Calculations still don't add up: the guy who put out the most power is probably the heaviest but not twice as heavy as me - the effect of weight should be linear - and the guy in the middle is about my weight I'd guess?
It won't be linear proportional to weight, the majority of the power created goes to overcoming aerodynamic drag which is (roughly) proportional to somewhere between the square and cube roots of the weight[1], but that's an approximation as it's related to the frontal area and drag coefficient which you can only estimate using rider weight/height.
1. That is (m=mass and CdA is effective drag area) CdA ~ m^x where 1/3 < x < 1/2
Shorter answer is that Strava's power calculator is, as you can see, highly flawed.
They claim their power figures will almost match those from a power meter but that ignores the fact that those rides where real power measurements are obtained:-
* One ride could have been into a headwind.
* Or a roaring tailwind
* One could have been part of a group ride with the rider buried in the group.
etc, etc.
Yet their algorithm which ignores wind and the like will magically get very close to the real numbers. Putting any differences down to transmission losses is just taking the piss.
The biggest variance in their formula looks to be the P(acceleration) component. So jittery GPS data will make the algorithm think the rider was accelerating and decelerating often and, therefore, requiring more power.