I'm not convinced that ratings from TopCoder's algorithm competitions are an accurate predictor of a participant's programming ability, except perhaps at the extremely low end.

This was me nine years ago, fresh out of college with a B.Sc. in Computer Science from a small state university:

http://www.topcoder.com/tc?module=MemberProfile&cr=27128...

A rating of 1529 or 1679 is "above average" (click the "Rating Distribution" button above the graph), but to be honest, I was a complete crap programmer at the time. I doubt I would have survived as a programmer (or even been considered for a programming position) at Google or similar back then. The same holds true now.

I hope that either the ratings distribution is artificially skewed towards the low end (maybe all the good programmers are too busy making lots of money to bother with TopCoder?) or the ratings really don't correspond with programming ability.