_JAX",
  "2013_07_CIN_DET",
  "2015_06_ARI_PIT",
  "2015_09_PHI_DAL",
  "2015_15_KC_BAL",
  "2016_01_MIN_TEN",
  "2016_05_NE_CLE",
  "2017_08_LAC_NE",
  "2017_13_DEN_MIA"
  "2018_14_JAX_TEN",
  "2020_15_KC_NO" 
)

pbp <- fast_scraper(wrong_scores, pp = T)
games <- readRDS(url("https://github.com/leesharpe/nfldata/blob/master/data/games.rds?raw=true"))

scores <- pbp %>% 
  group_by(game_id) %>% 
  summarise(
    # alternatively use max instead of last
    home_score_pbp = last(total_home_score),
    away_score_pbp = last(total_away_score)
  ) %>% 
  left_join(games %>% select(game_id, home_score, away_score), by = "game_id") %>% 
  filter(home_score != home_score_pbp | away_score != away_score_pbp)

we get

It looks like the safety on offense problem is solved but there are still 10 problematic games, where the pbp score is much more off than just 2 points. This could be another problem, most likely unrelated to the offensive safeties.