My knowledge of how a red-black tree or how hidden markov models work or that a quicksort is n^2 worst case and n log n typically has saved me literally hours of searching. Over the course of my 15 year career.
Good lord, DADS + ACM + CiteSeer and a healthy dose of curiosity and intelligence is enough to get most people through 99% of the "hard" stuff you're going to see even in the era of Big Data.
I agree you can self-teach a lot of it, and a lot of it isn't needed in a lot of jobs, either. I somewhat disagree on the last part. I think there's a significant amount of work in Big-Data-related areas where not having a solid knowledge of statistics greatly increases the odds of doing something completely wrongly, or interpreting the results wrongly. On the other hand, there's a significant amount of work where deep knowledge of statistics is optional at most. But companies like Palantir do seem to put a strong emphasis on the statistical background of people they hire.
Good lord, DADS + ACM + CiteSeer and a healthy dose of curiosity and intelligence is enough to get most people through 99% of the "hard" stuff you're going to see even in the era of Big Data.