- Published on
Clustering Says I Give Off a #1D2D8E Vibe
30 friends chose 30 colors they associate with me.
I'll keep this short, so you can spend more time playing with the colors.
Part 1: Why
One day, I read an article about how religion and social background influence people's perceptions of color. I wondered what color came to mind when my friends thought of me.
So I asked 30 friends I'd spent time with over the previous two years to each choose a color. I plotted their choices in the OKLab color space to look at their distances and distribution (Fig. 1).
Figure 1. Distribution of the 30 colors in OKLab. L represents lightness, while a and b describe the green–red and blue–yellow directions; distances between colors approximate their visual differences. Wireframes outline the clusters, tetrahedra mark their centers, and translucent sample points are unassigned colors. The faint background shows the sRGB gamut. Have fun exploring the colors.
Part 2: Analysis
I grouped the colors in OKLab using DBSCAN, choosing eps = 0.12 and min_samples = 3 after testing several settings.
This produced four clusters and five unassigned colors. I averaged the OKLab coordinates of the 13 colors in the largest cluster and converted the result to sRGB: #1D2D8E.
Somehow, this blue feels like a reflection of how my friends see me. ChatGPT described it as a deep, highly saturated blue with a hint of violet, calling it Deep Royal Blue and associating it with rationality, calm, and order. I wasn’t going to argue with that, so I decided to use it as this website's accent color.
Part 3: On this site
For links and buttons, I use this blue and two slightly darker shades:
| Color | Use |
|---|---|
| #1D2D8E | Main accent |
| #192679 | On hover |
| #141F63 | When pressed |
This blue carries a trace of how my friends see me.
Part 4: Data
Here are all 30 choices, sorted alphabetically by the names shown. Thank you, everyone.
| Name | Color |
|---|---|
| Fangying Xiang | #6082B6 |
| Fenghua Meng | #CFF536 |
| Hanming Hu | #6F2DA8 |
| Haotian Zheng | #FB7DA8 |
| Hidetoshi Kochi | #003153 |
| Hongling Ding | #BAE100 |
| Jiaxing Ye | #FFF5EE |
| Jiaying Gu | #A7E3EB |
| Jingchen Gong | #3457D5 |
| Liyang Fang | #4D371F |
| Mingchen Wang | #000080 |
| Pinyi Wu | #81D8D0 |
| Qiyue Wang | #5D8AA8 |
| Rui Chen | #6495ED |
| Shihao Zhong | #89CFF0 |
| Shulei He | #FF847C |
| Shuning Dong | #6300F2 |
| Tianran Chen | #E56334 |
| Xihe Yang | #010080 |
| Yang Zhong | #00F5FF |
| Yaqi Zhang | #270B3A |
| Ye Zhou | #191970 |
| Yifan Hou | #002FA7 |
| Yinren Zhao | #002395 |
| Yuying Xun | #225395 |
| Zhenyayuan Cao | #A7C7E7 |
| Zhicheng Yang | #00416A |
| Zhiyuan Zhong | #00FFFF |
| Ziyang Lu | #1F51FF |
| Ziyao Geng | #1F0954 |
Full sample data (CSV) contains all 30 names and chosen colors, together with their sRGB, OKLab and OKLCH coordinates and cluster assignments. This is the table to use if you'd like to explore the colors yourself.
Parameter tests (CSV) compares the three DBSCAN settings I tested, including their parameters, cluster sizes and the number of unassigned colors. The plot above uses the medium setting (eps = 0.12, min_samples = 3).