While not without value, you have to realize that the HQV test is to a somewhat large degree a marketing tool. In essence it was developed by Silicon Optix/Realta to show off the things their HQV video processors do well. For example a substantial portion of the score comes from handling odd cadences.
How accurate is it? Well when I played with the original DVD version, while it did rely on the user making determinations by eye, there were either objective marks to score from, or example of how things should be scored, so it's better than just subjective opinion.
Deinterlacing (video mode, film mode): Both are good
Chroma upscaling: Both are good or AMD is better
Image upscalng: Both are poor.
Black and white output range: AMD supports both full-range and limited-range, NVIDIA supports only limited-range (unless you hack the registry).
Refresh rate: AMD is near perfect, NVIDIA requires creating custom refresh rates.
Decoder: Both are equivalent (NVIDIA supports decoding 4K H.264 up to 30fps)
HQV Benchmark includes several other post-processing tests (denoise, edge sharpening, dynamic contrast enhancement etc.), I say both are good, but I would never turn them on.