Do you remember movies from the early 2000s as blue?
Not every movie, and not every scene. But the memory is easy to find: the midnight blues of Underworld, the cold disaster light of The Day After Tomorrow, or the pale, cool city in Minority Report. Put on a few thrillers or science-fiction films from that period and the impression can feel obvious.
Maybe that look was real. Maybe we remember the films that made the strongest impression, or the scenes we watched most often. Maybe the period feels blue because we connect it with rain, night, screens, and cold futures. The interesting question is not whether we can find a blue frame. It is whether movies from the early 2000s were, on average, more blue than movies from other periods.
A feeling needs a fair comparison
A movie can contain thousands of frames. Those frames are not thousands of separate movies. They share a colour grade, a cinematographer, locations, lighting choices, and the same story.
For the Blue Movies project, I selected the blue-ratio measure: blue intensity divided by the red, green, and blue intensities together. I average that measure across pixels and frames to get one mean blue ratio per movie, then compare those movie-level summaries across periods. This keeps extra frames useful for estimating one film without letting one film count as hundreds of independent movies, while leaving a separate question: whether the selected films represent the whole era.
The spreadsheet can fool you
Suppose I take 100 frames from each of ten films. The spreadsheet has 1,000 rows. The sample still has ten films.
More frames can help me describe a film. They do not give me the same evidence as more independently selected films. If one movie contributes more frames than another, it should not quietly get more influence just because it produced more rows.
That is the small trap behind a large-looking dataset. When a cultural memory feels obvious, the hard part is not finding supporting examples. It is deciding what should count as one example in the first place.