We honestly haven't hired enough engineers to make any sort of statistical measure on it but anecdotally it seems to be effective. We've yet to end up with a false positive and I know that we've hired at least two engineers who are awesome that I don't think we would have hired based solely on an "algorithms at the whiteboard" interview. Plus our engineer-time-invested per declined candidate is way down. Rather than spending 60 - 90 minutes watching a candidate flounder at the whiteboard we can axe them in a 2 minute code review.
