RChung

He appears not to believe in the "V" shape. He appears to believe that RR is flat, or nearly so, and then increases. Because he doesn't think there's a "V," he doesn't believe that the slope of the left part of the V is shallower and that the right part of the V is steeper. And I *think* his claim that resistance can decrease again is because that's what he's seen from his (poorly executed) tests. He doesn't believe in roller tests, and he doesn't believe in pendulum tests. [Edited to add:] He only believes in certain kinds of tests using power meters. He thinks that power meters introduce another layer of error that's uncontrolled so they make tests less reliable, not more. On those occasions when he's used power meters (such as laps he takes on the flat apron of a track) he tries to hold power and speed constant. I, of course, told him that since he has a power meter and speedometer and a way to record the data, he doesn't have to hold those things constant -- all he has to do is account for them in his calculations. He thinks that adds too much error.

So, he's been doing his tests based on roll downs but his measurement is based on a single data point, i.e., elapsed time. I haven't gotten him to explain how he's getting hundredths of seconds on his timing, while trying to control entry speed. I asked him once if he were using electric eyes triggered by the passing of his wheel. He never replied. I told him that (if he wasn't going to use a power meter) a better way to do his tests was instead of putting all of his analytical weight on a single measurement of elapsed time, to record speed at each second (so for a 15 second elapsed time test, he'd have 15 observations of speed). That way, he could spot anomalies both with the speed and with the accelerations. I also suggested that since he would have second-by-second data, he could improve his tests by starting at two (or more) different entry speeds. Then he could use the accelerations and he wouldn't have to control the speed. He said that would introduce the uncontrollable variable of CdA, and wind. I told him that measuring Crr and CdA was a feature, not a bug and the additional data would help him improve the sensitivity and robustness of his tests by helping to identify when the wind tainted his results. He asked me about the std error on my tests, and didn't believe me when I told him. I suggested he try a couple of different protocols but he has collected a lot of data and considers that part of his testing closed. So I've given up.