4.5 Points previous vs current season

Let’s see if a team’s average points scored from the previous season is related to their points scored in the current season. First, we’ll have to rearrange the data a little since teams can appear in both the home and away column. Ideally, we would have one column with the team name, and one column with the score. Each game would then have two rows, one for the away team and one for the home team. Here is one way to do that:

Code

da = d %>% select(date, away, ascore, home, hscore, season, gid) %>% mutate(ha = 'away')
dh = d %>% select(date, home, hscore, away, ascore, season, gid) %>% mutate(ha = 'home')

colnames(da) = c('date', 'team', 'score',  'opp', 'opp.score', 'season', 'gid', 'ha')
colnames(dh) = c('date', 'team', 'score',  'opp', 'opp.score', 'season', 'gid', 'ha')

dd = bind_rows(da, dh) %>% 
  arrange(date, gid)

head(dd)

        date team score opp opp.score season      gid   ha
1 2020-12-22  GSW    99 BKN       125   2021 22000001 away
2 2020-12-22  BKN   125 GSW        99   2021 22000001 home
3 2020-12-22  LAC   116 LAL       109   2021 22000002 away
4 2020-12-22  LAL   109 LAC       116   2021 22000002 home
5 2020-12-23  MIL   121 BOS       122   2021 22000003 away
6 2020-12-23  BOS   122 MIL       121   2021 22000003 home

Note that, for example, the first two rows correspond to the first game, and contain the same information that was in the first row of the previous data frame.

Now we can compute average points scored by team for each season.

Code

ds = dd %>%
  group_by(team, season) %>%
  summarise(score = mean(score))

head(ds)

# A tibble: 6 × 3
# Groups:   team [3]
  team  season score
  <chr> <chr>  <dbl>
1 ATL   2021    114.
2 ATL   2022    114.
3 BKN   2021    119.
4 BKN   2022    113.
5 BOS   2021    113.
6 BOS   2022    112.

We now have two rows per team, one for each season. If we want a scatter plot, we’ll pivot_wider to have a column for each season. We don’t want column names that start with a number (we would have to use the tick marks `2022` all the time), so we’ll rename those too.

Code

ds = ds %>%
  pivot_wider(id_cols     = team, 
              names_from  = season, 
              values_from = score) %>%
  rename(s2021 = `2021`, 
         s2022 = `2022`)

head(ds)

# A tibble: 6 × 3
# Groups:   team [6]
  team  s2021 s2022
  <chr> <dbl> <dbl>
1 ATL    114.  114.
2 BKN    119.  113.
3 BOS    113.  112.
4 CHA    109.  115.
5 CHI    111.  112.
6 CLE    104.  108.

Now we can make a scatter plot.

Code

ggplot(ds, 
       aes(x = s2021, 
           y = s2022, 
           label = team))+
  geom_point()+
  geom_text(hjust = -.1)

Current and previous season performances are related, despite the fact that some players, coaches, and front office personnel change teams in the offseason. Also, note the correlation is about 0.55:

Code

cor(ds$s2021, 
    ds$s2022)

[1] 0.5504614