ADC Sample rate for e-bike power meter

[Axk]

What current sample rate is enough for accurate (e.g. 5% accuracy) consumed power measurement given the current waveform on the screenshots attached?

I've recorded the waveform with no load applied to the wheel accelerating from 0 to full throttle. Used a thin wire as the shunt with the voltage across the shunt amplified with an om-amp (100x if I remember correctly), not sure if this measurement setup is correct. I was not interested in absolute values, just the shape of the current waveform.

It looks like 1 Msps should be enough for this, but I wonder if I could get away with a lower rate of e.g. 100Ksps and a low-pass filter and still keep the accuracy within 5%.

[schmitt trigger]

Since there is so much noise, and the current ramp is quite slow (about 100 msecs), if you decide to do 1 Megasample/sec then you will require to significantly average the readings.
Like 1000 averages.

Another is to low-pass filter the input waveform prior to the ADC, and sample at a much slower rate.