NJD_Goalies.R

library(readr);library(tidyverse);library(MASS) # For Bayes
library(ggplot2);library(ggbeeswarm);library(forcats) # For Graphing
select=dplyr::select

df <- read_csv("EH_std_gl_stats_ev_regular_adj_2022-04-22.csv") #Devils Goalie Report from Evolving-Hockey.com
dfall <- read_csv("EH_std_gl_stats_ev_regular_adj_2022-04-22 (1).csv") #All NHL Goalie report from Evolving-Hockey.com

players = df %>% select(Player,xGA,GA) %>%
  mutate(GperXG = GA/xGA) 
temp = dfall %>% select(Player,xGA,GA) %>%
  mutate(GperXG = GA/xGA) 



# Fit Posterior Distributions
dumb_prior = fitdistr(temp %>% filter(GperXG>0) %>%`$`('GperXG'),"gamma")

player_names = players$Player
player_posts = data.frame()

for(player in player_names){

  xG = players %>% filter(Player==player) %>% select(xGA) %>% `$`("xGA") %>% sum()
  G = players %>% filter(Player==player) %>% select(GA)%>% `$`("GA") %>% sum()
  draws = data.frame(
    samples=c(rgamma(10000,
                     shape = dumb_prior$estimate[1] + G, 
                     rate = dumb_prior$estimate[2] + xG)))
  smart_prior = fitdistr(draws$samples,"gamma")
  
  player_posts = player_posts %>% bind_rows(
    data.frame(player = player,
               est1 = smart_prior$estimate[1],
               est2 = smart_prior$estimate[2])
  )
}


# Simulate results of distributions
sims = data.frame()
for(final_player in player_names){
  final_fit = player_posts %>% filter(player==final_player)
  
  sims = sims %>% bind_rows(
    data.frame(Player = rep(final_player,10000),
               simmed_GperxG = rgamma(10000,shape = final_fit$est1, rate = final_fit$est2))
  )
  
}


# Aggregate simulation
sims = sims %>% 
  mutate(good = ifelse(simmed_GperxG<0.78,1,0),
         avg = ifelse(simmed_GperxG<0.98,1,0),
         backup = ifelse(simmed_GperxG<1.37,1,0)) %>%
  group_by(Player) %>% 
  mutate(rank=mean(simmed_GperxG),
         grank=mean(good),
         arank=mean(avg),
         brank=mean(backup)) %>% ungroup

# Plot distribution
sims$Player=fct_reorder(sims$Player,sims$rank,max,.desc=T)
ggplot(sims, aes(x = simmed_GperxG, y = Player,color=simmed_GperxG)) +
  geom_quasirandom(groupOnX = FALSE) +
  labs(title = 'Devils Goalies in Goals Allowed per xG',
       y = NULL,
       x = "Goals per xG (10K simulations)") +
  theme(legend.position = "none")

# Elite
sims$Player=fct_reorder(sims$Player,sims$grank,max,.desc=F)
ggplot(sims, aes(x = simmed_GperxG, y = Player,color=factor(good))) +
  geom_quasirandom(groupOnX = FALSE) +
  labs(title = 'Devils Goalies in Goals Allowed per xG',
       y = NULL,
       x = "Goals per xG (10K simulations)")   +
  scale_color_manual(values = c("1" = "royalblue3",
                                "0"="darkgrey"))+
  theme(legend.position = "none")

# Starter
sims$Player=fct_reorder(sims$Player,sims$arank,max,.desc=F)
ggplot(sims, aes(x = simmed_GperxG, y = Player,color=factor(avg))) +
  geom_quasirandom(groupOnX = FALSE) +
  labs(title = 'Devils Goalies in Goals Allowed per xG',
       y = NULL,
       x = "Goals per xG (10K simulations)")   +
  scale_color_manual(values = c("1" = "royalblue3",
                                "0"="darkgrey"))+
  theme(legend.position = "none")

# Backup
sims$Player=fct_reorder(sims$Player,sims$brank,max,.desc=F)
ggplot(sims, aes(x = simmed_GperxG, y = Player,color=factor(backup))) +
  geom_quasirandom(groupOnX = FALSE) +
  labs(title = 'Devils Goalies in Goals Allowed per xG',
       y = NULL,
       x = "Goals per xG (10K simulations)")   +
  scale_color_manual(values = c("1" = "royalblue3",
                                "0"="darkgrey"))+
  theme(legend.position = "none")