NBA All Stars: the exploration continues

The All Stars

First lets take a look at our starters:

asg_starters %>%
  select(player, team_slug) %>%
  gt() %>%
  tab_header(
    title = md("**NBA All-Star Game Starters**"),
  ) %>%
  cols_label(player = md("**Player Name**"),
             team_slug = md("**Team**")) %>%
  tab_footnote(
    footnote = "Team captain",
    locations = cells_data(
      columns = vars(player),
      rows = c(1, 8))
  ) %>%
  tab_options(footnote.glyph = c("*, †, ‡"))

asg_reserves %>%
  select(player, team_slug) %>%
  gt() %>%
  tab_header(
    title = md("**NBA All-Star Game Reserves**"),
  ) %>%
  cols_label(player = md("**Player Name**"),
             team_slug = md("**Team**"))

We’ve also got a category I’m calling “specials.” Brooklyn Nets guard D’Angelo Russell was named by NBA Commissioner Adam Silver as the replacement for Victor Oladipo, who ruptured a quad tendon in a game against the Raptors on January 23.¹

Silver also added Dirk Nowitzki and Dwyane Wade as “special team roster additions” because, well, they’re Dirk Nowitzki and D. Wade.²

asg_special %>%
  select(player, team_slug) %>%
  gt() %>%
  tab_header(
    title = md("**NBA All-Star Game Special Selections**"),
  ) %>%
  cols_label(player = md("**Player Name**"),
             team_slug = md("**Team**"))

The Votes

We’ve got voting data, which I scraped and wrangled in an earlier missive.³

Munging data in

Let’s get some more data on our actual selected All-Star team. First I’ll combine the starters, reserves, and “specials” into a single data frame.

asg_players <- bind_rows(asg_starters, asg_reserves, asg_special)

Next will bring in the All-Star voting data Note that I’m using two columns for the join not because I necessarily need them, but to avoid name repair which (by default) would leave me with two variables, team_slug.x and team_slug.y, in the resulting data frame.

asg_data <- asg_players %>%
  left_join(asg_votes, by = c("player", "team_slug"))

I’m also gonna ditch those last four columns, since they were a mess of my own making from the initial data scraping, and I’m pretty sure they have variable names that will bork something up somewhere.

asg_data <- asg_data[,1:13]

Selections vs. Voted Rankings

Next, let’s take a look at the players who were “least deserving” of their selection according to player, fan, and media rankings. We’ll group by conf_court, since the voting is does that way, and, thus each combination of Eastern/Western conferences and front/back courts has their own ranking.

rank_vars <- c("player", "team_slug", "player_rank", "fan_rank", "media_rank", "weighted_score", "conf_court")

asg_data %>%
  group_by(conf_court) %>%
  filter(player_rank == max(player_rank)) %>%
  select(rank_vars) %>%
  gt() %>%
  tab_header(
    title = md("**Least Deserving: Player Rank**"),
  ) %>%
  cols_label(player = md("**Player Name**"),
             team_slug = md("**Team**"),
             fan_rank = md("**Fan Rank**"),
             player_rank = md("**Player Rank**"),
             media_rank = md("**Media Rank**"),
             weighted_score = md("**Weighted Score**")) %>%
  tab_style(
    style = cells_styles(
      bkgd_color = "lightpink"),
    locations = cells_data(
      columns = vars(player_rank)
      )
    ) %>%
  tab_source_note(md("source: [nba.com](http://www.nba.com/article/2019/01/24/2019-nba-all-star-starters-revealed-official-release)")) %>%
  tab_options(stub_group.font.weight = "600",
              sourcenote.padding = px(10))

Why so many in this last round? Well, the media happened to have quite a few ties. Don’t worry, they haven’t totally lost their minds. They’ve got Damian Lillard ranked number 4 for Western Conference Guards, tied with Klay Thompson. Klay’s an interesting case, because though he’s definitely got the fans and media behind him, he’s only ranked 11th by his peers.

It’s not surprising that we see Nowitzki in all three. Dude’s a legend, playing at 41, and this is a “sentimental” pick on a lot of levels. DLo (D’Angelo Russell) is another story, though. I’m not saying he doesn’t deserve it — he’s been great this season, but (if you take a look at the table above for Eastern Backcourt), by votes alone, the players, fans, and media all had several players lined up before him.

