if(!require(ggbump)) devtools::install_github("davidsjoberg/ggbump")
if(!require(ggflags)) devtools::install_github("rensa/ggflags")
if(!require(pacman)) install.packages("pacman")
pacman::p_load(tidyverse, countrycode, janitor, padr, hablar, ggflags, ggbump, lubridate)

# Source: https://www.kaggle.com/heesoo37/120-years-of-olympic-history-athletes-and-results/version/2
df_raw <- read.csv("athlete_events.csv") %>% 
  clean_names() %>% 
  retype()

# Only include summer olympics and years after 1990
df <- df_raw %>% 
  filter(season == "Summer",
         year >= 1990)

# Summarise data set be unique on year, country and number of medals
df <- df %>% 
  group_by(year, team) %>% 
  summarise(n_medals = sum_(!is.na(medal))) %>% 
  ungroup()

# replace Unified Team with Russia (because I want to use a flag)
df <- df %>% 
  mutate(team = case_when(team == "Unified Team" ~ "Russia",
                          T ~ team))

# Create rank per olympic games
df <- df %>% 
  group_by(year) %>% 
  mutate(rank = rank(-n_medals, ties.method = "random")) %>% 
  ungroup()

# Create average rank per country
df <- df %>% 
  group_by(team) %>% 
  mutate(mean_rank = mean_(rank)) %>% 
  ungroup() 

# Only keep top 10 mean medal countries
df <- df %>% 
  filter(team %in% (df %>% 
           distinct(team, mean_rank) %>% 
           top_n(10, -mean_rank) %>% 
           pull(team)))

# Rerank the countries
# Create rank per olympic games
df <- df %>% 
  group_by(year) %>% 
  mutate(rank = rank(-n_medals, ties.method = "random")) %>% 
  ungroup()

# Remove unnecessary columns
df <- df %>% 
  select(-mean_rank)

country_2_letters <- countrycode(df$team %>% unique() %>% sort(),
            origin = "country.name",
            destination = "genc2c") %>% 
  tolower() %>% 
  set_names(df$team %>% unique() %>% sort())

df <- df %>% 
  mutate(team_2_letters = country_2_letters[team])

host_countries <-  tibble(
  team = c("United States",
              "Australia",
              "Greece",
              "China",
              "Great Britain",
              "Brazil"),
  year    = c(1996,
              2000,
              2004,
              2008,
              2012,
              2016
              )
) %>% 
  mutate(host = 1)

df <- df %>% 
  left_join(host_countries, by = c("team", "year"))

# Has the country been host in this time period
df <- df %>% 
  group_by(team) %>% 
  mutate(has_been_host = dummy_(any(host == 1, na.rm = T))) %>% 
  ungroup()

df %>% 
  ggplot(aes(year, rank, group = team, color = team, fill = team)) +
  geom_bump(aes(smooth = 10), size = 1.5, lineend = "round") +
  geom_label(data = df %>% filter(host == 1),
            aes(label = "Host")) +
  geom_text(data = df %>% filter(host == 1),
            aes(label = "Host"),
            color = "black") +
  geom_flag(data = df %>% filter(year == min(year)), 
            aes(country = team_2_letters),
            size = 8,
            color = "black") +
  geom_flag(data = df %>% filter(year == max(year)), 
            aes(country = team_2_letters),
            size = 8,
            color = "black") +
  scale_color_manual(values = c(wesanderson::wes_palette("GrandBudapest2"), wesanderson::wes_palette("GrandBudapest1"), wesanderson::wes_palette("BottleRocket2"))) +
  scale_fill_manual(values = c(wesanderson::wes_palette("GrandBudapest2"), wesanderson::wes_palette("GrandBudapest1"), wesanderson::wes_palette("BottleRocket2"))) +
  scale_y_reverse(breaks = 1:100) +
  scale_x_continuous(breaks = df$year %>% unique()) +
  theme_minimal() +
  theme(legend.position = "none",
        panel.grid = element_blank(),
        panel.background = element_rect(fill = "gray60", color = "transparent"),
        plot.background = element_rect(fill = "gray60"),
        text = element_text(color = "white")) +
  labs(x = NULL,
       y = NULL,
       title = "Hosting is boosting the number of medals",
       subtitle = "Number of medals on Summer Olympics 1992 - 2016")