The basic premise of take-home test is that it is meant to entirely eliminate the bottom percentile of your candidate pool at the cost of also eliminating a chunk of the top percentile (which is supposed to be an acceptable trade off, especially if you are hiring for junior to mid roles). But I found it does not eliminate bad programmers at all. In fact, after we did some in-person interviewing we discovered a negative correlation between take-home test score and actual coding ability. Most likely explanation is that those with really good scores are the ones who paid 50 bucks to some comp-sci prodigy from <random_country> to sit with them in screen share session for an hour. In contrast, those with poor scores are the ones who actually tried to do the test honestly. Another unintended consequence here is that your candidate pool now has a higher percentage of unscupulous people than it did before the test. Our conclusion in the end was that we still have to test the coding ability in person, which means take-home test does not really eliminate any work for us as part of the interview process, but imposes additional costs as per above.