I do not object to the broad points being made here but I do object to this interpretation. It's beyond awkward for many good valuable programmers, stifling really, to do this kind of unrealistic audition programming for a live audience. You are not just under stress -- you are far, far, far outside of any demonstration of your true value to the employer.
Perhaps the hiring pool really is that bad, you have a giant pile of unqualified applicants, and you really must use this kind of blunt instrument rather than taking the time to talk with candidates and their references and really get to know their work and their reputations. I've never had to do hiring and I don't envy you if you do. But I've also never in my life had to do this kind of test, and I think I'd use it as a filter myself to avoid working for someone who uses it.
[EDIT: I assumed it would be clear that I'm not talking about the coding challenge of FizzBuzz. I'm talking about the experience of audition-style programming. You end up filtering on less important characteristics of applicants who might otherwise be your best candidates, but think you're filtering on FizzBuzz.]
It's a for-loop, some conditionals, and printing. If you can't produce this in a few minutes, you can not program.
The point isn't to make a perfect, working fizzbuzz in the first try, the point is to find out if your applicant can program at all in the first place.
Ridiculously enough, there are people that can't program that would pass a normal interview with flying colours, but who utterly fail at fizzbuzz, which makes it a very valuable tool.
Programming on paper opposite an examiner with a stopwatch is indeed unrealistic. I don't disagree that FizzBuzz is crazy simple as a programming challenge.
I agree as well, which is why it's very important that you don't do that as an interviewer. The important thing is to check if the applicant can write a little bit of code from scratch, not if they can do it on time, in front of a big audience, without errors, with extra stress added on top.
And it sounds like the OP did it in an ok way, and didn't scare away any great programmers, which makes his conclusion pretty valid, that 60% of his applicants simply weren't capable.
I do not object to the broad points being made here but I do object to this interpretation. It's beyond awkward for many good valuable programmers, stifling really, to do this kind of unrealistic audition programming for a live audience. You are not just under stress -- you are far, far, far outside of any demonstration of your true value to the employer.
Perhaps the hiring pool really is that bad, you have a giant pile of unqualified applicants, and you really must use this kind of blunt instrument rather than taking the time to talk with candidates and their references and really get to know their work and their reputations. I've never had to do hiring and I don't envy you if you do. But I've also never in my life had to do this kind of test, and I think I'd use it as a filter myself to avoid working for someone who uses it.
[EDIT: I assumed it would be clear that I'm not talking about the coding challenge of FizzBuzz. I'm talking about the experience of audition-style programming. You end up filtering on less important characteristics of applicants who might otherwise be your best candidates, but think you're filtering on FizzBuzz.]