[asher_]

Earlier this year I was in the Philippines recruiting for a development office there. Although we weren't targeting grad programmers, the market dynamics we faced were likely similar. There are a whole bunch of substandard programmers that work there, and so using an online test as an initial filter proved very useful; there was no other way we could deal with the volume.

A local company we partnered with to help us recruit suggested https://codility.com/ which we ended up using.

We choose a pair of medium level challenges for the candidates. What we discovered as we went was that their score on the task was not the most important information we got from their test. How they wrote their code was. Looking at how they break up their solutions, and even things such as choices of - or consistency with - syntax was very informative and a good predictor of quality.

My advice would be to not set the bar too high for this kind of test, use it as a filter that eliminates the candidates you definitely don't want to interview, then proceed with the normal interview process.