[OC] Der „Schwerkraftbrunnen 2003“: Die Aufzeichnung von 126.868 trivialen Vermutungen zeigt, dass das menschliche Gedächtnis die gesamte Musikgeschichte systematisch in Richtung der frühen 2000er Jahre komprimiert

**Data Source & Methodology:**

Data comes from 126,868 guesses across 115 songs on [YearToBeat.com](https://yeartobeat.com), a daily tool I built to test music memory. Players watch a music video and guess the exact release year. Visualized using Python/Matplotlib.

**The Insight: The Gravity Well**

I originally thought people would just be randomly wrong. But the data shows a massive, systematic directional bias. Almost no one guesses perfectly on the y=x line (except for Baby One More Time, which acts as a perfect cultural anchor).

Instead, human memory compresses everything toward ~2003. Older songs (Video Killed The Radio Star) get pulled forward. Newer songs (Get Lucky) get dragged backward.

**The Psychology (Why this happens):**

Looking at the data, this seems to be driven by Temporal Anchoring and the Reminiscence Bump. People hear a song, anchor it to „the 2000s“ because that’s when their own musical tastes crystallized, and then adjust poorly.

It creates this massive „Dunning-Kruger sweet spot.“ People are incredibly confident they know the exact year, but the data shows they are almost always off by exactly 1 to 3 years.

I update the test with 5 new songs every day at the link above. I’d love to know if the data nerds here can actually escape the gravity well, or if your brains compress the timeline too.

I tried and the average of the presented songs were 2000-2005. is this compensated for in the statistics?

I would guess this is more an artifact of the data gathering method. Any way to normalize by age of participant?

The people who are more likely to try out a website where you guess the ages of songs would likely cluster around a certain age range, meaning they like music from a certain era, so they are more likely to guess that era.