Isn't the spirit level method independent of any slope? The bubble is oriented to the local gravity vector, which should be the same at beginning and end of the rollout...?
it is, but the tangent angle of the tyre contact may be different at one full turn of the cranks (meaning that the distance travelled by the wheel centres is not quite the same as the distance between the start and finish tyre contact points..... and if the ground is ripply the distance travelled won't be quite right for this reason either. The latter effect (alone) will always result in the gear appearing slightly smaller than it really is, but may be overwhelmed by the former effect which can introduce an error in either direction.
Yeah but 'cranks horizontal according to the level' to 'cranks horizontal according to the level' is also a 360° rotation, no?
yup, it is, hence a slight slope is OK but ripplyness might not be.
In reality everyone would be judged by the same yardstick (ripply or not) at the event and what you would keep schtum about is that actually you have a small leeway so that you don't accidentally disqualify gears that are actually spot on but that you measured wrongly. You would be able to get the accuracy much improved vs traditional rollout test, anyway, quite easily.
When riding/loaded the wheel doesn't travel exactly the distance suggested by the circumference of the tyre (because the tyre deforms in a way that doesn't conserve tread length ) and indeed the deformation will vary with the tread thickness and carcass construction too.
Under torque load there is always some slippage of the tyre too, so the harder you pedal, the smaller the gear gets....
cheers