Player faves not selected

Based on the votes of their peers, who was low-key robbed?

We’ll grab the top seven players from each segment according to player rankings. Then we can see who among them was not selected to play in the All-Star Game.⁴

I like this little not in operator %ni%, from a post on Stack Overflow by Spencer Castro.⁵ So, I’ll use that to get my player list.

'%ni%' <- Negate('%in%')

player_top7 <- asg_votes %>%
  filter(player_rank <= 7) 

top7_nin_asg <- player_top7 %>%
  filter(player %ni% asg_players$player)

Looks like Jimmy Butler and Derrick Rose shoulda made the cut as far as their fellow players were concerned.

Lest you think everyone was in agreement about that, the media were a pretty consistent voting block, and, the way rankings were done, this means players with zero media votes all tied for… Well, let’s see:

asg_votes %>%
  group_by(conf_court) %>%
  filter(media_rank == max(media_rank)) %>%
  select(media_total_votes, media_rank, conf_court) %>%
  distinct() %>%
  gt() %>%
  tab_header(
    title = md("**Media All-Star No Votes and Rank**"),
    subtitle = glue::glue("NBA All-Star Voting 2019 Results")
    ) %>%
  cols_label(
    media_total_votes = md("**Media Votes**"),
    media_rank = md("**Media Rank**")
    ) %>%
    tab_source_note(md("source: [nba.com](http://www.nba.com/article/2019/01/24/2019-nba-all-star-starters-revealed-official-release)")) %>%
  tab_options(stub_group.font.weight = "600",
              sourcenote.padding = px(10))

So, yeah, Jimmy Butler was 7th in media rank…along with every other frontcourt player from the Eastern Conference who got zero media votes.

Team LeBron vs. Team Giannis

The conference captains, LeBron James and Giannis Antetokounmpo, drafted their teams on February 8th. Here’s how that shook out:

allstar_teams <- read_csv(here::here("data", "ASG", "allstar_teams.csv"))
## Parsed with column specification:
## cols(
##   draft_pick = col_double(),
##   player = col_character(),
##   allstar_team = col_character()
## )

allstar_teams %>%
  gt() %>%
  tab_header(
    title = md("**Draft Results: Team LeBron vs. Team Giannis**"),
    subtitle = glue::glue("2019 NBA All-Star Rosters")
  ) %>%
  cols_label(
    draft_pick = md("**Pick**"),
    player = md("**Player**"),
    allstar_team = md("**Team**")
  ) %>%
    tab_source_note(md("source: [nba.com](http://www.nba.com/allstar/2019/draft)")) %>%
    tab_footnote(
    footnote = "Traded after draft",
    locations = cells_data(
      columns = vars(player),
      rows = c(13, 16))
  ) %>%
  tab_options(footnote.glyph = c("*, †, ‡"))

Since there are no rules prohibiting trades (this is only the second year the NBA has drafted all-star teams rather than splitting them by conference), Giannis and LeBron went ahead and swapped Ben Simmons and Russell Westbrook after the draft.⁶

allstar_results <- read_csv(here::here("data", "ASG", "allstar_results.csv"))
## Parsed with column specification:
## cols(
##   player = col_character(),
##   allstar_team = col_character(),
##   starter = col_logical()
## )

allstar_results %>%
  mutate(status = case_when(
    starter == TRUE ~ "Starters",
    TRUE ~ "Reserves"
  )) %>%
  dplyr::select(player, allstar_team, status) %>%
  group_by(allstar_team, status) %>%
  gt() %>%
  tab_header(
    title = md("**2019 NBA All-Star Teams**")
  ) %>%
    tab_source_note(md("source: [nba.com](http://www.nba.com/allstar/2019/draft)")) %>%
  tab_options(
    stub_group.font.weight = "600"
  )

Since it’s pretty obvious that we’re looking at player names here, I wanted to get rid of player in the header. To do this, we can pass player to the rowname_col argument for gt(). However, in order to do so, you need to have at least one other variable in the table (otherwise it’s just rownames, and your table gets confused). So, I’ll bring the players’ NBA team slugs back into the mix

asg_player_teamslug <- asg_data %>%
  dplyr::select(player, team_slug)

allstar_results %>%
  left_join(asg_player_teamslug, by = "player") %>%
  mutate(status = case_when(
    starter == TRUE ~ "Starters",
    TRUE ~ "Reserves"
  )) %>%
  dplyr::select(player, allstar_team, status, team_slug) %>%
  group_by(allstar_team, status) %>%
  gt(rowname_col = "player") %>%
  tab_header(
    title = md("**2019 NBA All-Star Teams**")
  ) %>%
  cols_label(
    team_slug = md("**NBA Team**")
  ) %>%
  tab_source_note(md("source: [nba.com](http://www.nba.com/allstar/2019/draft)")) %>%
  tab_options(
    stub_group.font.weight = "600"
  )

