Scavenger AI reveals Eurovision data patterns
Scavenger AI has analyzed 68 years of Eurovision Song Contest data. The analysis reveals statistical patterns in winning entries' performance slots, song languages, and lyrics.
Scavenger AI, a data analytics firm, has released a deep dive into 68 years of the Eurovision Song Contest. Examining over 51,000 voting lines and 1,700 song lyrics, the findings highlight surprising patterns in what contributes to success in the competition.
The analysis indicates that fan predictions are not always accurate. While community favorites like the MyESC Community performed better, placing all recent winners within the top seven, the OGAE fan club votes were more inconsistent. The MyESC Community's larger membership base, over 10,000 members, combined with more data, appears to explain its superior predictive capability compared to OGAE's group of approximately 900 members.
Strategically, performance order significantly impacts outcomes. Data shows that early entries, particularly those in the first five slots, have not won in recent years. Instead, two specific windows, slots 11–15 and 21–26, have produced seven of the last eight winners, with average scores of 213 and 216 points respectively. However, the intervening slot 16–20 shows a notable dip and has not produced a single victory.
Language use has also seen a shift. Following the 1999 rule change that removed the requirement to sing in one's native language, English-language songs saw a surge. However, Portugal's win in 2017 with Salvador Sobral's Portuguese song marked a reversal. Since then, the proportion of native language songs has steadily increased, with four of the last eight winning songs performed in their country's official language.
Lyrical analysis reveals that winning songs tend to have a more positive tone and include words like "forever," while "La La La" appears more frequently in last-place entries. The gender split among performers is relatively even, but solo artists statistically outperform groups. Scavenger AI suggests an optimal winning profile might involve a native language song, performed by a solo artist in a performance slot between 11–15 or 21–26, with positive lyrical content.