I don't see how Strava could accurately estimate power consistently over time, like many previously wrote here, if without a strain gauge there are just to many variables that are not part of the data collection in Strava.