OKCupid's own data showed that women rate men is ugly but message them anyway.
Tinder's data shows "chance of approach", which is brutal for men who are not too good looking.
IIRC, OkCupid's data shows a bit different thing. The distribution of ratings determines chance of messages/likes - the larger stddev the better chances. IIRC, the most messaged men were ones with ratings characterized by bimodal distribution. Therefore, the most messaged men were not the ones with highest visual rating. I do not recall the distribution of messages, so can't comment much on that.
[0]http://web.archive.org/web/20100725135309/http://blog.okcupi...
You try to improve these things. Going to the gym etc, or focusing on one's career is a start.
What it didn't seem to account for was if you rated someone high enough (I can't remember if it required 4 or 5 stars), OKCupid would notify the person in question that you had done so. If you didn't want these notifications being fired off you had to adjust your ratings accordingly - now 3 stars is the highest rating you are going to give.
I think the stats really just ended up reflecting that men were generally more comfortable having their ratings broadcast than women were.