Yes, with all the sources of variability such an experiment is going to contend with, the observed difference between tires is likely to be swamped by random variation. If the OP does, say, five trials with tire A and five trials with tire B, he can then, by knowing the standard errors of the two means, calculate the likelihood of observing such a difference due to chance if there was, truly, no difference between the tires. You would also need to specify how large a difference you would be willing to miss (by erroneously concluding that there was no difference when in fact there was.) One typically finds that the number of trials needed to exclude a real difference turns out to be disappointingly large, much more than the number 5 which I just pulled out of a hat, especially if the difference you are seeking to discover is small, as is likely the case here.

A practical difficulty with using a bike computer to do coast-down measurements is that the speedometer has a lag during which it shows the speed over the last several wheel revolutions, (or measurements to the satellites if a GPS) not the truly instantaneous speed. (If you put the bike in a stand, spin the wheel till it shows a speed, then stop it instantly it will take some seconds to drop to zero.) For consistency, you would need to be watching the speedometer constantly and start/stop the watch the instant you saw the number cross the two threshold speeds -- dangerous to do while coasting downhill! Really the stoker should watch the speedometer and work the stopwatch while you watch the road.

Edit:
How do you measure distance rolled to 4 significant digits?

Good for you for doing this!