Mark Brandl: raw computed results
The raw ranking from the production similarity calculation (face-embedding centroid cosine → gender-cohort z-normalization). The web-opinion crown, curation, and fusion re-rank are removed.
- 1
85.5%z 4.47 - 2
84.1%z 4.35 - 3
71.2%z 3.92 - 4
55.9%z 3.75 - 5
54.2%z 3.73 - 6
53.1%z 3.72 - 7
50.4%z 3.69 - 8
47%z 3.60 - 9
46.2%z 3.55 - 10
46.1%z 3.54 - 11
45.5%z 3.51 - 12
44.4%z 3.43 - 13
44.4%z 3.43 - 14
43.7%z 3.39 - 15
43.6%z 3.38 - 16
42.8%z 3.33 - 17
42.7%z 3.32 - 18
42.3%z 3.30 - 19
42.1%z 3.29 - 20
41.9%z 3.27 - 21
41.6%z 3.25 - 22
41.3%z 3.23 - 23
40.4%z 3.18 - 24
39.5%z 3.11 - 25
39.2%z 3.10 - 26
39.2%z 3.09 - 27
38.9%z 3.08 - 28
38.8%z 3.07 - 29
38.6%z 3.06 - 30
38.4%z 3.04 - 31
38.2%z 3.03 - 32
37.3%z 2.97 - 33
37.3%z 2.97 - 34
37.1%z 2.96 - 35
37%z 2.95 - 36
36.7%z 2.93 - 37
36.3%z 2.91 - 38
36%z 2.89 - 39
35.8%z 2.88 - 40
35.8%z 2.87 - 41
35.8%z 2.87 - 42
35.4%z 2.85 - 43
35.3%z 2.85 - 44
35.3%z 2.84 - 45
35.2%z 2.84 - 46
35.2%z 2.83 - 47
34.8%z 2.81 - 48
34.6%z 2.80 - 49
34.5%z 2.79 - 50
34.3%z 2.78