Back when I was still at university I was on a hiring panel for a new prof.
In a first step we ranked all the applicants using qualitative and quantitative metrics (the rankings across qualitative/quantitative and people on the panel was very similar when we compared rankings at the end) and then we handed in a shortlist with the best male applicant and best female applicant. We did this to make the Gleichstellungsbeauftragte (="diversity officer") shut up and thought that since the ability gap was really noticeable it'd be a no contest. How wrong we were. Despite being in the upper part of the bottom half of our ranking, she got the job.
That was the day I found out that I wasn't at a meritocratic institution as I had previously thought and left a few months later. That sort of thing really isn't good for your self confidence.
IIRC, Google did the same thing and concluded their hiring metrics had very little predictive value beyond setting a certain bar. Once you rose to that level, most discriminatory power of the metric evaporated.
If that was the case where you were, then is probably very little predictive value between "top scores" and "median scores".
http://www.pnas.org/content/112/17/5360.abstract
It should also be noted that the earlier study, which showed a smaller effect and only if the resumes were calibrated to be (a) mediocre and (b) precisely calibrated to be the same, was done in Psychology. Which is, of course, a field dominated by women and maybe not the typical "STEM" that we think about when we talk about the underrepresentation of women in tech.
Those two findings in turn gel with a recent French study that showed bias in grading in favor of the underrepresented gender in a particular field (though the bias was stronger in favor of women than in favor of men).
http://www.the-scientist.com/?articles.view/articleNo/46681/...
And of course there was the recent trial (not an experiment AFAICT) in Australia to reduce gender bias:
http://www.abc.net.au/news/2017-06-30/bilnd-recruitment-tria...
"The trial found assigning a male name to a candidate made them 3.2 per cent less likely to get a job interview. Adding a woman's name to a CV made the candidate 2.9 per cent more likely to get a foot in the door."
UPDATE: I was looking for one more thing, and finally found it. Despite the fact that the study showing bias against women is much weaker, it has been cited several times more often than the study showing bias for women (table near the bottom). https://www.psychologytoday.com/blog/rabble-rouser/201707/wh...