In general, standardized testing based on CS trivia is a good gauge of how well the candidate has prepared for the interview. To put it bluntly, it tests how many piss boring problems the candidate is willing to put herself through to get into your company.

If the problem is uninteresting and tedious, it's likely that your candidate simply doesn't _want_ to solve the problem; to accurately judge whether she is _able_ to solve a problem would require giving her a problem she feels like solving in the first place. Yes, some candidates do get nervous and break; you might want to filter out such candidates on the basis that they can't handle "pressure".

What I don't understand is how you minimize risk: picking a rotten candidate who has prepared thoroughly for the specific interview, and leaving out good engineers is an expensive mistake. At best, it probably works heuristically for freshers because she's likely to retain atleast some amount of the toy exercises she was made to do as part of her coursework.

I've personally seen too many cases of "gamification" of the standardized testing process, and am disappointed that companies haven't been able to come up with a better system.

For the record: this is not sour grapes. I have no desire to work on the proprietary codebase of any company, and am quite happy just doing what I want (contributing to real open source projects). My comment is just a general remark on the state of affairs.