Skip to contents

NBA Net Rating Quadrant

Saiem Gilani
@saiemgilani @saiemgilani

Source: vignettes/recipe-nba-net-rating-quadrant.Rmd
recipe-nba-net-rating-quadrant.Rmd
On this page

The brief: a season-review blog post needs one chart of where all 30 teams finished on offense and defense, split by conference, at 1600 px wide, plus a square version for Instagram. In ggplot2 each fix is one more layer: logos with geom_sdv_logos(), per-conference averages with geom_mean_lines(), and facets for the conferences. The ratings come from hoopR’s ESPN team box scores.

library(sdvplotR)
library(ggplot2)
library(dplyr, warn.conflicts = FALSE)

season <- 2026 # the 2025-26 season: NBA seasons are named by the year they end
label <- paste0(season - 1, "-", season %% 100)
out_dir <- tempfile("sdvplotR-recipe-") # where the exports go; use your own folder
dir.create(out_dir)

1. Get the data

Offensive and defensive rating are points scored and allowed per 100 possessions. Possessions are estimated from the box score (FGA - OREB + TOV + 0.44 FTA) and averaged with the opponent’s, so both teams in a game share one count.

box <- hoopR::load_nba_team_box(season) |>
  filter(season_type == 2) # regular season

games <- box |>
  mutate(
    poss = field_goals_attempted - offensive_rebounds + total_turnovers +
      0.44 * free_throws_attempted
  )
filter(games, is.na(poss)) |>
  select(game_id, game_date, team_abbreviation, opponent_team_abbreviation)
#> # A tibble: 1 × 4
#>     game_id game_date  team_abbreviation opponent_team_abbreviation
#>       <int> <date>     <chr>             <chr>                     
#> 1 401810169 2025-12-01 CHI               ORL

One box score is missing a stat, so that game has no possession count. An NA inside sum() would turn a whole season into NA, so the game is dropped, for both teams together: after the join both rows of that game have no shared count.

opponents <- select(games, game_id, opponent_team_id = team_id, opp_poss = poss)
stopifnot(identical(class(games$opponent_team_id), class(opponents$opponent_team_id)))

ratings <- games |>
  inner_join(opponents, by = c("game_id", "opponent_team_id")) |>
  mutate(game_poss = (poss + opp_poss) / 2) |>
  filter(!is.na(game_poss)) |>
  group_by(team = team_abbreviation) |>
  summarise(
    ortg = 100 * sum(team_score) / sum(game_poss),
    drtg = 100 * sum(opponent_team_score) / sum(game_poss)
  ) |>
  mutate(net = ortg - drtg)
stopifnot(!anyNA(ratings$net))
nrow(ratings)
#> [1] 33

33 teams in a 30-team league. sdv_resolve() finds the extras: anything that is not an NBA team comes back NA (with a warning naming the values). They are the All-Star Game’s teams, which ESPN files as regular-season games. The conference comes from sdv_teams(), keyed by the same team id.

ratings <- ratings |>
  mutate(team_id = sdv_resolve(team, league = "nba"))
filter(ratings, is.na(team_id))$team
#> [1] "STARS"   "STRIPES" "WORLD"

conferences <- sdv_teams("nba") |>
  transmute(team_id, team_abbr = abbr, conference = sub("ern Conference$", "", conference))
ratings <- ratings |>
  filter(!is.na(team_id)) |>
  inner_join(conferences, by = "team_id") |>
  arrange(desc(net), team_abbr)
stopifnot(nrow(ratings) == 30)
head(ratings)
#> # A tibble: 6 × 7
#>   team   ortg  drtg   net team_id team_abbr conference
#>   <chr> <dbl> <dbl> <dbl> <chr>   <chr>     <chr>     
#> 1 OKC    116.  105. 10.9  25      OKC       West      
#> 2 DET    114.  106.  7.93 8       DET       East      
#> 3 SA     117.  109.  7.86 24      SA        West      
#> 4 BOS    117.  109.  7.85 2       BOS       East      
#> 5 NY     117.  110.  6.40 18      NY        East      
#> 6 HOU    115.  110.  5.21 10      HOU       West

2. The first draft

Points and abbreviations, ggplot2’s defaults.

ggplot(ratings, aes(ortg, drtg, label = team_abbr)) +
  geom_point() +
  geom_text(nudge_y = 0.4, size = 3)

First draft: a plain scatter of NBA teams' offensive rating against defensive rating, labelled with overlapping abbreviations.

