Your formula is the definition of a Watt. But that doesn't mean that you use that formula for a motor. You don't connect a wire to it, and have it lift x kilos of weight in y seconds. You can jump along and directly measure rpm and torque. Because torque is also defined in the same units. Amount of Newton and the length of the arm in meters. So torque times revolution is also representing that same Watt.

So how does it matter what your car weights when you put it in a wind tunnel? Except that the weight will affect when it blows away...

Weight is almost irrelevant as long as you don't change height. Energy spent on acceleration is also mostly irrelevant for road racing. It would matter more for some types of velodrome racing.

It's the amount of air (who happens to have a weight) to push away that matters more. And you can't get even close to that with your speedometer or an accelerometer or the weight of the rider and bike.

And the drag factor is often >> 50% of the total power. You can't compute any drag factor for the simple reason that it gets reduced with 33% just because you happen to find someone else to ride behind. You need a sensor for it, since it is the #1 factor in consumed power when you don't significantly change the height. And such a sensor is not easy to make since a such a sensor must cover a significant percentage of the perpendicular area of the bike. The best way found to this day have been to actually use the real bike+rider as the sensor. So the approximations done are to ignore a number of other friction sources and specifically measure the amount of power/torque needed to maintain the bike speed.

Climbing the 21 turns of l'Alp d'Huez, you can ignore the wind and concentrate on weight and speed. Or weight and height. But on the flat, it's the wind that is the monster. And even a very marginal guestimate error will give you a huge error in your power estimate. You get a 100W error in the blink of an eye. It really scales that fast because of the evil square. If you do run alone at 300W and get behind someone, you save 100W. Just from having a wheel to follow.

The commercial companies would go for "the simplest" if it was possible. They have already found it isn't.

So they either measure torque. Or they measure the tension of the chain. Or they specifically use wind sensors. Call them incompetent or wrong, but then also prove them wrong by showing a real solution that works. They haven't managed...

The amount of force applied to a pedal do not vary so much with number of sprockets. Why? Because when it goes uphill you select a lighter gear but you still push what you can. Because the other riders will. So you stay at almost same rpm on the pedals. And almost the same force. It's only the torque at the backwheel that will be different since it's on the other side of the gear system.

Real systems measuring torque is just because that is the simplest measurements. It's the only type of measurements where we do have all knowledge covered. The only (!) issue is to get the measurement from a rotating section of the bike. We do not have the technology to perform the wind analysis. We would need to film the rider from all 360 degrees to try to establish cross section. Obviously not just from profile since the wind drag isn't just a question of area but also the Cw. When biking, we would then need yet another camera that films forward, to figure out if the rider is alone or partially covered by someone else. Nothing simple there. And a 10% error in the guess is a 10% error in the computed wind drag. Potentially a 8% error in total power. I don't think you can get even near an error guestimate of just 10%. Just switching from a tight bike jersey to a rain jacket with flapping cloth makes a huge difference in drag with almost no change in area. How do you guestimate that?

Think about all wind tunnel tests. They are done because it's many times too hard to compute the air resistance even when the model is 100% known. You aren't even close to that with a bike rider. And you don't know amount of wind hitting the rider.