We can also un-tidy our data to take a look at the teams side-by-side. I’ll also add a little flare to this table by giving the table a background colour in tab_options() — the font colour switches automatically, which is pretty cool.

allstar_spread <- allstar_results %>%
  tibble::rowid_to_column() %>%
  spread(allstar_team, player) %>%
  mutate(status = case_when(
    starter == TRUE ~ "Starters",
    TRUE ~ "Reserves"
  )) %>%
  dplyr::select(status, `Team Giannis`, `Team LeBron`)

team_giannis <- allstar_spread %>%
  filter(!is.na(`Team Giannis`)) %>%
  dplyr::select(status, `Team Giannis`)

team_lebron <- allstar_spread %>%
  filter(!is.na(`Team LeBron`)) %>%
  dplyr::select(status, `Team LeBron`)

spread_teams <- team_giannis %>%
  bind_cols(team_lebron) %>%
  select(-status1)

spread_teams %>%
  group_by(status) %>%
  gt() %>%
  tab_options(
    table.background.color = "midnightblue",
    column_labels.font.weight = "800",
    stub_group.font.weight = "600"
  )

Actual sportsball data

Now we’ll grab the the most recent advanced metrics from basketball reference using the {nbastatR} package by Alex Bresler. Note, running bref_players_stats() will assign the output data frames, dataBREFPlayerTotals and dataBREFPlayerAdvanced, to the environment, so we don’t need to do anything else.

bref_players_stats(seasons = 2019, tables = c("advanced", "totals"), widen = TRUE, assign_to_environment = TRUE)

Let’s take a quick gander at what these data look like for everyone in the league using the {skimr} package.⁷

skim(dataBREFPlayerTotals) %>% skimr::kable()

skimmed_list <- skim_to_list(dataBREFPlayerTotals)
numeric_skim <- skimmed_list$numeric
numeric_skim %>%
  select(one_of(c("variable", "n", "mean", "sd", "p0", "p25", "p50", "p75", "p100", "hist"))) %>%
  mutate(mean = as.numeric(mean),
         sd = as.numeric(sd)) %>%
  mutate_at(vars(starts_with("p")), as.numeric) %>%
  select(-p0)
## # A tibble: 27 x 9
##    variable          n      mean     sd   p25   p50   p75  p100 hist    
##    <chr>             <chr> <dbl>  <dbl> <dbl> <dbl> <dbl> <dbl> <chr>   
##  1 agePlayer         489    26.1   4.18    23    25    29    42 ▃▇▇▆▃▁▁▁
##  2 astTotals         489    85.2  93.9     16    54   120   537 ▇▃▂▁▁▁▁▁
##  3 blkTotals         489    17.2  21.4      3    10    23   145 ▇▂▁▁▁▁▁▁
##  4 countGames        489    36.9  18.3     22    43    53    59 ▃▂▂▂▂▃▅▇
##  5 countGamesStarted 489    17.4  20.9      0     6    35    59 ▇▂▁▁▁▁▁▂
##  6 drbTotals         489   120.  110.      33    97   176   594 ▇▅▃▂▁▁▁▁
##  7 fg2aTotals        489   199.  194.      42   142   289   925 ▇▅▂▂▁▁▁▁
##  8 fg2mTotals        489   103.  104.      20    71   148   502 ▇▅▂▁▁▁▁▁
##  9 fg3aTotals        489   110.  115.      14    72   174   733 ▇▃▂▁▁▁▁▁
## 10 fg3mTotals        489    39.1  43.4      3    24    64   274 ▇▃▂▁▁▁▁▁
## # … with 17 more rows

skim(dataBREFPlayerAdvanced) %>% skimr::kable()