Three problems: the labels overlap, a low defensive rating is good but sits at the bottom, and the two conferences (which only meet a third of the time) are mixed together.

3. Logos, one panel per conference, and average lines

geom_sdv_logos() replaces the points, facet_wrap() splits the conferences, and geom_mean_lines() draws each panel’s own average offense and defense, so every team is read against its conference.

ggplot(ratings, aes(ortg, drtg)) +
  geom_mean_lines(aes(x0 = ortg, y0 = drtg), color = "grey60", linewidth = 0.5) +
  geom_sdv_logos(aes(team = team_abbr), league = "nba", width = 0.1) +
  facet_wrap(vars(conference))

NBA teams drawn as logos in two panels, East and West, with dashed lines at each conference's average offensive and defensive rating.

Better, but the best defenses are at the bottom, logos at the edges are cut off, and the axis titles are column names.

4. Point the axes the right way and add net-rating guides

scale_y_reverse() puts the best defenses on top, so up and to the right is good on both axes, and expansion() on both scales leaves room for the edge logos. Net rating is the gap between the two numbers, so teams with the same net rating sit on a diagonal: faint lines at -10, -5, 0, +5 and +10 show it without a third axis. Each line is cut to the data’s range and drawn as data (geom_segment()), so the reversed scale moves it with the logos; a data frame with no conference column repeats in both panels.

x_lim <- range(ratings$ortg) + c(-1, 1)
y_lim <- range(ratings$drtg) + c(-1, 1)
# the line drtg = ortg - net, cut to the box x_lim by y_lim
guides <- data.frame(net = c(-10, -5, 0, 5, 10)) |>
  mutate(
    x0 = pmax(x_lim[1], y_lim[1] + net),
    x1 = pmin(x_lim[2], y_lim[2] + net)
  ) |>
  filter(x0 < x1) |>
  mutate(
    y0 = x0 - net, y1 = x1 - net,
    lx = x0 + 0.88 * (x1 - x0), ly = lx - net, # a label near the lower-right end
    label = if_else(net > 0, paste0("+", net), as.character(net))
  )

net_guides <- list(
  geom_segment(
    aes(x = x0, y = y0, xend = x1, yend = y1),
    data = guides, inherit.aes = FALSE, color = "grey88", linewidth = 0.4
  ),
  geom_text(
    aes(x = lx, y = ly, label = label),
    data = guides, inherit.aes = FALSE, color = "grey60", size = 2.6
  ),
  geom_mean_lines(aes(x0 = ortg, y0 = drtg), color = "grey60", linewidth = 0.5)
)

p <- ggplot(ratings, aes(ortg, drtg)) +
  net_guides +
  geom_sdv_logos(aes(team = team_abbr), league = "nba", width = 0.1) +
  facet_wrap(vars(conference)) +
  scale_x_continuous(expand = expansion(mult = 0.06)) +
  scale_y_reverse(expand = expansion(mult = 0.08)) +
  labs(
    x = "Offensive rating (points per 100 possessions)",
    y = "Defensive rating (allowed per 100)"
  )
p

The two-panel logo chart with the defensive axis reversed and faint diagonal lines marking net ratings of -10, -5, 0, +5 and +10.

