Brad Keller: 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
83.4%z 4.30 - 2
82.7%z 4.24 - 3
76.7%z 3.98 - 4
52.1%z 3.71 - 5
46.4%z 3.56 - 6
45.3%z 3.49 - 7
45%z 3.47 - 8
44.6%z 3.45 - 9
43.9%z 3.40 - 10
43.3%z 3.36 - 11
41.9%z 3.27 - 12
41.9%z 3.27 - 13
40.9%z 3.21 - 14
40.5%z 3.18 - 15
40.2%z 3.16 - 16
40.2%z 3.16 - 17
39.9%z 3.14 - 18
39.9%z 3.14 - 19
39.3%z 3.10 - 20
39.2%z 3.10 - 21
39.2%z 3.09 - 22
38.9%z 3.08 - 23
38.7%z 3.06 - 24
38.6%z 3.06 - 25
38.1%z 3.03 - 26
37.9%z 3.02 - 27
37.9%z 3.01 - 28
37.9%z 3.01 - 29
37.7%z 3.00 - 30
37.5%z 2.99 - 31
37.5%z 2.98 - 32
37.4%z 2.98 - 33
37%z 2.95 - 34
37%z 2.95 - 35
36.2%z 2.90 - 36
36.2%z 2.90 - 37
36.1%z 2.89 - 38
35.7%z 2.87 - 39
35.6%z 2.86 - 40
35.4%z 2.85 - 41
35.4%z 2.85 - 42
35.3%z 2.84 - 43
35.2%z 2.84 - 44
34.8%z 2.81 - 45
34.6%z 2.80 - 46
34.3%z 2.78 - 47
34.2%z 2.77 - 48
34%z 2.76 - 49
34%z 2.76 - 50
33.6%z 2.74