This is a pretty cool data set. There are quite a few things to look into with a number of complete numeric fields which can be rare.
I decided to fit a model to predict the HP level of each PokΓ©mon based on speed, attack, defence, base experience, height, and weight.
I then found the top 16 PokΓ©mon that have the largest positive residual – those with a lot more HP than expected.
π Github

# https://github.com/rfordatascience/tidytuesday
library(tidyverse)
library(showtext)
library(patchwork)
library(janitor)
library(glue)
library(ggtext)
library(ggpath)
library(ggforce)
library(ggrepel)
# πΎ load data ---------------------------------------------------------------
pokemon_df <- readr::read_csv('https://raw.githubusercontent.com/rfordatascience/tidytuesday/main/data/2025/2025-04-01/pokemon_df.csv') |>
mutate(
pokemon = str_to_title(pokemon)
)
# βοΈ fonts and palettes ------------------------------------------------------
font_add("fa-brands", regular = "../../Assets/Fonts/fontawesome/webfonts/fa-brands-400.ttf")
font_add("fa-solid", regular = "../../Assets/Fonts/fontawesome/webfonts/fa-solid-900.ttf")
font_add("pokemon", regular = "assets/fonts/pokemon/Pokemon solid.ttf")
txt <- "grey20"
bg <- "white"
accent <- txt
pal <- c('#d34d60', '#e9e44c', '#68ae29', '#7abfb4', '#556eaf', '#847a9e', '#f6c8d5')
font_add_google("Raleway", "raleway", regular.wt = 200)
ft <- "raleway"
showtext_auto()
# π model -------------------------------------------------------------------
mod <- glm(hp ~ height + weight + base_experience + defense + attack + speed, data = pokemon_df)
summary(mod)
# π€Ό wrangle -----------------------------------------------------------------
df_mod <- tibble(
id = pokemon_df$id,
y = pokemon_df$hp,
y_hat = fitted.values(mod),
res = residuals(mod)
) |>
filter(!id %in% c(10120, 10057, 10026, 10056)) |>
mutate(
rank = n() - rank(res) + 1,
top_n = rank <= 16
)
df <- pokemon_df |>
select(id, pokemon, height, weight, base_experience, hp, attack, defense, speed) |>
pivot_longer(-c(id, pokemon), names_to = "var", values_to = "val") |>
group_by(var) |>
mutate(val_scaled = min_max(val, 0, 1)) |>
inner_join(df_mod, join_by(id)) |>
filter(top_n)
df_url <- pokemon_df |>
inner_join(df_mod, join_by(id)) |>
filter(top_n)
# π‘ text --------------------------------------------------------------------
caption <- make_caption(accent, bg)
title_url <- "https://upload.wikimedia.org/wikipedia/commons/thumb/9/98/International_Pok%C3%A9mon_logo.svg/1200px-International_Pok%C3%A9mon_logo.svg.png"
subtitle <- "A model was fit to estimate the Pokemon's HP based on height, weight, attack, defense, speed, base experience.
The 16 Pokemon below are those that have the largest positive difference between their actual and expected HP levels (i.e. the largest positive residual)."
# π plot --------------------------------------------------------------------
plt_point <- df_mod |>
ggplot(aes(y_hat, y)) +
geom_mark_hull(aes(fill = top_n), filter(df_mod, top_n), fill = pal[2], colour = pal[5], linewidth = 0.5) +
geom_point(alpha = 0.2, size = 3, colour = pal[5]) +
geom_smooth(method = "lm", se = FALSE, colour = pal[5]) +
geom_text_repel(aes(label = pokemon), filter(df_url, top_n), family = ft, colour = pal[5], size = 12) +
coord_cartesian(clip = "off") +
labs(
x = "Predicted HP",
y = "HP"
) +
theme_void() +
theme(
text = element_text(family = ft, size = 48, lineheight = 0.3, colour = txt),
plot.background = element_rect(fill = bg, colour = bg),
plot.title = element_text(size = 128, hjust = 0.5),
plot.subtitle = element_text(hjust = 0.5),
plot.caption = element_markdown(colour = txt, hjust = 0.5, margin = margin(t = 20)),
plot.margin = margin(b = 20, t = 50, r = 50, l = 30),
panel.spacing.y = unit(3, "cm"),
panel.spacing.x = unit(2, "cm"),
legend.position = "bottom",
axis.text = element_text(margin = margin(t = 10)),
axis.line = element_line(linewidth = 0.2),
axis.ticks = element_line(linewidth = 0.2),
axis.ticks.length = unit(0.25, "cm"),
axis.title = element_text()
)
plt_title <- ggplot() +
annotate("from_path", x = 1, y = 0.75, path = title_url, width = 0.8) +
annotate("text", x = 1, y = 0.5, label = str_wrap(subtitle, 50), family = ft, colour = txt, size = 16, vjust = 1, lineheight = 0.3) +
coord_cartesian(clip = "off") +
theme_void() +
xlim(0, 2) +
ylim(-1, 1)
plt_title +
inset_element(plt_point, left = 0, top = 0.6, bottom = 0.1, right = 1)
ggsave("scripts/2025/13 - pokemon/pokemon-title.png", height = 12, width = 6)
# π plot --------------------------------------------------------------------
df |>
mutate(var = snakecase::to_title_case(var)) |>
ggplot() +
geom_col(aes(var, 1.5), fill = "grey95") +
geom_col(aes(var, val_scaled+0.5, fill = var)) +
geom_text(aes(var, val_scaled+0.52, label = val), family = ft, colour = txt, size = 12, hjust = 0) +
geom_point(aes("Height", 0.01), df_url, size = 56, colour = txt) +
geom_point(aes("Height", 0), df_url, size = 56, colour = "white") +
geom_from_path(aes("Height", 0, path = url_image), df_url, width = 0.7) +
facet_wrap(~pokemon) +
scale_fill_manual(values = rev(pal)) +
labs(
caption = caption,
fill = "Attribute"
) +
coord_flip(clip = "off") +
guides(fill = guide_legend(nrow = 1)) +
theme_void() +
theme(
text = element_text(family = ft, size = 48, lineheight = 0.3, colour = txt),
plot.background = element_rect(fill = bg, colour = bg),
plot.title = element_text(size = 128, hjust = 0.5),
plot.subtitle = element_text(hjust = 0.5),
plot.caption = element_markdown(colour = txt, hjust = 0.5, margin = margin(t = 20)),
plot.margin = margin(b = 20, t = 50, r = 50, l = 100),
panel.spacing.y = unit(3, "cm"),
panel.spacing.x = unit(2, "cm"),
legend.position = "bottom",
legend.margin = margin(t = 20),
strip.text = element_text(family = "pokemon", margin = margin(b = 10), size = 64)
)
ggsave("scripts/2025/13 - pokemon/pokemon.png", height = 12, width = 20)
title <- image_read("scripts/2025/13 - pokemon/pokemon-title.png")
body <- image_read("scripts/2025/13 - pokemon/pokemon-stats.png")
image_append(c(title, body)) |>
image_write("scripts/2025/13 - pokemon/pokemon.png")

