Per current standards, amplifier power are tested using a single tone sine wave. This means the peak voltage of the signal is 1.41 times (3 dB) higher than the RMS voltage. This ratio of peak to RMS voltage is the crest factor. Therefore, if your amp can output a maximum voltage swing of +/-50 volts, its "power rating" at 8 ohm would be: (50/1.41)^2 / 8 = 157 W. (The 1.41 number in the calculation is the crest factor.)
[Note that the multi-tone is a steady signal, i.e. crest factor is different and separate from the dynamic range of the signal.]
However, music (and multi-tones) is not a single tone sinewave, and has a much higher crest factor. If we choose a more realistic value of 12 dB (= 4x), the power at which clipping starts will be: (50/4)^2 / 8 = 20 W.
I agree with
@March Audio that the FTC test method is nowhere representative of the 99.9% use case. Nobody (sane and with intact hearing) drives their amplifer continuously to their max power and listen to single tone sine waves. Why should amplifiers be designed to this requirement?