Analyzing Twitch data using the twitchETL package

Code base for twitchETL and this example can be found on GitHub

Retrieving top games using the twitchETL package:

require(twitchETL)
require(dplyr)
require(readr)
library(httr)

# get access token
clientID <- 'xxx'
clientSecret <- 'xxx'
r <- POST(paste0("https://id.twitch.tv/oauth2/token?client_id=", clientID, "&client_secret=", clientSecret, "&grant_type=client_credentials"))
stop_for_status(r)
access_token <- content(r, "parsed", "application/json")$access_token

query_timestamp <- Sys.time() %>% as.character

topGames <- getTopGameIDs(clientID, access_token, 10)
gameIDs <- topGames$game_id %>% as.character

topGamesStreams <- lapply(gameIDs, function(g) {
      Sys.sleep(2)
      return(getCurrentStreams(clientID, access_token, game_id = g))
  }) %>% bind_rows

df <- topGamesStreams %>%
  left_join(topGames, by = 'game_id')

# write to CSV
df %>%
  write_csv(paste0('~/twitch_snapshot_', Sys.time(), '.csv'))

# OR write to database
db <- RMySQL::dbConnect(RMySQL::MySQL(), user = "xxx", password = 'xxx', dbname = "xxx", host = "localhost")
RMySQL::dbWriteTable(db, value = df %>% mutate(query_timestamp = query_timestamp), name = "snapshots", append = TRUE, row.names = F)
dbDisconnect(db)

A brief analysis of Twitch data

options(warn = -1) 

require(RMySQL)
require(dplyr)
require(ggplot2)
require(ggdark)
require(ggiraph)
require(stringr)
require(lubridate)
require(highcharter)

# db_permissions.Rdata should contain the following environment variables:
## host: 'xxx.xxx.xx.xx'
## dbname: 'abc'
## user: def'
## password: 'ghi'

#save(host, dbname, user, password, file = 'db_permissions.Rdata')
load('db_permissions.Rdata') # permissions loaded from a local SQL database

conn <- dbConnect(MySQL(), host = host, dbname = 'twitch',  user = user, password = password)
#dbListTables(conn)

# Aggregate ranking of games as of latest pull of all time
df <- tbl(conn, 'snapshots') %>%
   group_by(game_name) %>%
   summarise(viewer_count = sum(viewer_count)) %>% 
   collect

# Daily snapshots of average views
df2 <- tbl(conn, 'snapshots') %>%
  collect %>%
  mutate(Period = ceiling_date(as_datetime(pull_timestamp), unit = 'days')) %>%
  group_by(game_name, Period) %>%
  summarise(viewer_count = mean(viewer_count))
DT::datatable(df)

g <- ggplot(df, aes(x = reorder(game_name, viewer_count), y = viewer_count)) + 
  geom_bar_interactive(stat = 'identity', aes(tooltip = viewer_count, data_id = str_replace_all(game_name, "'", ""))) + 
  xlab('') + ylab('Total Viewer Count') + 
  labs(caption = 'Source: databasement.org; Twitch API') +
  ggtitle('Most viewed games on Twitch since May') +
  coord_flip() + 
  dark_theme_minimal() + 
  theme(text = element_text(size = 9), 
        plot.title.position = 'plot',
        plot.caption.position = 'plot',
        plot.caption = element_text(size = 6, hjust = 0)) 
ggiraph::ggiraph(ggobj = g, height_svg = 8)

g <- ggplot(df2 %>% filter(Period >= '2020-05-15'), aes(x = Period, y = reorder(game_name, viewer_count))) + 
  geom_tile_interactive(aes(fill = log(viewer_count), tooltip = viewer_count, data_id = Period)) + 
  xlab('') + ylab('Average views per day') + 
  labs(caption = 'Source: databasement.org; Twitch API') +
  ggtitle('Twitch views over time') +
  dark_theme_minimal() + 
  theme(text = element_text(size = 9), 
        plot.title.position = 'plot',
        plot.caption.position = 'plot',
        plot.caption = element_text(size = 6, hjust = 0)) +
  scale_fill_viridis_c_interactive('total views per day', option = 'inferno') +
  theme(legend.position = 'hidden')
gg <- ggiraph::ggiraph(ggobj = g, height_svg = 12)
gg <- girafe_options(gg, opts_toolbar(position = 'top'))
gg