In this OkCupid blog post they argued that the distribution of messages "dramatically illustrates just how much more important a woman’s looks are than a guy’s".

I transcribed this graph from Dataclysm into numbers and computed the share of message that go to the top 20 percentiles.

import numpy as np m = np.array([0.313, 0.322, 0.33 , 0.338, 0.347, 0.355, 0.363, 0.37 , 0.376,               0.383, 0.389, 0.395, 0.402, 0.408, 0.417, 0.427, 0.437, 0.446,               0.456, 0.466, 0.476, 0.486, 0.496, 0.506, 0.506, 0.506, 0.506,               0.506, 0.506, 0.506, 0.506, 0.506, 0.506, 0.506, 0.506, 0.515,               0.526, 0.536, 0.547, 0.557, 0.568, 0.578, 0.589, 0.599, 0.61 ,               0.62 , 0.631, 0.642, 0.652, 0.663, 0.673, 0.684, 0.694, 0.705,               0.715, 0.726, 0.736, 0.747, 0.757, 0.768, 0.778, 0.789, 0.801,               0.813, 0.825, 0.838, 0.85 , 0.862, 0.875, 0.887, 0.899, 0.912,               0.924, 0.939, 0.969, 1.   , 1.03 , 1.06 , 1.091, 1.121, 1.151,               1.182, 1.212, 1.239, 1.265, 1.291, 1.318, 1.344, 1.37 , 1.396,               1.454, 1.555, 1.656, 1.757, 1.858, 2.019, 2.205, 2.459, 3.125,               4.236])  f = np.array([ 0.973,  1.106,  1.24 ,  1.373,  1.506,  1.64 ,  1.773,  1.906,                2.021,  2.111,  2.201,  2.291,  2.381,  2.471,  2.561,  2.651,                2.741,  2.826,  2.906,  2.986,  3.066,  3.145,  3.225,  3.305,                3.385,  3.465,  3.544,  3.63 ,  3.732,  3.834,  3.935,  4.037,                4.139,  4.24 ,  4.342,  4.443,  4.545,  4.647,  4.748,  4.841,                4.927,  5.013,  5.099,  5.185,  5.271,  5.356,  5.442,  5.528,                5.614,  5.7  ,  5.786,  5.872,  5.958,  6.064,  6.178,  6.292,                6.406,  6.521,  6.635,  6.749,  6.863,  6.977,  7.091,  7.214,                7.358,  7.503,  7.647,  7.791,  7.935,  8.08 ,  8.224,  8.368,                8.513,  8.675,  8.844,  9.013,  9.182,  9.352,  9.521,  9.69 ,                9.859, 10.029, 10.203, 10.384, 10.566, 10.747, 10.928, 11.109,               11.291, 11.472, 11.827, 12.616, 13.405, 14.193, 14.982, 16.707,               18.432, 20.156, 22.818, 26.411])  m[80:].sum() / m.sum() # 0.4144903913161593 f[80:].sum() / f.sum() # 0.41054106074829816 m.mean() # 0.8466300000000001 f.mean() # 6.774839999999999 f.mean() / m.mean() # 8.002126076326139

So it is 41/20 for both sexes. No evidence that men preferentially target the most attractive members more than women at all!

We can also see that, on average, women receive 8 times as many messages as men, and even accounting for the slight surplus of males on OkCupid (54:46 M:F), it would still be (50/46) 0.85 = .92, (50/54) 6.77 = 6.27, so still 6.8 times as many messages.

I could create a similar graph to the one in the OkCupid blog by re-binning the data from Dataclysm with very weird bin edges [0, 1, 25, 50, 75, 95, 98, 100] and by dividing by the 1st percentile (dashed lines is the re-binned Dataclysm data):

https://i.imgur.com/jGwjwYm.png

It does not fit perfectly, so maybe they even used different bins for each sex.

So either…

  1. I'm misunderstanding something. (Not a scientist so that could be…)
  2. The graph in Datacylsm is wrong, or…
  3. The OkCupid blog was wrong. Maybe that's also why they deleted this article?