I have the same experience as Bob and J.A. with Google Earth, though I use it only when I am faced with a course design that appears to be a "tight fit" with the available real estate. Going to the highest magnification and clicking as often as needed to replicate a calibration ride often proves to be close to the certified distance when I compare after the road measurement, as J.A. experiences. If you have used GE this way, you know that this is a tedious procedure, and even this level of exactitude does not approach a certification ride for accuracy, as J.A. states.
GMaps does a good job of estimating where you have some real estate to work with, in my experience. I use a combination of the "For Runners" and "Straight Line" measurement tools in GMaps to get a facsimile of the course to be measured. The more turns in the course, the less accurate I find GMaps (or any on line tool) to be - overstating the course distance, naturally. My brief trial using MapMyRun was so unsatisfactory that I stopped using it years ago.
Mark, I have not tried MapMyRun in the last couple of years, but like most of us, I suspect, I sometimes get a MMR of the course I am asked to measure. Based on my experience with MMR, I have no difficulty saying "MMR is not accurate enough to display your course on your race web site" or something like that. I think to say otherwise could be misinterpreted to imply that RDs can just use MMR and forget about certification. While I suppose you or I or any of us reading this could create a MMR that might be close, I suspect we all agree MMR will never supplant certification measurements.
Some MMR courses I get turn out to be laughably short. I do not know how others of us do this, but I always take the MMR, pencil drawing, turn-by-turns or whatever I am given by my client, and then create a GMap of the course, showing what I believe will achieve the certified distance. I send it to the client for confirmation. Sometimes the client is surprised at the amount of distance that is likely to be added to their course.
Mark, you educated me on the fact some or most new GPS watches are accurate, and that much more of the error in course measurement by GPS device-wearing runners is typically due the lack of tangent running than GPS inaccuracy. However, I break with you about educating anyone about how to use MMR. I see no value in showing anyone how to use MMR or any on line tool for course measurement. Maybe you have had a different experience with MMR. For me, it is good enough that they can just rough out the desired path so I get the gist. I agree with the state certifier Pete talked to who stated that "...MMR is good enough (only) for training runs and course layout".