Figure 5

Types of over-represented accounts. 249 of the 1000 accounts with the most Tweets in the 1% Sample. Blue: Sample cheaters. Accounts with low entropy score and many days with ≥48 Tweets/day are defined as likely cheaters. Green: Corporate spammers. Accounts over-represented on many days but without suspicious entropy scores; all of these accounts can be linked to commercial activity. Yellow: In-syncs. Frequency bots that are temporarily synced with the sampling mechanism. Black: Lucky accounts that are randomly over-represented but show no suspicious characteristics