My spectrum analyzer program is working pretty well, so I tested it on some of the built-in MSP filters. The results were educational.
With the svf~ object, the "cutoff" frequency is actually where the frequency response crosses 0dB when resonance is set to zero. When resonance is very high the peak is actually at a lower frequency than cutoff. For example, when set to 80Hz the actual peak was at 78Hz.
With the onepole~ object, the "cutoff" frequency is the -3dB point of the filter.
With the lores~ object, the "cutoff" frequency is the frequency where resonance will create a spike. It doesn't represent any specific gain by itself.
I just figured the cutoff would be the -3dB point in all cases of zero resonance, but that didn't turn out to be the case at all.