skimmed_list <- skim_to_list(dataBREFPlayerAdvanced)
numeric_skim <- skimmed_list$numeric
numeric_skim %>%
  select(one_of(c("variable", "n", "mean", "sd", "p0", "p25", "p50", "p75", "p100", "hist"))) %>%
  mutate(mean = as.numeric(mean),
         sd = as.numeric(sd)) %>%
  mutate_at(vars(starts_with("p")), as.numeric) %>%
  select(-p0)
## # A tibble: 24 x 9
##    variable   n        mean      sd     p25     p50     p75    p100 hist   
##    <chr>      <chr>   <dbl>   <dbl>   <dbl>   <dbl>   <dbl>   <dbl> <chr>  
##  1 agePlayer  489    26.1     4.18   23      25     2.90e+1   42    ▃▇▇▆▃▁…
##  2 countGames 489    36.9    18.3    22      43     5.30e+1   59    ▃▂▂▂▂▃…
##  3 minutes    489   840.    610.    256     824     1.38e+3 2158    ▇▅▃▅▃▅…
##  4 pct3PRate  489     0.36    0.21    0.22    0.38  5.10e-1    0.91 ▆▃▇▇▆▅…
##  5 pctAST     489     0.13    0.096   0.068   0.11  1.80e-1    0.59 ▆▇▃▂▁▁…
##  6 pctBLK     489     0.17    0.27    0.013   0.023 1.50e-1    0.9  ▇▁▁▁▁▁…
##  7 pctDRB     489     0.15    0.084   0.1     0.14  1.90e-1    0.71 ▂▇▃▁▁▁…
##  8 pctFTRate  489     0.24    0.14    0.16    0.22  3.20e-1    0.83 ▃▇▇▃▂▁…
##  9 pctORB     489     0.075   0.14    0.019   0.035 7.50e-2    1    ▇▁▁▁▁▁…
## 10 pctSTL     489     0.13    0.27    0.012   0.017 2.50e-2    0.9  ▇▁▁▁▁▁…
## # … with 14 more rows

allstar_data <- asg_data %>%
  left_join(dataBREFPlayerAdvanced, by = c("player" = "namePlayer"))

skim(allstar_data) %>% skimr::kable()

skimmed_list <- skim_to_list(allstar_data)
numeric_skim <- skimmed_list$numeric
numeric_skim %>%
  select(one_of(c("variable", "mean", "sd", "p0", "p25", "p50", "p75", "p100", "hist"))) %>%
  mutate(mean = as.numeric(mean),
         sd = as.numeric(sd)) %>%
  mutate_at(vars(starts_with("p")), as.numeric) %>%
  select(-p0)
## # A tibble: 31 x 8
##    variable        mean       sd      p25       p50       p75    p100 hist 
##    <chr>          <dbl>    <dbl>    <dbl>     <dbl>     <dbl>   <dbl> <chr>
##  1 agePlayer    2.81e+1  4.38e+0  2.50e+1   2.80e+1   3.00e+1 4.00e+1 ▇▅▇▆…
##  2 countGames   5.08e+1  7.95e+0  4.65e+1   5.40e+1   5.65e+1 5.80e+1 ▁▁▁▁…
##  3 fan_rank     5.89e+0  4.35e+0  3.00e+0   5.00e+0   8.50e+0 1.50e+1 ▇▇▇▁…
##  4 fan_total_…  1.90e+6  1.41e+6  6.73e+5   1.45e+6   3.00e+6 4.62e+6 ▇▆▃▁…
##  5 media_rank   4.07e+0  2.53e+0  2.00e+0   4.00e+0   6.50e+0 8.00e+0 ▇▃▃▇…
##  6 media_tota…  3.68e+1  4.13e+1  1.00e+0   1.10e+1   7.75e+1 1.00e+2 ▇▁▁▁…
##  7 minutes      1.69e+3  3.77e+2  1.59e+3   1.77e+3   1.93e+3 2.16e+3 ▁▁▁▂…
##  8 pct3PRate    3.30e-1  1.60e-1  2.30e-1   3.40e-1   4.20e-1 6.50e-1 ▃▃▆▇…
##  9 pctAST       2.60e-1  1.00e-1  1.80e-1   2.60e-1   3.60e-1 4.70e-1 ▁▂▇▅…
## 10 pctBLK       1.20e-1  2.60e-1  1.30e-2   1.70e-2   4.10e-2 9.00e-1 ▇▁▁▁…
## # … with 21 more rows