Songs come from the Song A Day spreadsheet and catalog, cleaned and checked against YouTube. Each song's place was matched to coordinates and its day's weather fetched from Open-Meteo. Moon phase, planet positions, retrogrades and eclipses were computed for 5 pm New York time. Headlines are from Wikipedia's Current events portal, 99,951 of them. Views are from YouTube in October 2026.
67 questions were tested with shuffle tests that compare songs only with others from the same month, week or year, then corrected for how many questions were asked. Two reviewers re-ran every result that looked real with different methods. They kept the ones above and downgraded four: the moods don't really change between the cities he lived in, seasons don't shift the mood mix, the politics link is mostly election weeks, and “darker over time” turned out to be a jump in angry songs.
Everything is public: the cleaned data (one row per song, 121 columns), the code, every result and both reviews, at github.com/songadaymann/songaday-data.