5. Polish and tell the story

A theme strips the chart junk (minor grid, panel background) and sets the type; the strip titles become plain bold labels on the left. The title states the finding, the subtitle explains the reading, and the caption carries the source. The two conference leaders get their net rating printed beside the logo.

leaders <- ratings |>
  group_by(conference) |>
  slice_max(net, n = 1, with_ties = FALSE) |>
  ungroup() |>
  mutate(note = sprintf("%s %+.1f", team_abbr, net))
top <- slice_max(leaders, net, n = 1)
east <- filter(leaders, conference == "East")

title <- if (top$conference == "West") {
  sprintf("%s stood alone at %+.1f; %s led the East", top$team_abbr, top$net, east$team_abbr)
} else {
  sprintf("%s led the league at %+.1f", top$team_abbr, top$net)
}

blog_theme <- theme_minimal(base_size = 10) +
  theme(
    plot.title.position = "plot",
    plot.title = ggtext::element_textbox_simple(face = "bold", size = 14, margin = margin(b = 4)),
    plot.subtitle = ggtext::element_textbox_simple(color = "grey40", size = 9, margin = margin(b = 8)),
    plot.caption = element_text(color = "grey45", size = 7.5),
    strip.text = element_text(face = "bold", size = 11, hjust = 0),
    axis.title = element_text(color = "grey40", size = 9),
    axis.text = element_text(color = "grey55"),
    panel.grid.minor = element_blank(),
    panel.grid.major = element_line(color = "grey94"),
    plot.margin = margin(12, 14, 8, 12),
    plot.background = element_rect(fill = "white", color = NA)
  )

net_chart <- function(logo_width = 0.1, ncol = 2) {
  ggplot(ratings, aes(ortg, drtg)) +
    net_guides +
    geom_sdv_logos(aes(team = team_abbr), league = "nba", width = logo_width) +
    geom_text(
      aes(label = note), data = leaders,
      hjust = 0, nudge_x = 1.1, size = 3, fontface = "bold", color = "grey20"
    ) +
    facet_wrap(vars(conference), ncol = ncol) +
    scale_x_continuous(expand = expansion(mult = 0.06)) +
    scale_y_reverse(expand = expansion(mult = 0.08)) +
    labs(
      title = title,
      subtitle = paste0(
        "Points scored and allowed per 100 possessions, ", label,
        " regular season; up and right is better. Dashed lines: each conference's ",
        "average. Diagonals: net rating, in steps of 5."
      ),
      x = "Offensive rating (points per 100 possessions)",
      y = "Defensive rating (allowed per 100)",
      caption = "Data: ESPN via hoopR  |  Chart: sdvplotR"
    ) +
    blog_theme
}

net_chart()

Final two-panel chart of NBA offensive and defensive ratings by conference, with net-rating diagonals, conference average lines, and each conference leader's net rating labelled.

6. Export for the blog and for Instagram

ggsave() takes the size in inches and a dpi: 8 x 4.5 in at 200 dpi is the blog’s 1600 x 900. The square post stacks the conferences (ncol = 1). Each panel is then about twice as wide as before, and width is a share of the panel’s width, so the square gets about half the share to keep the logos the same size.

blog <- file.path(out_dir, "nba_net_rating_1600x900.png")
ggsave(blog, net_chart(), width = 8, height = 4.5, dpi = 200)

square <- file.path(out_dir, "nba_net_rating_1080x1080.png")
ggsave(square, net_chart(logo_width = 0.055, ncol = 1), width = 7.2, height = 7.2, dpi = 150)

magick::image_info(magick::image_read(c(blog, square)))[c("width", "height")]
#> # A tibble: 2 × 2
#>   width height
#>   <int>  <int>
#> 1  1600    900
#> 2  1080   1080
magick::image_read(blog)

The finished NBA net rating chart exported at 1600 by 900 pixels for the blog.

The square cut, stacked:

magick::image_read(square)

The same chart with the conferences stacked, exported at 1080 by 1080 pixels for Instagram.

Our Authors

Citation

To cite the sdvplotR R package in publications, use:

BibTeX Citation

@misc{gilani_carl_lee_kay_weatherman_sdvplotR,
  author = {Gilani, Saiem and Carl, Sebastian and Lee, Jared and Kay, Camden and Weatherman, Andrew},
  title = {sdvplotR: The SportsDataverse's R Package for Multi-League Sports Plotting.},
  url = {https://sdvplotR.sportsdataverse.org},
  year = {2026}
}