R para visualizar datos deportivos espaciales

devtools::install_github("statsbomb/StatsBombR")
library(StatsBombR)

competitions <- FreeCompetitions() 
liga_spain <- competitions %>% filter(competition_id == 11)
range(liga_spain$season_name)

games <- FreeMatches(liga_spain)
glimpse(games)
unique(paste0(games$home_team.home_team_name, "-", 
              games$away_team.away_team_name))
barcelona_games <- games %>% 
                   filter("Barcelona" %in% c(home_team.home_team_name, 
                                             away_team.away_team_name))

barcelona_events <- StatsBombFreeEvents(MatchesDF = barcelona_games)
library(readr)
#write_rds(barcelona_events, "barcelona_events.rds")
#barcelona_events <- read_rds("barcelona_events.rds")
messi_events <- barcelona_events %>% 
                filter(str_detect(player.name, "Messi"))

glimpse(messi_events)
table(messi_events$type.name)

messi_shots <- messi_events %>% 
               filter(type.name == "Shot")

messi_shots_cleaned <- allclean(messi_shots)

data <- messi_shots_cleaned %>% 
        left_join(liga_spain %>% select(season_id, season_name)) %>%
        select(match_id, period, timestamp, type.name, possession_team.name, play_pattern.name,
               player.name, position.name, shot.type.name, location.x, location.y,
               shot.body_part.name, shot.technique.name, shot.statsbomb_xg, 
               shot.outcome.name, shot.end_location.x, shot.end_location.y, season_name)

devtools::install_github("boxuancui/DataExplorer")
library(DataExplorer)
plot_bar(data %>% select(shot.type.name, shot.outcome.name), ncol = 2)
plot_density(data %>% select(location.x, location.y), ncol = 1)

Paquetes: {dplyr}{tidyr}{lubridate}{forcats}{ggplot2}{DataExplorer}

Coordenadas según proveedor

data <- data %>%
        mutate_at(vars(contains('.x')), funs( . / 120 * 110)) %>% 
        mutate_at(vars(contains('.y')), funs( (80 - .) / 80 * 73))

data_free_kick <- data %>%
        filter(shot.type.name == "Free Kick")

• Sugerencias:
  
• Invertir coordenadas Y
• Transformar de yardas a metros: queda en 110x73 [m]. Otra opción es cancha estándar promedio 105x68 [m]
• No hay que invertir las coordenadas X según Period (siempre considera la misma dirección de ataque)
• Trabajar con grupos de tiros separados según "type.name"

library(jpeg)
img <- jpeg::readJPEG("image_path")

library(grid)
p <- rasterGrob(img, width = unit(1, "npc"), height = unit(1, "npc"))

library(ggplot2)
g <- ggplot() +
        annotation_custom(p, xmin = -0.5, xmax = 110.5, 
                             ymin = -0.5, ymax = 73.5) +
        scale_x_continuous(limits = c(-0.5, 110.5), expand = c(0,0)) +
        scale_y_continuous(limits = c(-0.5, 73.5), expand = c(0,0))

g + geom_point(data = data_free_kick, 
               mapping = aes(x = location.x, y = location.y), 
               pch = 21, alpha = 0.2, fill = "red", size = 2)

library(png)
img2 <- png::readPNG("image2_path")
p2 <- rasterGrob(img2, width = unit(1, "npc"), height = unit(1, "npc"))
g2 <- ggplot() +
      annotation_custom(p2, xmin = - Inf, xmax = Inf, ymin = - Inf, ymax = Inf) +
      scale_x_continuous(limits = c(-15, 15), expand = c(0,0)) +
      scale_y_continuous(limits = c(7, 30.2), expand = c(0,0))
g2

library(ggforce)
# función para dibujar la mitad ofensiva de la cancha invertida
get_pitch <- function(gp, pitch_fill = "white", pitch_col = "lightgrey"){
        contorno_df <- data.frame(x = c(55, 55, 110, 110), 
                                  xend = c(55, 110, 110, 55),
                                  y = c(0, 73, 73, 0),
                                  yend = c(73, 73, 0, 0))

        areas_df <- data.frame(x = c(110, 93, 93, 110, 104, 104), 
                               xend = c(93, 93, 110, 104, 104, 110), 
                               y = c(16, 16, 57, 27, 27, 46),
                               yend = c(16, 57, 57, 27, 46, 46))
        gp +
                theme_void() +
                theme(panel.background = element_rect(fill = pitch_fill),
                      plot.margin = unit(c(0.3, 0.3, 0.3, 0.3), "cm")) +
                # rectángulos
                geom_segment(data = contorno_df,
                             aes(x = x, xend = xend, y = y, yend = yend), 
                             col = pitch_col) +
                geom_segment(data = areas_df,
                             aes(x = x, xend = xend, y = y, yend = yend), 
                             col = pitch_col) +
                geom_rect(aes(xmin = 110, xmax = 111.5, ymin = 33, ymax = 40), 
                          fill = "transparent", col = pitch_col) +
                # puntos
                geom_point(aes(x = 110/2, y = 73/2), col = pitch_col) +
                geom_point(aes(x = 98, y = 73/2), col = pitch_col) +
                # semi círculos
                geom_arc(aes(x0 = 98, y0 = 73/2, r = 9.15, start = -34*pi/180, 
                             end = -146*pi/180), col = pitch_col) + #sigue...
}

                geom_arc(aes(x0 = 110/2, y0 = 73/2, r = 9.15, start = 0, 
                             end = 180*pi/180), col = pitch_col) +
                geom_arc(aes(x0 = 110, y0 = 73, r = 1, start = 270*pi/180, 
                             end = 360*pi/180), col = pitch_col) +
                geom_arc(aes(x0 = 110, y0 = 0, r = 1, start = 180*pi/180, 
                             end = 270*pi/180), col = pitch_col) +
                coord_flip() +
                scale_y_reverse()
}
get_pitch(gp = ggplot(), 
          pitch_fill = "white", pitch_col = "black") +
geom_point(data = data_free_kick, 
           mapping = aes(x = location.x, y = location.y), 
           alpha = 0.3, col = "darkblue", size = 2)

library(remotes)
install_github("Dato-Futbol/soccerAnimate")

• Repositorio con tutorial